Revert "[DAGCombine] Make sure combined nodes are added back to the worklist in topological order."
This reverts commit e69fa03ddd85812be3143d79a0359c3e8d43bd45.
This patch lead to build time outs on the AMDGPU OpenMP runtime
buildbot.
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 0579f50..082fd10 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -1814,11 +1814,11 @@
// Add any operands of the new node which have not yet been combined to the
// worklist as well. Because the worklist uniques things already, this
// won't repeatedly process the same operand.
+ CombinedNodes.insert(N);
for (const SDValue &ChildN : N->op_values())
if (!CombinedNodes.count(ChildN.getNode()))
AddToWorklist(ChildN.getNode());
- CombinedNodes.insert(N);
SDValue RV = combine(N);
if (!RV.getNode())
@@ -1852,8 +1852,10 @@
// out), because re-visiting the EntryToken and its users will not uncover
// any additional opportunities, but there may be a large number of such
// users, potentially causing compile time explosion.
- if (RV.getOpcode() != ISD::EntryToken)
- AddToWorklistWithUsers(RV.getNode());
+ if (RV.getOpcode() != ISD::EntryToken) {
+ AddToWorklist(RV.getNode());
+ AddUsersToWorklist(RV.getNode());
+ }
// Finally, if the node is now dead, remove it from the graph. The node
// may not be dead if the replacement process recursively simplified to
diff --git a/llvm/test/CodeGen/AMDGPU/combine-cond-add-sub.ll b/llvm/test/CodeGen/AMDGPU/combine-cond-add-sub.ll
index 27a3117..1ee9ed2 100644
--- a/llvm/test/CodeGen/AMDGPU/combine-cond-add-sub.ll
+++ b/llvm/test/CodeGen/AMDGPU/combine-cond-add-sub.ll
@@ -268,9 +268,10 @@
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: buffer_load_dword v4, v[2:3], s[4:7], 0 addr64
; GCN-NEXT: v_cmp_gt_u32_e32 vcc, v0, v1
+; GCN-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc
; GCN-NEXT: s_waitcnt vmcnt(0)
-; GCN-NEXT: v_subbrev_u32_e32 v0, vcc, 0, v4, vcc
-; GCN-NEXT: v_sub_i32_e32 v0, vcc, s0, v0
+; GCN-NEXT: v_sub_i32_e32 v0, vcc, v0, v4
+; GCN-NEXT: v_add_i32_e32 v0, vcc, s0, v0
; GCN-NEXT: v_add_i32_e32 v0, vcc, 0x64, v0
; GCN-NEXT: buffer_store_dword v0, v[2:3], s[4:7], 0 addr64
; GCN-NEXT: s_endpgm
@@ -281,11 +282,12 @@
; GFX9-NEXT: s_load_dword s4, s[0:1], 0x2c
; GFX9-NEXT: v_lshlrev_b32_e32 v2, 2, v0
; GFX9-NEXT: v_cmp_gt_u32_e32 vcc, v0, v1
+; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: global_load_dword v3, v2, s[2:3]
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_subbrev_co_u32_e32 v0, vcc, 0, v3, vcc
-; GFX9-NEXT: v_sub_u32_e32 v0, s4, v0
+; GFX9-NEXT: v_sub_u32_e32 v0, v0, v3
+; GFX9-NEXT: v_add_u32_e32 v0, s4, v0
; GFX9-NEXT: v_add_u32_e32 v0, 0x64, v0
; GFX9-NEXT: global_store_dword v2, v0, s[2:3]
; GFX9-NEXT: s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/dagcombine-setcc-select.ll b/llvm/test/CodeGen/AMDGPU/dagcombine-setcc-select.ll
index 8125094..3473682 100644
--- a/llvm/test/CodeGen/AMDGPU/dagcombine-setcc-select.ll
+++ b/llvm/test/CodeGen/AMDGPU/dagcombine-setcc-select.ll
@@ -23,8 +23,8 @@
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dword s0, s[0:1], 0x24
; GCN-NEXT: s_waitcnt lgkmcnt(0)
-; GCN-NEXT: v_cmp_nlt_f32_e64 s[0:1], s0, 1.0
-; GCN-NEXT: v_cndmask_b32_e64 v0, 2.0, 4.0, s[0:1]
+; GCN-NEXT: v_cmp_lt_f32_e64 s[0:1], s0, 1.0
+; GCN-NEXT: v_cndmask_b32_e64 v0, 4.0, 2.0, s[0:1]
; GCN-NEXT: flat_store_dword v[0:1], v0
; GCN-NEXT: s_endpgm
%c1 = fcmp olt float %x, 1.0
@@ -40,8 +40,8 @@
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dword s0, s[0:1], 0x24
; GCN-NEXT: s_waitcnt lgkmcnt(0)
-; GCN-NEXT: v_cmp_nlt_f32_e64 s[0:1], s0, 1.0
-; GCN-NEXT: v_cndmask_b32_e64 v0, 2.0, 4.0, s[0:1]
+; GCN-NEXT: v_cmp_lt_f32_e64 s[0:1], s0, 1.0
+; GCN-NEXT: v_cndmask_b32_e64 v0, 4.0, 2.0, s[0:1]
; GCN-NEXT: flat_store_dword v[0:1], v0
; GCN-NEXT: s_endpgm
%c1 = fcmp olt float %x, 1.0
diff --git a/llvm/test/CodeGen/AMDGPU/ds-alignment.ll b/llvm/test/CodeGen/AMDGPU/ds-alignment.ll
index 439ff32..c60da23 100644
--- a/llvm/test/CodeGen/AMDGPU/ds-alignment.ll
+++ b/llvm/test/CodeGen/AMDGPU/ds-alignment.ll
@@ -224,12 +224,12 @@
; ALIGNED-SDAG-NEXT: ds_write_b8 v7, v6 offset:5
; ALIGNED-SDAG-NEXT: ds_write_b8 v7, v1
; ALIGNED-SDAG-NEXT: ds_write_b8 v7, v2 offset:1
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v7, v8 offset:6
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v7, v0 offset:7
; ALIGNED-SDAG-NEXT: ds_write_b8 v7, v3 offset:2
; ALIGNED-SDAG-NEXT: ds_write_b8 v7, v4 offset:3
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(7)
-; ALIGNED-SDAG-NEXT: ds_write_b8 v7, v8 offset:6
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(7)
-; ALIGNED-SDAG-NEXT: ds_write_b8 v7, v0 offset:7
; ALIGNED-SDAG-NEXT: s_endpgm
;
; ALIGNED-GISEL-LABEL: ds8align1:
@@ -296,17 +296,17 @@
; ALIGNED-SDAG-NEXT: v_mov_b32_e32 v0, s0
; ALIGNED-SDAG-NEXT: ds_read_u16 v1, v0 offset:4
; ALIGNED-SDAG-NEXT: ds_read_u16 v2, v0
-; ALIGNED-SDAG-NEXT: ds_read_u16 v3, v0 offset:2
-; ALIGNED-SDAG-NEXT: ds_read_u16 v0, v0 offset:6
+; ALIGNED-SDAG-NEXT: ds_read_u16 v3, v0 offset:6
+; ALIGNED-SDAG-NEXT: ds_read_u16 v0, v0 offset:2
; ALIGNED-SDAG-NEXT: v_mov_b32_e32 v4, s1
; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
; ALIGNED-SDAG-NEXT: ds_write_b16 v4, v1 offset:4
; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
; ALIGNED-SDAG-NEXT: ds_write_b16 v4, v2
; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v4, v3 offset:2
+; ALIGNED-SDAG-NEXT: ds_write_b16 v4, v3 offset:6
; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v4, v0 offset:6
+; ALIGNED-SDAG-NEXT: ds_write_b16 v4, v0 offset:2
; ALIGNED-SDAG-NEXT: s_endpgm
;
; ALIGNED-GISEL-LABEL: ds8align2:
@@ -395,22 +395,24 @@
; ALIGNED-SDAG-NEXT: ds_read_u8 v11, v0 offset:10
; ALIGNED-SDAG-NEXT: ds_read_u8 v0, v0 offset:11
; ALIGNED-SDAG-NEXT: v_mov_b32_e32 v12, s1
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
-; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v9 offset:8
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
-; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v10 offset:9
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(7)
; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v5 offset:4
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(7)
; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v6 offset:5
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v9 offset:8
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v10 offset:9
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v11 offset:10
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(5)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v0 offset:11
; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v1
; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v2 offset:1
; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v3 offset:2
; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v4 offset:3
; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v7 offset:6
; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v8 offset:7
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(11)
-; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v11 offset:10
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(11)
-; ALIGNED-SDAG-NEXT: ds_write_b8 v12, v0 offset:11
; ALIGNED-SDAG-NEXT: s_endpgm
;
; ALIGNED-GISEL-LABEL: ds12align1:
@@ -492,23 +494,23 @@
; ALIGNED-SDAG-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(0)
; ALIGNED-SDAG-NEXT: v_mov_b32_e32 v0, s0
-; ALIGNED-SDAG-NEXT: ds_read_u16 v1, v0 offset:8
-; ALIGNED-SDAG-NEXT: ds_read_u16 v2, v0
-; ALIGNED-SDAG-NEXT: ds_read_u16 v3, v0 offset:2
-; ALIGNED-SDAG-NEXT: ds_read_u16 v4, v0 offset:4
-; ALIGNED-SDAG-NEXT: ds_read_u16 v5, v0 offset:6
+; ALIGNED-SDAG-NEXT: ds_read_u16 v1, v0
+; ALIGNED-SDAG-NEXT: ds_read_u16 v2, v0 offset:2
+; ALIGNED-SDAG-NEXT: ds_read_u16 v3, v0 offset:4
+; ALIGNED-SDAG-NEXT: ds_read_u16 v4, v0 offset:8
+; ALIGNED-SDAG-NEXT: ds_read_u16 v5, v0 offset:10
+; ALIGNED-SDAG-NEXT: ds_read_u16 v0, v0 offset:6
; ALIGNED-SDAG-NEXT: v_mov_b32_e32 v6, s1
-; ALIGNED-SDAG-NEXT: ds_read_u16 v0, v0 offset:10
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(5)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v1 offset:8
; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v4 offset:4
-; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v2
-; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v3 offset:2
+; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v3 offset:4
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
+; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v4 offset:8
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
+; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v5 offset:10
+; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v1
+; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v2 offset:2
; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(5)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v5 offset:6
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(5)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v0 offset:10
+; ALIGNED-SDAG-NEXT: ds_write_b16 v6, v0 offset:6
; ALIGNED-SDAG-NEXT: s_endpgm
;
; ALIGNED-GISEL-LABEL: ds12align2:
@@ -693,25 +695,30 @@
; ALIGNED-SDAG-NEXT: ds_read_u8 v15, v0 offset:14
; ALIGNED-SDAG-NEXT: ds_read_u8 v0, v0 offset:15
; ALIGNED-SDAG-NEXT: v_mov_b32_e32 v16, s1
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
-; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v13 offset:12
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(3)
-; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v14 offset:13
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(13)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v3 offset:2
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(13)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v4 offset:3
; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v1
; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v2 offset:1
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(13)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v7 offset:6
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(13)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v8 offset:7
; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v5 offset:4
; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v6 offset:5
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(13)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v11 offset:10
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(13)
+; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v12 offset:11
; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v9 offset:8
; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v10 offset:9
-; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v3 offset:2
-; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v4 offset:3
-; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v7 offset:6
-; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v8 offset:7
-; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v11 offset:10
-; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v12 offset:11
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(14)
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(13)
; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v15 offset:14
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(13)
; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v0 offset:15
+; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v13 offset:12
+; ALIGNED-SDAG-NEXT: ds_write_b8 v16, v14 offset:13
; ALIGNED-SDAG-NEXT: s_endpgm
;
; ALIGNED-GISEL-LABEL: ds16align1:
@@ -809,29 +816,27 @@
; ALIGNED-SDAG-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x24
; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(0)
; ALIGNED-SDAG-NEXT: v_mov_b32_e32 v0, s0
-; ALIGNED-SDAG-NEXT: ds_read_u16 v1, v0 offset:12
; ALIGNED-SDAG-NEXT: ds_read_u16 v2, v0
; ALIGNED-SDAG-NEXT: ds_read_u16 v3, v0 offset:2
; ALIGNED-SDAG-NEXT: ds_read_u16 v4, v0 offset:4
; ALIGNED-SDAG-NEXT: ds_read_u16 v5, v0 offset:6
; ALIGNED-SDAG-NEXT: ds_read_u16 v6, v0 offset:8
; ALIGNED-SDAG-NEXT: ds_read_u16 v7, v0 offset:10
-; ALIGNED-SDAG-NEXT: v_mov_b32_e32 v8, s1
+; ALIGNED-SDAG-NEXT: ds_read_u16 v8, v0 offset:12
; ALIGNED-SDAG-NEXT: ds_read_u16 v0, v0 offset:14
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(7)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v8, v1 offset:12
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(7)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v8, v2
+; ALIGNED-SDAG-NEXT: v_mov_b32_e32 v1, s1
; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(6)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v8, v4 offset:4
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(5)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v8, v6 offset:8
-; ALIGNED-SDAG-NEXT: ds_write_b16 v8, v3 offset:2
-; ALIGNED-SDAG-NEXT: ds_write_b16 v8, v5 offset:6
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(7)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v8, v7 offset:10
-; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(7)
-; ALIGNED-SDAG-NEXT: ds_write_b16 v8, v0 offset:14
+; ALIGNED-SDAG-NEXT: ds_write_b16 v1, v3 offset:2
+; ALIGNED-SDAG-NEXT: ds_write_b16 v1, v2
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(6)
+; ALIGNED-SDAG-NEXT: ds_write_b16 v1, v5 offset:6
+; ALIGNED-SDAG-NEXT: ds_write_b16 v1, v4 offset:4
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(6)
+; ALIGNED-SDAG-NEXT: ds_write_b16 v1, v7 offset:10
+; ALIGNED-SDAG-NEXT: ds_write_b16 v1, v6 offset:8
+; ALIGNED-SDAG-NEXT: s_waitcnt lgkmcnt(6)
+; ALIGNED-SDAG-NEXT: ds_write_b16 v1, v0 offset:14
+; ALIGNED-SDAG-NEXT: ds_write_b16 v1, v8 offset:12
; ALIGNED-SDAG-NEXT: s_endpgm
;
; ALIGNED-GISEL-LABEL: ds16align2:
diff --git a/llvm/test/CodeGen/AMDGPU/ds_write2.ll b/llvm/test/CodeGen/AMDGPU/ds_write2.ll
index 02abe69..db3dc48 100644
--- a/llvm/test/CodeGen/AMDGPU/ds_write2.ll
+++ b/llvm/test/CodeGen/AMDGPU/ds_write2.ll
@@ -657,20 +657,20 @@
; CI-NEXT: ds_write_b8 v0, v1 offset:5
; CI-NEXT: v_lshrrev_b32_e32 v4, 16, v1
; CI-NEXT: v_lshrrev_b32_e32 v5, 8, v1
-; CI-NEXT: ds_write_b8 v0, v1 offset:9
; CI-NEXT: ds_write_b8 v0, v2 offset:13
+; CI-NEXT: ds_write_b8 v0, v1 offset:9
; CI-NEXT: v_lshrrev_b32_e32 v1, 24, v2
; CI-NEXT: v_lshrrev_b32_e32 v6, 16, v2
; CI-NEXT: v_lshrrev_b32_e32 v2, 8, v2
; CI-NEXT: ds_write_b8 v0, v3 offset:8
; CI-NEXT: ds_write_b8 v0, v4 offset:7
; CI-NEXT: ds_write_b8 v0, v5 offset:6
-; CI-NEXT: ds_write_b8 v0, v3 offset:12
-; CI-NEXT: ds_write_b8 v0, v4 offset:11
-; CI-NEXT: ds_write_b8 v0, v5 offset:10
; CI-NEXT: ds_write_b8 v0, v1 offset:16
; CI-NEXT: ds_write_b8 v0, v6 offset:15
; CI-NEXT: ds_write_b8 v0, v2 offset:14
+; CI-NEXT: ds_write_b8 v0, v3 offset:12
+; CI-NEXT: ds_write_b8 v0, v4 offset:11
+; CI-NEXT: ds_write_b8 v0, v5 offset:10
; CI-NEXT: s_endpgm
;
; GFX9-ALIGNED-LABEL: unaligned_offset_simple_write2_one_val_f64:
@@ -686,18 +686,18 @@
; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v0 offset:5
; GFX9-ALIGNED-NEXT: v_lshrrev_b32_e32 v3, 24, v0
; GFX9-ALIGNED-NEXT: v_lshrrev_b32_e32 v4, 8, v0
-; GFX9-ALIGNED-NEXT: ds_write_b8_d16_hi v2, v0 offset:11
-; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v0 offset:9
; GFX9-ALIGNED-NEXT: ds_write_b8_d16_hi v2, v1 offset:15
; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v1 offset:13
+; GFX9-ALIGNED-NEXT: ds_write_b8_d16_hi v2, v0 offset:11
+; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v0 offset:9
; GFX9-ALIGNED-NEXT: v_lshrrev_b32_e32 v0, 24, v1
; GFX9-ALIGNED-NEXT: v_lshrrev_b32_e32 v1, 8, v1
; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v3 offset:8
; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v4 offset:6
-; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v3 offset:12
-; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v4 offset:10
; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v0 offset:16
; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v1 offset:14
+; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v3 offset:12
+; GFX9-ALIGNED-NEXT: ds_write_b8 v2, v4 offset:10
; GFX9-ALIGNED-NEXT: s_endpgm
;
; GFX9-UNALIGNED-LABEL: unaligned_offset_simple_write2_one_val_f64:
diff --git a/llvm/test/CodeGen/AMDGPU/idot4u.ll b/llvm/test/CodeGen/AMDGPU/idot4u.ll
index 8370f5a..5aad8c6 100644
--- a/llvm/test/CodeGen/AMDGPU/idot4u.ll
+++ b/llvm/test/CodeGen/AMDGPU/idot4u.ll
@@ -1851,14 +1851,14 @@
; GFX7-NEXT: buffer_load_ushort v1, off, s[0:3], 0
; GFX7-NEXT: s_waitcnt vmcnt(2)
; GFX7-NEXT: v_and_b32_e32 v3, 0xff00, v2
-; GFX7-NEXT: v_bfe_u32 v4, v2, 16, 8
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, 24, v2
; GFX7-NEXT: s_waitcnt vmcnt(1)
; GFX7-NEXT: v_and_b32_e32 v6, 0xff00, v0
-; GFX7-NEXT: v_lshrrev_b32_e32 v5, 24, v2
+; GFX7-NEXT: v_bfe_u32 v5, v2, 16, 8
; GFX7-NEXT: v_and_b32_e32 v2, 0xff, v2
; GFX7-NEXT: v_lshlrev_b32_e32 v3, 8, v3
-; GFX7-NEXT: v_bfe_u32 v7, v0, 16, 8
-; GFX7-NEXT: v_lshrrev_b32_e32 v8, 24, v0
+; GFX7-NEXT: v_lshrrev_b32_e32 v7, 24, v0
+; GFX7-NEXT: v_bfe_u32 v8, v0, 16, 8
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
; GFX7-NEXT: v_lshlrev_b32_e32 v6, 8, v6
; GFX7-NEXT: v_alignbit_b32 v3, s10, v3, 16
@@ -1866,8 +1866,8 @@
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_mad_u32_u24 v0, v2, v0, v1
; GFX7-NEXT: v_mad_u32_u24 v0, v3, v6, v0
-; GFX7-NEXT: v_mad_u32_u24 v0, v4, v7, v0
; GFX7-NEXT: v_mad_u32_u24 v0, v5, v8, v0
+; GFX7-NEXT: v_mad_u32_u24 v0, v4, v7, v0
; GFX7-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX7-NEXT: s_endpgm
;
diff --git a/llvm/test/CodeGen/AMDGPU/idot8s.ll b/llvm/test/CodeGen/AMDGPU/idot8s.ll
index 7dc006b..de59b32 100644
--- a/llvm/test/CodeGen/AMDGPU/idot8s.ll
+++ b/llvm/test/CodeGen/AMDGPU/idot8s.ll
@@ -2557,50 +2557,50 @@
; GFX7-NEXT: buffer_load_ubyte v1, off, s[0:3], 0
; GFX7-NEXT: s_addc_u32 s13, s13, 0
; GFX7-NEXT: s_waitcnt vmcnt(2)
-; GFX7-NEXT: v_bfe_i32 v7, v2, 0, 4
-; GFX7-NEXT: v_bfe_i32 v3, v2, 24, 4
+; GFX7-NEXT: v_bfe_i32 v8, v2, 0, 4
+; GFX7-NEXT: v_ashrrev_i32_e32 v3, 28, v2
; GFX7-NEXT: s_waitcnt vmcnt(1)
-; GFX7-NEXT: v_bfe_i32 v14, v0, 0, 4
-; GFX7-NEXT: v_bfe_i32 v4, v2, 20, 4
-; GFX7-NEXT: v_bfe_i32 v5, v2, 16, 4
-; GFX7-NEXT: v_bfe_i32 v6, v2, 8, 4
-; GFX7-NEXT: v_ashrrev_i32_e32 v8, 28, v2
+; GFX7-NEXT: v_bfe_i32 v15, v0, 0, 4
+; GFX7-NEXT: v_bfe_i32 v4, v2, 24, 4
+; GFX7-NEXT: v_bfe_i32 v5, v2, 20, 4
+; GFX7-NEXT: v_bfe_i32 v6, v2, 16, 4
+; GFX7-NEXT: v_bfe_i32 v7, v2, 8, 4
; GFX7-NEXT: v_bfe_i32 v9, v2, 12, 4
; GFX7-NEXT: v_bfe_i32 v2, v2, 4, 4
-; GFX7-NEXT: v_and_b32_e32 v7, 0xff, v7
-; GFX7-NEXT: v_bfe_i32 v10, v0, 24, 4
-; GFX7-NEXT: v_bfe_i32 v11, v0, 20, 4
-; GFX7-NEXT: v_bfe_i32 v12, v0, 16, 4
-; GFX7-NEXT: v_bfe_i32 v13, v0, 8, 4
-; GFX7-NEXT: v_ashrrev_i32_e32 v15, 28, v0
+; GFX7-NEXT: v_and_b32_e32 v8, 0xff, v8
+; GFX7-NEXT: v_ashrrev_i32_e32 v10, 28, v0
+; GFX7-NEXT: v_bfe_i32 v11, v0, 24, 4
+; GFX7-NEXT: v_bfe_i32 v12, v0, 20, 4
+; GFX7-NEXT: v_bfe_i32 v13, v0, 16, 4
+; GFX7-NEXT: v_bfe_i32 v14, v0, 8, 4
; GFX7-NEXT: v_bfe_i32 v16, v0, 12, 4
; GFX7-NEXT: v_bfe_i32 v0, v0, 4, 4
-; GFX7-NEXT: v_and_b32_e32 v14, 0xff, v14
+; GFX7-NEXT: v_and_b32_e32 v15, 0xff, v15
; GFX7-NEXT: v_and_b32_e32 v2, 0xff, v2
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_mad_u32_u24 v1, v7, v14, v1
-; GFX7-NEXT: v_and_b32_e32 v6, 0xff, v6
+; GFX7-NEXT: v_mad_u32_u24 v1, v8, v15, v1
+; GFX7-NEXT: v_and_b32_e32 v7, 0xff, v7
; GFX7-NEXT: v_lshlrev_b32_e32 v9, 24, v9
-; GFX7-NEXT: v_and_b32_e32 v13, 0xff, v13
+; GFX7-NEXT: v_and_b32_e32 v14, 0xff, v14
; GFX7-NEXT: v_lshlrev_b32_e32 v16, 24, v16
; GFX7-NEXT: v_mad_u32_u24 v0, v2, v0, v1
; GFX7-NEXT: v_alignbit_b32 v9, 0, v9, 24
; GFX7-NEXT: v_alignbit_b32 v16, 0, v16, 24
-; GFX7-NEXT: v_mad_u32_u24 v0, v6, v13, v0
+; GFX7-NEXT: v_mad_u32_u24 v0, v7, v14, v0
+; GFX7-NEXT: v_and_b32_e32 v6, 0xff, v6
+; GFX7-NEXT: v_and_b32_e32 v13, 0xff, v13
+; GFX7-NEXT: v_mad_u32_u24 v0, v9, v16, v0
; GFX7-NEXT: v_and_b32_e32 v5, 0xff, v5
; GFX7-NEXT: v_and_b32_e32 v12, 0xff, v12
-; GFX7-NEXT: v_mad_u32_u24 v0, v9, v16, v0
+; GFX7-NEXT: v_mad_u32_u24 v0, v6, v13, v0
; GFX7-NEXT: v_and_b32_e32 v4, 0xff, v4
; GFX7-NEXT: v_and_b32_e32 v11, 0xff, v11
; GFX7-NEXT: v_mad_u32_u24 v0, v5, v12, v0
; GFX7-NEXT: v_and_b32_e32 v3, 0xff, v3
; GFX7-NEXT: v_and_b32_e32 v10, 0xff, v10
; GFX7-NEXT: v_mad_u32_u24 v0, v4, v11, v0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xff, v8
-; GFX7-NEXT: v_and_b32_e32 v15, 0xff, v15
; GFX7-NEXT: v_mad_u32_u24 v0, v3, v10, v0
-; GFX7-NEXT: v_mad_u32_u24 v0, v8, v15, v0
; GFX7-NEXT: buffer_store_byte v0, off, s[0:3], 0
; GFX7-NEXT: s_endpgm
;
diff --git a/llvm/test/CodeGen/AMDGPU/idot8u.ll b/llvm/test/CodeGen/AMDGPU/idot8u.ll
index 01ee809..1c092fc 100644
--- a/llvm/test/CodeGen/AMDGPU/idot8u.ll
+++ b/llvm/test/CodeGen/AMDGPU/idot8u.ll
@@ -2445,35 +2445,35 @@
; GFX7-NEXT: s_addc_u32 s13, s13, 0
; GFX7-NEXT: s_waitcnt vmcnt(2)
; GFX7-NEXT: v_and_b32_e32 v8, 15, v2
-; GFX7-NEXT: v_bfe_u32 v7, v2, 4, 4
+; GFX7-NEXT: v_bfe_u32 v3, v2, 24, 4
; GFX7-NEXT: s_waitcnt vmcnt(1)
; GFX7-NEXT: v_and_b32_e32 v15, 15, v0
-; GFX7-NEXT: v_lshlrev_b32_e32 v9, 12, v2
+; GFX7-NEXT: v_bfe_u32 v4, v2, 20, 4
+; GFX7-NEXT: v_bfe_u32 v5, v2, 16, 4
+; GFX7-NEXT: v_bfe_u32 v6, v2, 8, 4
+; GFX7-NEXT: v_bfe_u32 v7, v2, 4, 4
+; GFX7-NEXT: v_lshrrev_b32_e32 v9, 28, v2
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 12, v2
+; GFX7-NEXT: v_bfe_u32 v10, v0, 24, 4
+; GFX7-NEXT: v_bfe_u32 v11, v0, 20, 4
+; GFX7-NEXT: v_bfe_u32 v12, v0, 16, 4
+; GFX7-NEXT: v_bfe_u32 v13, v0, 8, 4
; GFX7-NEXT: v_bfe_u32 v14, v0, 4, 4
-; GFX7-NEXT: v_lshlrev_b32_e32 v16, 12, v0
+; GFX7-NEXT: v_lshrrev_b32_e32 v16, 28, v0
+; GFX7-NEXT: v_lshlrev_b32_e32 v0, 12, v0
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_mad_u32_u24 v1, v8, v15, v1
-; GFX7-NEXT: v_bfe_u32 v6, v2, 8, 4
-; GFX7-NEXT: v_bfe_u32 v13, v0, 8, 4
-; GFX7-NEXT: v_and_b32_e32 v9, 0xf000000, v9
-; GFX7-NEXT: v_and_b32_e32 v16, 0xf000000, v16
+; GFX7-NEXT: v_and_b32_e32 v2, 0xf000000, v2
+; GFX7-NEXT: v_and_b32_e32 v0, 0xf000000, v0
; GFX7-NEXT: v_mad_u32_u24 v1, v7, v14, v1
-; GFX7-NEXT: v_alignbit_b32 v9, s10, v9, 24
-; GFX7-NEXT: v_alignbit_b32 v8, 0, v16, 24
+; GFX7-NEXT: v_alignbit_b32 v2, s10, v2, 24
+; GFX7-NEXT: v_alignbit_b32 v0, 0, v0, 24
; GFX7-NEXT: v_mad_u32_u24 v1, v6, v13, v1
-; GFX7-NEXT: v_bfe_u32 v5, v2, 16, 4
-; GFX7-NEXT: v_bfe_u32 v12, v0, 16, 4
-; GFX7-NEXT: v_mad_u32_u24 v1, v9, v8, v1
-; GFX7-NEXT: v_bfe_u32 v4, v2, 20, 4
-; GFX7-NEXT: v_bfe_u32 v11, v0, 20, 4
-; GFX7-NEXT: v_mad_u32_u24 v1, v5, v12, v1
-; GFX7-NEXT: v_bfe_u32 v3, v2, 24, 4
-; GFX7-NEXT: v_bfe_u32 v10, v0, 24, 4
-; GFX7-NEXT: v_mad_u32_u24 v1, v4, v11, v1
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 28, v2
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, 28, v0
-; GFX7-NEXT: v_mad_u32_u24 v1, v3, v10, v1
; GFX7-NEXT: v_mad_u32_u24 v0, v2, v0, v1
+; GFX7-NEXT: v_mad_u32_u24 v0, v5, v12, v0
+; GFX7-NEXT: v_mad_u32_u24 v0, v4, v11, v0
+; GFX7-NEXT: v_mad_u32_u24 v0, v3, v10, v0
+; GFX7-NEXT: v_mad_u32_u24 v0, v9, v16, v0
; GFX7-NEXT: buffer_store_byte v0, off, s[0:3], 0
; GFX7-NEXT: s_endpgm
;
diff --git a/llvm/test/CodeGen/AMDGPU/load-local-redundant-copies.ll b/llvm/test/CodeGen/AMDGPU/load-local-redundant-copies.ll
index 03ee5a6..f8a58fa 100644
--- a/llvm/test/CodeGen/AMDGPU/load-local-redundant-copies.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-local-redundant-copies.ll
@@ -30,22 +30,22 @@
define amdgpu_vs void @test_2(<4 x i32> inreg %arg1, i32 %arg2, i32 inreg %arg3, ptr addrspace(3) %arg4) {
; CHECK-LABEL: test_2:
; CHECK: ; %bb.0:
+; CHECK-NEXT: v_add_i32_e32 v5, vcc, 28, v1
+; CHECK-NEXT: v_add_i32_e32 v2, vcc, 24, v1
; CHECK-NEXT: v_add_i32_e32 v3, vcc, 20, v1
-; CHECK-NEXT: v_add_i32_e32 v2, vcc, 16, v1
-; CHECK-NEXT: v_add_i32_e32 v4, vcc, 28, v1
-; CHECK-NEXT: v_add_i32_e32 v6, vcc, 24, v1
+; CHECK-NEXT: v_add_i32_e32 v6, vcc, 16, v1
; CHECK-NEXT: v_add_i32_e32 v7, vcc, 12, v1
; CHECK-NEXT: v_add_i32_e32 v8, vcc, 8, v1
; CHECK-NEXT: v_add_i32_e32 v10, vcc, 4, v1
; CHECK-NEXT: s_mov_b32 m0, -1
-; CHECK-NEXT: ds_read_b32 v2, v2
-; CHECK-NEXT: ds_read_b32 v5, v4
-; CHECK-NEXT: ds_read_b32 v4, v6
+; CHECK-NEXT: ds_read_b32 v4, v2
+; CHECK-NEXT: ds_read_b32 v3, v3
+; CHECK-NEXT: ds_read_b32 v2, v6
; CHECK-NEXT: ds_read_b32 v9, v7
; CHECK-NEXT: ds_read_b32 v8, v8
; CHECK-NEXT: ds_read_b32 v7, v10
; CHECK-NEXT: ds_read_b32 v6, v1
-; CHECK-NEXT: ds_read_b32 v3, v3
+; CHECK-NEXT: ds_read_b32 v5, v5
; CHECK-NEXT: s_waitcnt lgkmcnt(1)
; CHECK-NEXT: tbuffer_store_format_xyzw v[6:9], v0, s[0:3], s4 format:[BUF_DATA_FORMAT_32_32_32,BUF_NUM_FORMAT_UINT] idxen glc slc
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
diff --git a/llvm/test/CodeGen/AMDGPU/store-local.128.ll b/llvm/test/CodeGen/AMDGPU/store-local.128.ll
index 01ad966..0663b96 100644
--- a/llvm/test/CodeGen/AMDGPU/store-local.128.ll
+++ b/llvm/test/CodeGen/AMDGPU/store-local.128.ll
@@ -83,42 +83,42 @@
; GFX9-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x10
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v0, s2
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: v_mov_b32_e32 v2, s6
-; GFX9-NEXT: ds_write_b8 v0, v1 offset:12
-; GFX9-NEXT: ds_write_b8_d16_hi v0, v1 offset:14
-; GFX9-NEXT: ds_write_b8 v0, v2 offset:8
-; GFX9-NEXT: ds_write_b8_d16_hi v0, v2 offset:10
-; GFX9-NEXT: v_mov_b32_e32 v1, s5
-; GFX9-NEXT: ds_write_b8 v0, v1 offset:4
-; GFX9-NEXT: ds_write_b8_d16_hi v0, v1 offset:6
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: v_mov_b32_e32 v2, s7
+; GFX9-NEXT: ds_write_b8 v0, v1 offset:8
+; GFX9-NEXT: ds_write_b8_d16_hi v0, v1 offset:10
+; GFX9-NEXT: ds_write_b8 v0, v2 offset:12
+; GFX9-NEXT: ds_write_b8_d16_hi v0, v2 offset:14
; GFX9-NEXT: v_mov_b32_e32 v1, s4
-; GFX9-NEXT: s_lshr_b32 s0, s7, 8
; GFX9-NEXT: ds_write_b8 v0, v1
; GFX9-NEXT: ds_write_b8_d16_hi v0, v1 offset:2
-; GFX9-NEXT: v_mov_b32_e32 v1, s0
-; GFX9-NEXT: s_lshr_b32 s0, s7, 24
-; GFX9-NEXT: ds_write_b8 v0, v1 offset:13
-; GFX9-NEXT: v_mov_b32_e32 v1, s0
+; GFX9-NEXT: v_mov_b32_e32 v1, s5
; GFX9-NEXT: s_lshr_b32 s0, s6, 8
-; GFX9-NEXT: ds_write_b8 v0, v1 offset:15
+; GFX9-NEXT: ds_write_b8 v0, v1 offset:4
+; GFX9-NEXT: ds_write_b8_d16_hi v0, v1 offset:6
; GFX9-NEXT: v_mov_b32_e32 v1, s0
; GFX9-NEXT: s_lshr_b32 s0, s6, 24
; GFX9-NEXT: ds_write_b8 v0, v1 offset:9
; GFX9-NEXT: v_mov_b32_e32 v1, s0
-; GFX9-NEXT: s_lshr_b32 s0, s5, 8
+; GFX9-NEXT: s_lshr_b32 s0, s7, 8
; GFX9-NEXT: ds_write_b8 v0, v1 offset:11
; GFX9-NEXT: v_mov_b32_e32 v1, s0
-; GFX9-NEXT: s_lshr_b32 s0, s5, 24
-; GFX9-NEXT: ds_write_b8 v0, v1 offset:5
+; GFX9-NEXT: s_lshr_b32 s0, s7, 24
+; GFX9-NEXT: ds_write_b8 v0, v1 offset:13
; GFX9-NEXT: v_mov_b32_e32 v1, s0
; GFX9-NEXT: s_lshr_b32 s0, s4, 8
-; GFX9-NEXT: ds_write_b8 v0, v1 offset:7
+; GFX9-NEXT: ds_write_b8 v0, v1 offset:15
; GFX9-NEXT: v_mov_b32_e32 v1, s0
; GFX9-NEXT: s_lshr_b32 s0, s4, 24
; GFX9-NEXT: ds_write_b8 v0, v1 offset:1
; GFX9-NEXT: v_mov_b32_e32 v1, s0
+; GFX9-NEXT: s_lshr_b32 s0, s5, 8
; GFX9-NEXT: ds_write_b8 v0, v1 offset:3
+; GFX9-NEXT: v_mov_b32_e32 v1, s0
+; GFX9-NEXT: s_lshr_b32 s0, s5, 24
+; GFX9-NEXT: ds_write_b8 v0, v1 offset:5
+; GFX9-NEXT: v_mov_b32_e32 v1, s0
+; GFX9-NEXT: ds_write_b8 v0, v1 offset:7
; GFX9-NEXT: s_endpgm
;
; GFX7-LABEL: store_lds_v4i32_align1:
@@ -128,50 +128,50 @@
; GFX7-NEXT: s_mov_b32 m0, -1
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v0, s4
-; GFX7-NEXT: v_mov_b32_e32 v1, s3
-; GFX7-NEXT: v_mov_b32_e32 v2, s2
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:12
-; GFX7-NEXT: ds_write_b8 v0, v2 offset:8
-; GFX7-NEXT: v_mov_b32_e32 v1, s1
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:4
+; GFX7-NEXT: v_mov_b32_e32 v1, s2
+; GFX7-NEXT: v_mov_b32_e32 v2, s3
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:8
+; GFX7-NEXT: ds_write_b8 v0, v2 offset:12
; GFX7-NEXT: v_mov_b32_e32 v1, s0
-; GFX7-NEXT: s_lshr_b32 s4, s3, 8
; GFX7-NEXT: ds_write_b8 v0, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, s1
+; GFX7-NEXT: s_lshr_b32 s4, s2, 8
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:4
; GFX7-NEXT: v_mov_b32_e32 v1, s4
-; GFX7-NEXT: s_lshr_b32 s4, s3, 24
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:13
-; GFX7-NEXT: v_mov_b32_e32 v1, s4
-; GFX7-NEXT: s_lshr_b32 s3, s3, 16
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:15
-; GFX7-NEXT: v_mov_b32_e32 v1, s3
-; GFX7-NEXT: s_lshr_b32 s3, s2, 8
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:14
-; GFX7-NEXT: v_mov_b32_e32 v1, s3
-; GFX7-NEXT: s_lshr_b32 s3, s2, 24
+; GFX7-NEXT: s_lshr_b32 s4, s2, 24
; GFX7-NEXT: ds_write_b8 v0, v1 offset:9
-; GFX7-NEXT: v_mov_b32_e32 v1, s3
+; GFX7-NEXT: v_mov_b32_e32 v1, s4
; GFX7-NEXT: s_lshr_b32 s2, s2, 16
; GFX7-NEXT: ds_write_b8 v0, v1 offset:11
; GFX7-NEXT: v_mov_b32_e32 v1, s2
-; GFX7-NEXT: s_lshr_b32 s2, s1, 8
+; GFX7-NEXT: s_lshr_b32 s2, s3, 8
; GFX7-NEXT: ds_write_b8 v0, v1 offset:10
; GFX7-NEXT: v_mov_b32_e32 v1, s2
-; GFX7-NEXT: s_lshr_b32 s2, s1, 24
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:5
+; GFX7-NEXT: s_lshr_b32 s2, s3, 24
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:13
; GFX7-NEXT: v_mov_b32_e32 v1, s2
-; GFX7-NEXT: s_lshr_b32 s1, s1, 16
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:7
-; GFX7-NEXT: v_mov_b32_e32 v1, s1
-; GFX7-NEXT: s_lshr_b32 s1, s0, 8
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:6
-; GFX7-NEXT: v_mov_b32_e32 v1, s1
-; GFX7-NEXT: s_lshr_b32 s1, s0, 24
+; GFX7-NEXT: s_lshr_b32 s2, s3, 16
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:15
+; GFX7-NEXT: v_mov_b32_e32 v1, s2
+; GFX7-NEXT: s_lshr_b32 s2, s0, 8
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:14
+; GFX7-NEXT: v_mov_b32_e32 v1, s2
+; GFX7-NEXT: s_lshr_b32 s2, s0, 24
; GFX7-NEXT: ds_write_b8 v0, v1 offset:1
-; GFX7-NEXT: v_mov_b32_e32 v1, s1
+; GFX7-NEXT: v_mov_b32_e32 v1, s2
; GFX7-NEXT: s_lshr_b32 s0, s0, 16
; GFX7-NEXT: ds_write_b8 v0, v1 offset:3
; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: s_lshr_b32 s0, s1, 8
; GFX7-NEXT: ds_write_b8 v0, v1 offset:2
+; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: s_lshr_b32 s0, s1, 24
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:5
+; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: s_lshr_b32 s0, s1, 16
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:7
+; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:6
; GFX7-NEXT: s_endpgm
;
; GFX6-LABEL: store_lds_v4i32_align1:
@@ -181,50 +181,50 @@
; GFX6-NEXT: s_mov_b32 m0, -1
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v0, s4
-; GFX6-NEXT: v_mov_b32_e32 v1, s3
-; GFX6-NEXT: v_mov_b32_e32 v2, s2
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:12
-; GFX6-NEXT: ds_write_b8 v0, v2 offset:8
-; GFX6-NEXT: v_mov_b32_e32 v1, s1
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:4
+; GFX6-NEXT: v_mov_b32_e32 v1, s2
+; GFX6-NEXT: v_mov_b32_e32 v2, s3
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:8
+; GFX6-NEXT: ds_write_b8 v0, v2 offset:12
; GFX6-NEXT: v_mov_b32_e32 v1, s0
-; GFX6-NEXT: s_lshr_b32 s4, s3, 8
; GFX6-NEXT: ds_write_b8 v0, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, s1
+; GFX6-NEXT: s_lshr_b32 s4, s2, 8
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:4
; GFX6-NEXT: v_mov_b32_e32 v1, s4
-; GFX6-NEXT: s_lshr_b32 s4, s3, 24
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:13
-; GFX6-NEXT: v_mov_b32_e32 v1, s4
-; GFX6-NEXT: s_lshr_b32 s3, s3, 16
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:15
-; GFX6-NEXT: v_mov_b32_e32 v1, s3
-; GFX6-NEXT: s_lshr_b32 s3, s2, 8
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:14
-; GFX6-NEXT: v_mov_b32_e32 v1, s3
-; GFX6-NEXT: s_lshr_b32 s3, s2, 24
+; GFX6-NEXT: s_lshr_b32 s4, s2, 24
; GFX6-NEXT: ds_write_b8 v0, v1 offset:9
-; GFX6-NEXT: v_mov_b32_e32 v1, s3
+; GFX6-NEXT: v_mov_b32_e32 v1, s4
; GFX6-NEXT: s_lshr_b32 s2, s2, 16
; GFX6-NEXT: ds_write_b8 v0, v1 offset:11
; GFX6-NEXT: v_mov_b32_e32 v1, s2
-; GFX6-NEXT: s_lshr_b32 s2, s1, 8
+; GFX6-NEXT: s_lshr_b32 s2, s3, 8
; GFX6-NEXT: ds_write_b8 v0, v1 offset:10
; GFX6-NEXT: v_mov_b32_e32 v1, s2
-; GFX6-NEXT: s_lshr_b32 s2, s1, 24
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:5
+; GFX6-NEXT: s_lshr_b32 s2, s3, 24
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:13
; GFX6-NEXT: v_mov_b32_e32 v1, s2
-; GFX6-NEXT: s_lshr_b32 s1, s1, 16
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:7
-; GFX6-NEXT: v_mov_b32_e32 v1, s1
-; GFX6-NEXT: s_lshr_b32 s1, s0, 8
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:6
-; GFX6-NEXT: v_mov_b32_e32 v1, s1
-; GFX6-NEXT: s_lshr_b32 s1, s0, 24
+; GFX6-NEXT: s_lshr_b32 s2, s3, 16
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:15
+; GFX6-NEXT: v_mov_b32_e32 v1, s2
+; GFX6-NEXT: s_lshr_b32 s2, s0, 8
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:14
+; GFX6-NEXT: v_mov_b32_e32 v1, s2
+; GFX6-NEXT: s_lshr_b32 s2, s0, 24
; GFX6-NEXT: ds_write_b8 v0, v1 offset:1
-; GFX6-NEXT: v_mov_b32_e32 v1, s1
+; GFX6-NEXT: v_mov_b32_e32 v1, s2
; GFX6-NEXT: s_lshr_b32 s0, s0, 16
; GFX6-NEXT: ds_write_b8 v0, v1 offset:3
; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: s_lshr_b32 s0, s1, 8
; GFX6-NEXT: ds_write_b8 v0, v1 offset:2
+; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: s_lshr_b32 s0, s1, 24
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:5
+; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: s_lshr_b32 s0, s1, 16
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:7
+; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:6
; GFX6-NEXT: s_endpgm
;
; GFX10-LABEL: store_lds_v4i32_align1:
@@ -234,42 +234,42 @@
; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x10
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, s2
-; GFX10-NEXT: v_mov_b32_e32 v1, s7
-; GFX10-NEXT: s_lshr_b32 s3, s6, 24
-; GFX10-NEXT: v_mov_b32_e32 v2, s6
-; GFX10-NEXT: s_lshr_b32 s0, s7, 8
-; GFX10-NEXT: s_lshr_b32 s2, s6, 8
-; GFX10-NEXT: s_lshr_b32 s6, s5, 8
-; GFX10-NEXT: v_mov_b32_e32 v3, s5
-; GFX10-NEXT: s_lshr_b32 s1, s7, 24
-; GFX10-NEXT: s_lshr_b32 s5, s5, 24
+; GFX10-NEXT: v_mov_b32_e32 v1, s6
+; GFX10-NEXT: s_lshr_b32 s3, s7, 24
+; GFX10-NEXT: v_mov_b32_e32 v2, s7
+; GFX10-NEXT: s_lshr_b32 s0, s6, 8
+; GFX10-NEXT: s_lshr_b32 s1, s6, 24
+; GFX10-NEXT: s_lshr_b32 s6, s4, 8
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
+; GFX10-NEXT: s_lshr_b32 s2, s7, 8
+; GFX10-NEXT: s_lshr_b32 s4, s4, 24
; GFX10-NEXT: v_mov_b32_e32 v8, s3
; GFX10-NEXT: v_mov_b32_e32 v5, s0
; GFX10-NEXT: v_mov_b32_e32 v9, s6
-; GFX10-NEXT: s_lshr_b32 s0, s4, 8
-; GFX10-NEXT: v_mov_b32_e32 v4, s4
+; GFX10-NEXT: s_lshr_b32 s0, s5, 8
+; GFX10-NEXT: v_mov_b32_e32 v4, s5
; GFX10-NEXT: v_mov_b32_e32 v6, s1
; GFX10-NEXT: v_mov_b32_e32 v7, s2
-; GFX10-NEXT: ds_write_b8 v0, v1 offset:12
-; GFX10-NEXT: ds_write_b8_d16_hi v0, v1 offset:14
-; GFX10-NEXT: ds_write_b8 v0, v2 offset:8
-; GFX10-NEXT: ds_write_b8_d16_hi v0, v2 offset:10
-; GFX10-NEXT: ds_write_b8 v0, v3 offset:4
-; GFX10-NEXT: ds_write_b8_d16_hi v0, v3 offset:6
-; GFX10-NEXT: ds_write_b8 v0, v4
-; GFX10-NEXT: ds_write_b8_d16_hi v0, v4 offset:2
-; GFX10-NEXT: ds_write_b8 v0, v5 offset:13
-; GFX10-NEXT: ds_write_b8 v0, v6 offset:15
-; GFX10-NEXT: ds_write_b8 v0, v7 offset:9
-; GFX10-NEXT: v_mov_b32_e32 v1, s5
-; GFX10-NEXT: s_lshr_b32 s1, s4, 24
+; GFX10-NEXT: ds_write_b8 v0, v1 offset:8
+; GFX10-NEXT: ds_write_b8_d16_hi v0, v1 offset:10
+; GFX10-NEXT: ds_write_b8 v0, v2 offset:12
+; GFX10-NEXT: ds_write_b8_d16_hi v0, v2 offset:14
+; GFX10-NEXT: ds_write_b8 v0, v3
+; GFX10-NEXT: ds_write_b8_d16_hi v0, v3 offset:2
+; GFX10-NEXT: ds_write_b8 v0, v4 offset:4
+; GFX10-NEXT: ds_write_b8_d16_hi v0, v4 offset:6
+; GFX10-NEXT: ds_write_b8 v0, v5 offset:9
+; GFX10-NEXT: ds_write_b8 v0, v6 offset:11
+; GFX10-NEXT: ds_write_b8 v0, v7 offset:13
+; GFX10-NEXT: v_mov_b32_e32 v1, s4
+; GFX10-NEXT: s_lshr_b32 s1, s5, 24
; GFX10-NEXT: v_mov_b32_e32 v2, s0
; GFX10-NEXT: v_mov_b32_e32 v3, s1
-; GFX10-NEXT: ds_write_b8 v0, v8 offset:11
-; GFX10-NEXT: ds_write_b8 v0, v9 offset:5
-; GFX10-NEXT: ds_write_b8 v0, v1 offset:7
-; GFX10-NEXT: ds_write_b8 v0, v2 offset:1
-; GFX10-NEXT: ds_write_b8 v0, v3 offset:3
+; GFX10-NEXT: ds_write_b8 v0, v8 offset:15
+; GFX10-NEXT: ds_write_b8 v0, v9 offset:1
+; GFX10-NEXT: ds_write_b8 v0, v1 offset:3
+; GFX10-NEXT: ds_write_b8 v0, v2 offset:5
+; GFX10-NEXT: ds_write_b8 v0, v3 offset:7
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: store_lds_v4i32_align1:
@@ -278,37 +278,38 @@
; GFX11-NEXT: s_load_b32 s4, s[0:1], 0x0
; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x10
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s3
-; GFX11-NEXT: s_lshr_b32 s4, s3, 8
+; GFX11-NEXT: v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s2
+; GFX11-NEXT: v_dual_mov_b32 v2, s3 :: v_dual_mov_b32 v3, s0
+; GFX11-NEXT: s_lshr_b32 s4, s2, 8
+; GFX11-NEXT: s_lshr_b32 s5, s3, 8
; GFX11-NEXT: s_lshr_b32 s3, s3, 24
-; GFX11-NEXT: s_lshr_b32 s5, s2, 8
-; GFX11-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s1
-; GFX11-NEXT: s_lshr_b32 s2, s2, 24
-; GFX11-NEXT: s_lshr_b32 s6, s1, 8
-; GFX11-NEXT: v_dual_mov_b32 v6, s3 :: v_dual_mov_b32 v7, s5
-; GFX11-NEXT: v_dual_mov_b32 v8, s2 :: v_dual_mov_b32 v9, s6
-; GFX11-NEXT: v_dual_mov_b32 v4, s0 :: v_dual_mov_b32 v5, s4
-; GFX11-NEXT: s_lshr_b32 s1, s1, 24
-; GFX11-NEXT: s_lshr_b32 s7, s0, 8
+; GFX11-NEXT: s_lshr_b32 s6, s0, 8
; GFX11-NEXT: s_lshr_b32 s0, s0, 24
-; GFX11-NEXT: ds_store_b8 v0, v2 offset:8
-; GFX11-NEXT: ds_store_b8_d16_hi v0, v2 offset:10
-; GFX11-NEXT: ds_store_b8 v0, v1 offset:12
-; GFX11-NEXT: ds_store_b8 v0, v4
-; GFX11-NEXT: ds_store_b8_d16_hi v0, v4 offset:2
-; GFX11-NEXT: ds_store_b8 v0, v3 offset:4
-; GFX11-NEXT: ds_store_b8 v0, v5 offset:13
-; GFX11-NEXT: ds_store_b8_d16_hi v0, v1 offset:14
-; GFX11-NEXT: ds_store_b8 v0, v6 offset:15
-; GFX11-NEXT: v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v10, s7
-; GFX11-NEXT: v_mov_b32_e32 v11, s0
-; GFX11-NEXT: ds_store_b8 v0, v7 offset:9
-; GFX11-NEXT: ds_store_b8 v0, v8 offset:11
-; GFX11-NEXT: ds_store_b8 v0, v9 offset:5
-; GFX11-NEXT: ds_store_b8_d16_hi v0, v3 offset:6
-; GFX11-NEXT: ds_store_b8 v0, v1 offset:7
-; GFX11-NEXT: ds_store_b8 v0, v10 offset:1
-; GFX11-NEXT: ds_store_b8 v0, v11 offset:3
+; GFX11-NEXT: v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v5, s4
+; GFX11-NEXT: s_lshr_b32 s2, s2, 24
+; GFX11-NEXT: s_lshr_b32 s7, s1, 8
+; GFX11-NEXT: v_dual_mov_b32 v8, s3 :: v_dual_mov_b32 v9, s6
+; GFX11-NEXT: v_mov_b32_e32 v10, s0
+; GFX11-NEXT: v_dual_mov_b32 v6, s2 :: v_dual_mov_b32 v7, s5
+; GFX11-NEXT: ds_store_b8 v0, v1 offset:8
+; GFX11-NEXT: ds_store_b8 v0, v3
+; GFX11-NEXT: ds_store_b8_d16_hi v0, v1 offset:10
+; GFX11-NEXT: ds_store_b8 v0, v5 offset:9
+; GFX11-NEXT: ds_store_b8 v0, v2 offset:12
+; GFX11-NEXT: ds_store_b8 v0, v6 offset:11
+; GFX11-NEXT: ds_store_b8 v0, v7 offset:13
+; GFX11-NEXT: ds_store_b8_d16_hi v0, v2 offset:14
+; GFX11-NEXT: ds_store_b8 v0, v8 offset:15
+; GFX11-NEXT: v_mov_b32_e32 v1, s7
+; GFX11-NEXT: s_lshr_b32 s0, s1, 24
+; GFX11-NEXT: ds_store_b8_d16_hi v0, v3 offset:2
+; GFX11-NEXT: ds_store_b8 v0, v9 offset:1
+; GFX11-NEXT: v_mov_b32_e32 v2, s0
+; GFX11-NEXT: ds_store_b8 v0, v4 offset:4
+; GFX11-NEXT: ds_store_b8 v0, v10 offset:3
+; GFX11-NEXT: ds_store_b8 v0, v1 offset:5
+; GFX11-NEXT: ds_store_b8_d16_hi v0, v4 offset:6
+; GFX11-NEXT: ds_store_b8 v0, v2 offset:7
; GFX11-NEXT: s_endpgm
store <4 x i32> %x, ptr addrspace(3) %out, align 1
ret void
@@ -321,18 +322,18 @@
; GFX9-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x10
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v0, s2
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: v_mov_b32_e32 v2, s6
-; GFX9-NEXT: ds_write_b16 v0, v1 offset:12
-; GFX9-NEXT: ds_write_b16_d16_hi v0, v1 offset:14
-; GFX9-NEXT: ds_write_b16 v0, v2 offset:8
-; GFX9-NEXT: ds_write_b16_d16_hi v0, v2 offset:10
-; GFX9-NEXT: v_mov_b32_e32 v1, s5
-; GFX9-NEXT: ds_write_b16 v0, v1 offset:4
-; GFX9-NEXT: ds_write_b16_d16_hi v0, v1 offset:6
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: v_mov_b32_e32 v2, s7
+; GFX9-NEXT: ds_write_b16 v0, v1 offset:8
+; GFX9-NEXT: ds_write_b16_d16_hi v0, v1 offset:10
+; GFX9-NEXT: ds_write_b16 v0, v2 offset:12
+; GFX9-NEXT: ds_write_b16_d16_hi v0, v2 offset:14
; GFX9-NEXT: v_mov_b32_e32 v1, s4
; GFX9-NEXT: ds_write_b16 v0, v1
; GFX9-NEXT: ds_write_b16_d16_hi v0, v1 offset:2
+; GFX9-NEXT: v_mov_b32_e32 v1, s5
+; GFX9-NEXT: ds_write_b16 v0, v1 offset:4
+; GFX9-NEXT: ds_write_b16_d16_hi v0, v1 offset:6
; GFX9-NEXT: s_endpgm
;
; GFX7-LABEL: store_lds_v4i32_align2:
@@ -342,26 +343,26 @@
; GFX7-NEXT: s_mov_b32 m0, -1
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v0, s4
-; GFX7-NEXT: v_mov_b32_e32 v1, s3
-; GFX7-NEXT: v_mov_b32_e32 v2, s2
-; GFX7-NEXT: ds_write_b16 v0, v1 offset:12
-; GFX7-NEXT: ds_write_b16 v0, v2 offset:8
-; GFX7-NEXT: v_mov_b32_e32 v1, s1
-; GFX7-NEXT: ds_write_b16 v0, v1 offset:4
-; GFX7-NEXT: v_mov_b32_e32 v1, s0
-; GFX7-NEXT: s_lshr_b32 s3, s3, 16
-; GFX7-NEXT: ds_write_b16 v0, v1
-; GFX7-NEXT: v_mov_b32_e32 v1, s3
-; GFX7-NEXT: s_lshr_b32 s2, s2, 16
-; GFX7-NEXT: ds_write_b16 v0, v1 offset:14
; GFX7-NEXT: v_mov_b32_e32 v1, s2
-; GFX7-NEXT: s_lshr_b32 s1, s1, 16
-; GFX7-NEXT: ds_write_b16 v0, v1 offset:10
-; GFX7-NEXT: v_mov_b32_e32 v1, s1
-; GFX7-NEXT: s_lshr_b32 s0, s0, 16
-; GFX7-NEXT: ds_write_b16 v0, v1 offset:6
+; GFX7-NEXT: v_mov_b32_e32 v2, s3
+; GFX7-NEXT: ds_write_b16 v0, v1 offset:8
+; GFX7-NEXT: ds_write_b16 v0, v2 offset:12
; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: ds_write_b16 v0, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, s1
+; GFX7-NEXT: s_lshr_b32 s2, s2, 16
+; GFX7-NEXT: ds_write_b16 v0, v1 offset:4
+; GFX7-NEXT: v_mov_b32_e32 v1, s2
+; GFX7-NEXT: s_lshr_b32 s2, s3, 16
+; GFX7-NEXT: ds_write_b16 v0, v1 offset:10
+; GFX7-NEXT: v_mov_b32_e32 v1, s2
+; GFX7-NEXT: s_lshr_b32 s0, s0, 16
+; GFX7-NEXT: ds_write_b16 v0, v1 offset:14
+; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: s_lshr_b32 s0, s1, 16
; GFX7-NEXT: ds_write_b16 v0, v1 offset:2
+; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: ds_write_b16 v0, v1 offset:6
; GFX7-NEXT: s_endpgm
;
; GFX6-LABEL: store_lds_v4i32_align2:
@@ -371,26 +372,26 @@
; GFX6-NEXT: s_mov_b32 m0, -1
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v0, s4
-; GFX6-NEXT: v_mov_b32_e32 v1, s3
-; GFX6-NEXT: v_mov_b32_e32 v2, s2
-; GFX6-NEXT: ds_write_b16 v0, v1 offset:12
-; GFX6-NEXT: ds_write_b16 v0, v2 offset:8
-; GFX6-NEXT: v_mov_b32_e32 v1, s1
-; GFX6-NEXT: ds_write_b16 v0, v1 offset:4
-; GFX6-NEXT: v_mov_b32_e32 v1, s0
-; GFX6-NEXT: s_lshr_b32 s3, s3, 16
-; GFX6-NEXT: ds_write_b16 v0, v1
-; GFX6-NEXT: v_mov_b32_e32 v1, s3
-; GFX6-NEXT: s_lshr_b32 s2, s2, 16
-; GFX6-NEXT: ds_write_b16 v0, v1 offset:14
; GFX6-NEXT: v_mov_b32_e32 v1, s2
-; GFX6-NEXT: s_lshr_b32 s1, s1, 16
-; GFX6-NEXT: ds_write_b16 v0, v1 offset:10
-; GFX6-NEXT: v_mov_b32_e32 v1, s1
-; GFX6-NEXT: s_lshr_b32 s0, s0, 16
-; GFX6-NEXT: ds_write_b16 v0, v1 offset:6
+; GFX6-NEXT: v_mov_b32_e32 v2, s3
+; GFX6-NEXT: ds_write_b16 v0, v1 offset:8
+; GFX6-NEXT: ds_write_b16 v0, v2 offset:12
; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: ds_write_b16 v0, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, s1
+; GFX6-NEXT: s_lshr_b32 s2, s2, 16
+; GFX6-NEXT: ds_write_b16 v0, v1 offset:4
+; GFX6-NEXT: v_mov_b32_e32 v1, s2
+; GFX6-NEXT: s_lshr_b32 s2, s3, 16
+; GFX6-NEXT: ds_write_b16 v0, v1 offset:10
+; GFX6-NEXT: v_mov_b32_e32 v1, s2
+; GFX6-NEXT: s_lshr_b32 s0, s0, 16
+; GFX6-NEXT: ds_write_b16 v0, v1 offset:14
+; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: s_lshr_b32 s0, s1, 16
; GFX6-NEXT: ds_write_b16 v0, v1 offset:2
+; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: ds_write_b16 v0, v1 offset:6
; GFX6-NEXT: s_endpgm
;
; GFX10-LABEL: store_lds_v4i32_align2:
@@ -400,18 +401,18 @@
; GFX10-NEXT: s_load_dwordx4 s[4:7], s[0:1], 0x10
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, s2
-; GFX10-NEXT: v_mov_b32_e32 v1, s7
-; GFX10-NEXT: v_mov_b32_e32 v2, s6
-; GFX10-NEXT: v_mov_b32_e32 v3, s5
-; GFX10-NEXT: v_mov_b32_e32 v4, s4
-; GFX10-NEXT: ds_write_b16 v0, v1 offset:12
-; GFX10-NEXT: ds_write_b16_d16_hi v0, v1 offset:14
-; GFX10-NEXT: ds_write_b16 v0, v2 offset:8
-; GFX10-NEXT: ds_write_b16_d16_hi v0, v2 offset:10
-; GFX10-NEXT: ds_write_b16 v0, v3 offset:4
-; GFX10-NEXT: ds_write_b16_d16_hi v0, v3 offset:6
-; GFX10-NEXT: ds_write_b16 v0, v4
-; GFX10-NEXT: ds_write_b16_d16_hi v0, v4 offset:2
+; GFX10-NEXT: v_mov_b32_e32 v1, s6
+; GFX10-NEXT: v_mov_b32_e32 v2, s7
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
+; GFX10-NEXT: v_mov_b32_e32 v4, s5
+; GFX10-NEXT: ds_write_b16 v0, v1 offset:8
+; GFX10-NEXT: ds_write_b16_d16_hi v0, v1 offset:10
+; GFX10-NEXT: ds_write_b16 v0, v2 offset:12
+; GFX10-NEXT: ds_write_b16_d16_hi v0, v2 offset:14
+; GFX10-NEXT: ds_write_b16 v0, v3
+; GFX10-NEXT: ds_write_b16_d16_hi v0, v3 offset:2
+; GFX10-NEXT: ds_write_b16 v0, v4 offset:4
+; GFX10-NEXT: ds_write_b16_d16_hi v0, v4 offset:6
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: store_lds_v4i32_align2:
@@ -420,17 +421,17 @@
; GFX11-NEXT: s_load_b32 s4, s[0:1], 0x0
; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x10
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
-; GFX11-NEXT: v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s3
-; GFX11-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
-; GFX11-NEXT: v_mov_b32_e32 v4, s2
-; GFX11-NEXT: ds_store_b16_d16_hi v0, v1 offset:14
-; GFX11-NEXT: ds_store_b16 v0, v2
-; GFX11-NEXT: ds_store_b16 v0, v3 offset:4
-; GFX11-NEXT: ds_store_b16 v0, v4 offset:8
-; GFX11-NEXT: ds_store_b16 v0, v1 offset:12
-; GFX11-NEXT: ds_store_b16_d16_hi v0, v4 offset:10
-; GFX11-NEXT: ds_store_b16_d16_hi v0, v3 offset:6
-; GFX11-NEXT: ds_store_b16_d16_hi v0, v2 offset:2
+; GFX11-NEXT: v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s2
+; GFX11-NEXT: v_dual_mov_b32 v2, s3 :: v_dual_mov_b32 v3, s0
+; GFX11-NEXT: v_mov_b32_e32 v4, s1
+; GFX11-NEXT: ds_store_b16_d16_hi v0, v1 offset:10
+; GFX11-NEXT: ds_store_b16 v0, v2 offset:12
+; GFX11-NEXT: ds_store_b16_d16_hi v0, v2 offset:14
+; GFX11-NEXT: ds_store_b16 v0, v3
+; GFX11-NEXT: ds_store_b16_d16_hi v0, v3 offset:2
+; GFX11-NEXT: ds_store_b16 v0, v4 offset:4
+; GFX11-NEXT: ds_store_b16 v0, v1 offset:8
+; GFX11-NEXT: ds_store_b16_d16_hi v0, v4 offset:6
; GFX11-NEXT: s_endpgm
store <4 x i32> %x, ptr addrspace(3) %out, align 2
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/store-local.96.ll b/llvm/test/CodeGen/AMDGPU/store-local.96.ll
index 507b411..1a7d6d9 100644
--- a/llvm/test/CodeGen/AMDGPU/store-local.96.ll
+++ b/llvm/test/CodeGen/AMDGPU/store-local.96.ll
@@ -80,32 +80,32 @@
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v0, s2
; GFX9-NEXT: v_mov_b32_e32 v1, s6
-; GFX9-NEXT: v_mov_b32_e32 v2, s5
+; GFX9-NEXT: v_mov_b32_e32 v2, s4
; GFX9-NEXT: ds_write_b8 v0, v1 offset:8
; GFX9-NEXT: ds_write_b8_d16_hi v0, v1 offset:10
-; GFX9-NEXT: ds_write_b8 v0, v2 offset:4
-; GFX9-NEXT: ds_write_b8_d16_hi v0, v2 offset:6
-; GFX9-NEXT: v_mov_b32_e32 v1, s4
+; GFX9-NEXT: ds_write_b8 v0, v2
+; GFX9-NEXT: ds_write_b8_d16_hi v0, v2 offset:2
+; GFX9-NEXT: v_mov_b32_e32 v1, s5
; GFX9-NEXT: s_lshr_b32 s0, s6, 8
-; GFX9-NEXT: ds_write_b8 v0, v1
-; GFX9-NEXT: ds_write_b8_d16_hi v0, v1 offset:2
+; GFX9-NEXT: ds_write_b8 v0, v1 offset:4
+; GFX9-NEXT: ds_write_b8_d16_hi v0, v1 offset:6
; GFX9-NEXT: v_mov_b32_e32 v1, s0
; GFX9-NEXT: s_lshr_b32 s0, s6, 24
; GFX9-NEXT: ds_write_b8 v0, v1 offset:9
; GFX9-NEXT: v_mov_b32_e32 v1, s0
-; GFX9-NEXT: s_lshr_b32 s0, s5, 8
-; GFX9-NEXT: ds_write_b8 v0, v1 offset:11
-; GFX9-NEXT: v_mov_b32_e32 v1, s0
-; GFX9-NEXT: s_lshr_b32 s0, s5, 24
-; GFX9-NEXT: ds_write_b8 v0, v1 offset:5
-; GFX9-NEXT: v_mov_b32_e32 v1, s0
; GFX9-NEXT: s_lshr_b32 s0, s4, 8
-; GFX9-NEXT: ds_write_b8 v0, v1 offset:7
+; GFX9-NEXT: ds_write_b8 v0, v1 offset:11
; GFX9-NEXT: v_mov_b32_e32 v1, s0
; GFX9-NEXT: s_lshr_b32 s0, s4, 24
; GFX9-NEXT: ds_write_b8 v0, v1 offset:1
; GFX9-NEXT: v_mov_b32_e32 v1, s0
+; GFX9-NEXT: s_lshr_b32 s0, s5, 8
; GFX9-NEXT: ds_write_b8 v0, v1 offset:3
+; GFX9-NEXT: v_mov_b32_e32 v1, s0
+; GFX9-NEXT: s_lshr_b32 s0, s5, 24
+; GFX9-NEXT: ds_write_b8 v0, v1 offset:5
+; GFX9-NEXT: v_mov_b32_e32 v1, s0
+; GFX9-NEXT: ds_write_b8 v0, v1 offset:7
; GFX9-NEXT: s_endpgm
;
; GFX7-LABEL: store_lds_v3i32_align1:
@@ -116,12 +116,12 @@
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v0, s4
; GFX7-NEXT: v_mov_b32_e32 v1, s2
-; GFX7-NEXT: v_mov_b32_e32 v2, s1
+; GFX7-NEXT: v_mov_b32_e32 v2, s0
; GFX7-NEXT: ds_write_b8 v0, v1 offset:8
-; GFX7-NEXT: ds_write_b8 v0, v2 offset:4
-; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: ds_write_b8 v0, v2
+; GFX7-NEXT: v_mov_b32_e32 v1, s1
; GFX7-NEXT: s_lshr_b32 s3, s2, 8
-; GFX7-NEXT: ds_write_b8 v0, v1
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:4
; GFX7-NEXT: v_mov_b32_e32 v1, s3
; GFX7-NEXT: s_lshr_b32 s3, s2, 24
; GFX7-NEXT: ds_write_b8 v0, v1 offset:9
@@ -129,25 +129,25 @@
; GFX7-NEXT: s_lshr_b32 s2, s2, 16
; GFX7-NEXT: ds_write_b8 v0, v1 offset:11
; GFX7-NEXT: v_mov_b32_e32 v1, s2
-; GFX7-NEXT: s_lshr_b32 s2, s1, 8
+; GFX7-NEXT: s_lshr_b32 s2, s0, 8
; GFX7-NEXT: ds_write_b8 v0, v1 offset:10
; GFX7-NEXT: v_mov_b32_e32 v1, s2
-; GFX7-NEXT: s_lshr_b32 s2, s1, 24
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:5
-; GFX7-NEXT: v_mov_b32_e32 v1, s2
-; GFX7-NEXT: s_lshr_b32 s1, s1, 16
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:7
-; GFX7-NEXT: v_mov_b32_e32 v1, s1
-; GFX7-NEXT: s_lshr_b32 s1, s0, 8
-; GFX7-NEXT: ds_write_b8 v0, v1 offset:6
-; GFX7-NEXT: v_mov_b32_e32 v1, s1
-; GFX7-NEXT: s_lshr_b32 s1, s0, 24
+; GFX7-NEXT: s_lshr_b32 s2, s0, 24
; GFX7-NEXT: ds_write_b8 v0, v1 offset:1
-; GFX7-NEXT: v_mov_b32_e32 v1, s1
+; GFX7-NEXT: v_mov_b32_e32 v1, s2
; GFX7-NEXT: s_lshr_b32 s0, s0, 16
; GFX7-NEXT: ds_write_b8 v0, v1 offset:3
; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: s_lshr_b32 s0, s1, 8
; GFX7-NEXT: ds_write_b8 v0, v1 offset:2
+; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: s_lshr_b32 s0, s1, 24
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:5
+; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: s_lshr_b32 s0, s1, 16
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:7
+; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: ds_write_b8 v0, v1 offset:6
; GFX7-NEXT: s_endpgm
;
; GFX6-LABEL: store_lds_v3i32_align1:
@@ -158,12 +158,12 @@
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v0, s4
; GFX6-NEXT: v_mov_b32_e32 v1, s2
-; GFX6-NEXT: v_mov_b32_e32 v2, s1
+; GFX6-NEXT: v_mov_b32_e32 v2, s0
; GFX6-NEXT: ds_write_b8 v0, v1 offset:8
-; GFX6-NEXT: ds_write_b8 v0, v2 offset:4
-; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: ds_write_b8 v0, v2
+; GFX6-NEXT: v_mov_b32_e32 v1, s1
; GFX6-NEXT: s_lshr_b32 s3, s2, 8
-; GFX6-NEXT: ds_write_b8 v0, v1
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:4
; GFX6-NEXT: v_mov_b32_e32 v1, s3
; GFX6-NEXT: s_lshr_b32 s3, s2, 24
; GFX6-NEXT: ds_write_b8 v0, v1 offset:9
@@ -171,25 +171,25 @@
; GFX6-NEXT: s_lshr_b32 s2, s2, 16
; GFX6-NEXT: ds_write_b8 v0, v1 offset:11
; GFX6-NEXT: v_mov_b32_e32 v1, s2
-; GFX6-NEXT: s_lshr_b32 s2, s1, 8
+; GFX6-NEXT: s_lshr_b32 s2, s0, 8
; GFX6-NEXT: ds_write_b8 v0, v1 offset:10
; GFX6-NEXT: v_mov_b32_e32 v1, s2
-; GFX6-NEXT: s_lshr_b32 s2, s1, 24
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:5
-; GFX6-NEXT: v_mov_b32_e32 v1, s2
-; GFX6-NEXT: s_lshr_b32 s1, s1, 16
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:7
-; GFX6-NEXT: v_mov_b32_e32 v1, s1
-; GFX6-NEXT: s_lshr_b32 s1, s0, 8
-; GFX6-NEXT: ds_write_b8 v0, v1 offset:6
-; GFX6-NEXT: v_mov_b32_e32 v1, s1
-; GFX6-NEXT: s_lshr_b32 s1, s0, 24
+; GFX6-NEXT: s_lshr_b32 s2, s0, 24
; GFX6-NEXT: ds_write_b8 v0, v1 offset:1
-; GFX6-NEXT: v_mov_b32_e32 v1, s1
+; GFX6-NEXT: v_mov_b32_e32 v1, s2
; GFX6-NEXT: s_lshr_b32 s0, s0, 16
; GFX6-NEXT: ds_write_b8 v0, v1 offset:3
; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: s_lshr_b32 s0, s1, 8
; GFX6-NEXT: ds_write_b8 v0, v1 offset:2
+; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: s_lshr_b32 s0, s1, 24
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:5
+; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: s_lshr_b32 s0, s1, 16
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:7
+; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: ds_write_b8 v0, v1 offset:6
; GFX6-NEXT: s_endpgm
;
; GFX10-LABEL: store_lds_v3i32_align1:
@@ -200,32 +200,32 @@
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, s2
; GFX10-NEXT: v_mov_b32_e32 v1, s6
-; GFX10-NEXT: v_mov_b32_e32 v2, s5
-; GFX10-NEXT: v_mov_b32_e32 v3, s4
+; GFX10-NEXT: v_mov_b32_e32 v2, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s5
; GFX10-NEXT: s_lshr_b32 s0, s6, 8
; GFX10-NEXT: s_lshr_b32 s1, s6, 24
-; GFX10-NEXT: s_lshr_b32 s2, s5, 8
-; GFX10-NEXT: s_lshr_b32 s3, s5, 24
-; GFX10-NEXT: s_lshr_b32 s5, s4, 8
-; GFX10-NEXT: s_lshr_b32 s4, s4, 24
+; GFX10-NEXT: s_lshr_b32 s2, s4, 8
+; GFX10-NEXT: s_lshr_b32 s3, s4, 24
+; GFX10-NEXT: s_lshr_b32 s4, s5, 8
+; GFX10-NEXT: s_lshr_b32 s5, s5, 24
; GFX10-NEXT: v_mov_b32_e32 v4, s0
; GFX10-NEXT: v_mov_b32_e32 v5, s1
; GFX10-NEXT: v_mov_b32_e32 v6, s2
; GFX10-NEXT: v_mov_b32_e32 v7, s3
-; GFX10-NEXT: v_mov_b32_e32 v8, s5
-; GFX10-NEXT: v_mov_b32_e32 v9, s4
+; GFX10-NEXT: v_mov_b32_e32 v8, s4
+; GFX10-NEXT: v_mov_b32_e32 v9, s5
; GFX10-NEXT: ds_write_b8 v0, v1 offset:8
; GFX10-NEXT: ds_write_b8_d16_hi v0, v1 offset:10
-; GFX10-NEXT: ds_write_b8 v0, v2 offset:4
-; GFX10-NEXT: ds_write_b8_d16_hi v0, v2 offset:6
-; GFX10-NEXT: ds_write_b8 v0, v3
-; GFX10-NEXT: ds_write_b8_d16_hi v0, v3 offset:2
+; GFX10-NEXT: ds_write_b8 v0, v2
+; GFX10-NEXT: ds_write_b8_d16_hi v0, v2 offset:2
+; GFX10-NEXT: ds_write_b8 v0, v3 offset:4
+; GFX10-NEXT: ds_write_b8_d16_hi v0, v3 offset:6
; GFX10-NEXT: ds_write_b8 v0, v4 offset:9
; GFX10-NEXT: ds_write_b8 v0, v5 offset:11
-; GFX10-NEXT: ds_write_b8 v0, v6 offset:5
-; GFX10-NEXT: ds_write_b8 v0, v7 offset:7
-; GFX10-NEXT: ds_write_b8 v0, v8 offset:1
-; GFX10-NEXT: ds_write_b8 v0, v9 offset:3
+; GFX10-NEXT: ds_write_b8 v0, v6 offset:1
+; GFX10-NEXT: ds_write_b8 v0, v7 offset:3
+; GFX10-NEXT: ds_write_b8 v0, v8 offset:5
+; GFX10-NEXT: ds_write_b8 v0, v9 offset:7
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: store_lds_v3i32_align1:
@@ -235,28 +235,29 @@
; GFX11-NEXT: s_load_b128 s[0:3], s[0:1], 0x10
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
; GFX11-NEXT: v_dual_mov_b32 v0, s4 :: v_dual_mov_b32 v1, s2
-; GFX11-NEXT: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v3, s0
; GFX11-NEXT: s_lshr_b32 s3, s2, 8
+; GFX11-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
; GFX11-NEXT: s_lshr_b32 s2, s2, 24
-; GFX11-NEXT: s_lshr_b32 s4, s1, 8
-; GFX11-NEXT: s_lshr_b32 s1, s1, 24
-; GFX11-NEXT: s_lshr_b32 s5, s0, 8
-; GFX11-NEXT: s_lshr_b32 s0, s0, 24
+; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: v_dual_mov_b32 v4, s3 :: v_dual_mov_b32 v5, s2
-; GFX11-NEXT: v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s1
-; GFX11-NEXT: v_dual_mov_b32 v8, s5 :: v_dual_mov_b32 v9, s0
+; GFX11-NEXT: s_lshr_b32 s4, s0, 8
+; GFX11-NEXT: s_lshr_b32 s0, s0, 24
+; GFX11-NEXT: s_lshr_b32 s5, s1, 8
+; GFX11-NEXT: s_lshr_b32 s1, s1, 24
+; GFX11-NEXT: v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s0
+; GFX11-NEXT: v_dual_mov_b32 v8, s5 :: v_dual_mov_b32 v9, s1
; GFX11-NEXT: ds_store_b8 v0, v1 offset:8
-; GFX11-NEXT: ds_store_b8 v0, v3
-; GFX11-NEXT: ds_store_b8_d16_hi v0, v3 offset:2
-; GFX11-NEXT: ds_store_b8 v0, v2 offset:4
+; GFX11-NEXT: ds_store_b8 v0, v2
; GFX11-NEXT: ds_store_b8 v0, v4 offset:9
; GFX11-NEXT: ds_store_b8_d16_hi v0, v1 offset:10
; GFX11-NEXT: ds_store_b8 v0, v5 offset:11
-; GFX11-NEXT: ds_store_b8 v0, v6 offset:5
-; GFX11-NEXT: ds_store_b8_d16_hi v0, v2 offset:6
-; GFX11-NEXT: ds_store_b8 v0, v7 offset:7
-; GFX11-NEXT: ds_store_b8 v0, v8 offset:1
-; GFX11-NEXT: ds_store_b8 v0, v9 offset:3
+; GFX11-NEXT: ds_store_b8_d16_hi v0, v2 offset:2
+; GFX11-NEXT: ds_store_b8 v0, v6 offset:1
+; GFX11-NEXT: ds_store_b8 v0, v3 offset:4
+; GFX11-NEXT: ds_store_b8 v0, v7 offset:3
+; GFX11-NEXT: ds_store_b8 v0, v8 offset:5
+; GFX11-NEXT: ds_store_b8_d16_hi v0, v3 offset:6
+; GFX11-NEXT: ds_store_b8 v0, v9 offset:7
; GFX11-NEXT: s_endpgm
store <3 x i32> %x, ptr addrspace(3) %out, align 1
ret void
@@ -270,14 +271,14 @@
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v0, s2
; GFX9-NEXT: v_mov_b32_e32 v1, s6
-; GFX9-NEXT: v_mov_b32_e32 v2, s5
+; GFX9-NEXT: v_mov_b32_e32 v2, s4
; GFX9-NEXT: ds_write_b16 v0, v1 offset:8
; GFX9-NEXT: ds_write_b16_d16_hi v0, v1 offset:10
-; GFX9-NEXT: ds_write_b16 v0, v2 offset:4
-; GFX9-NEXT: ds_write_b16_d16_hi v0, v2 offset:6
-; GFX9-NEXT: v_mov_b32_e32 v1, s4
-; GFX9-NEXT: ds_write_b16 v0, v1
-; GFX9-NEXT: ds_write_b16_d16_hi v0, v1 offset:2
+; GFX9-NEXT: ds_write_b16 v0, v2
+; GFX9-NEXT: ds_write_b16_d16_hi v0, v2 offset:2
+; GFX9-NEXT: v_mov_b32_e32 v1, s5
+; GFX9-NEXT: ds_write_b16 v0, v1 offset:4
+; GFX9-NEXT: ds_write_b16_d16_hi v0, v1 offset:6
; GFX9-NEXT: s_endpgm
;
; GFX7-LABEL: store_lds_v3i32_align2:
@@ -288,20 +289,20 @@
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v0, s4
; GFX7-NEXT: v_mov_b32_e32 v1, s2
-; GFX7-NEXT: v_mov_b32_e32 v2, s1
+; GFX7-NEXT: v_mov_b32_e32 v2, s0
; GFX7-NEXT: ds_write_b16 v0, v1 offset:8
-; GFX7-NEXT: ds_write_b16 v0, v2 offset:4
-; GFX7-NEXT: v_mov_b32_e32 v1, s0
-; GFX7-NEXT: s_lshr_b32 s2, s2, 16
-; GFX7-NEXT: ds_write_b16 v0, v1
-; GFX7-NEXT: v_mov_b32_e32 v1, s2
-; GFX7-NEXT: s_lshr_b32 s1, s1, 16
-; GFX7-NEXT: ds_write_b16 v0, v1 offset:10
+; GFX7-NEXT: ds_write_b16 v0, v2
; GFX7-NEXT: v_mov_b32_e32 v1, s1
+; GFX7-NEXT: s_lshr_b32 s2, s2, 16
+; GFX7-NEXT: ds_write_b16 v0, v1 offset:4
+; GFX7-NEXT: v_mov_b32_e32 v1, s2
; GFX7-NEXT: s_lshr_b32 s0, s0, 16
-; GFX7-NEXT: ds_write_b16 v0, v1 offset:6
+; GFX7-NEXT: ds_write_b16 v0, v1 offset:10
; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: s_lshr_b32 s0, s1, 16
; GFX7-NEXT: ds_write_b16 v0, v1 offset:2
+; GFX7-NEXT: v_mov_b32_e32 v1, s0
+; GFX7-NEXT: ds_write_b16 v0, v1 offset:6
; GFX7-NEXT: s_endpgm
;
; GFX6-LABEL: store_lds_v3i32_align2:
@@ -312,20 +313,20 @@
; GFX6-NEXT: s_waitcnt lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v0, s4
; GFX6-NEXT: v_mov_b32_e32 v1, s2
-; GFX6-NEXT: v_mov_b32_e32 v2, s1
+; GFX6-NEXT: v_mov_b32_e32 v2, s0
; GFX6-NEXT: ds_write_b16 v0, v1 offset:8
-; GFX6-NEXT: ds_write_b16 v0, v2 offset:4
-; GFX6-NEXT: v_mov_b32_e32 v1, s0
-; GFX6-NEXT: s_lshr_b32 s2, s2, 16
-; GFX6-NEXT: ds_write_b16 v0, v1
-; GFX6-NEXT: v_mov_b32_e32 v1, s2
-; GFX6-NEXT: s_lshr_b32 s1, s1, 16
-; GFX6-NEXT: ds_write_b16 v0, v1 offset:10
+; GFX6-NEXT: ds_write_b16 v0, v2
; GFX6-NEXT: v_mov_b32_e32 v1, s1
+; GFX6-NEXT: s_lshr_b32 s2, s2, 16
+; GFX6-NEXT: ds_write_b16 v0, v1 offset:4
+; GFX6-NEXT: v_mov_b32_e32 v1, s2
; GFX6-NEXT: s_lshr_b32 s0, s0, 16
-; GFX6-NEXT: ds_write_b16 v0, v1 offset:6
+; GFX6-NEXT: ds_write_b16 v0, v1 offset:10
; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: s_lshr_b32 s0, s1, 16
; GFX6-NEXT: ds_write_b16 v0, v1 offset:2
+; GFX6-NEXT: v_mov_b32_e32 v1, s0
+; GFX6-NEXT: ds_write_b16 v0, v1 offset:6
; GFX6-NEXT: s_endpgm
;
; GFX10-LABEL: store_lds_v3i32_align2:
@@ -336,14 +337,14 @@
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v0, s2
; GFX10-NEXT: v_mov_b32_e32 v1, s6
-; GFX10-NEXT: v_mov_b32_e32 v2, s5
-; GFX10-NEXT: v_mov_b32_e32 v3, s4
+; GFX10-NEXT: v_mov_b32_e32 v2, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s5
; GFX10-NEXT: ds_write_b16 v0, v1 offset:8
; GFX10-NEXT: ds_write_b16_d16_hi v0, v1 offset:10
-; GFX10-NEXT: ds_write_b16 v0, v2 offset:4
-; GFX10-NEXT: ds_write_b16_d16_hi v0, v2 offset:6
-; GFX10-NEXT: ds_write_b16 v0, v3
-; GFX10-NEXT: ds_write_b16_d16_hi v0, v3 offset:2
+; GFX10-NEXT: ds_write_b16 v0, v2
+; GFX10-NEXT: ds_write_b16_d16_hi v0, v2 offset:2
+; GFX10-NEXT: ds_write_b16 v0, v3 offset:4
+; GFX10-NEXT: ds_write_b16_d16_hi v0, v3 offset:6
; GFX10-NEXT: s_endpgm
;
; GFX11-LABEL: store_lds_v3i32_align2:
@@ -356,10 +357,10 @@
; GFX11-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
; GFX11-NEXT: ds_store_b16_d16_hi v0, v1 offset:10
; GFX11-NEXT: ds_store_b16 v0, v2
+; GFX11-NEXT: ds_store_b16_d16_hi v0, v2 offset:2
; GFX11-NEXT: ds_store_b16 v0, v3 offset:4
; GFX11-NEXT: ds_store_b16 v0, v1 offset:8
; GFX11-NEXT: ds_store_b16_d16_hi v0, v3 offset:6
-; GFX11-NEXT: ds_store_b16_d16_hi v0, v2 offset:2
; GFX11-NEXT: s_endpgm
store <3 x i32> %x, ptr addrspace(3) %out, align 2
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/widen-smrd-loads.ll b/llvm/test/CodeGen/AMDGPU/widen-smrd-loads.ll
index 633a69d..2bf74d9 100644
--- a/llvm/test/CodeGen/AMDGPU/widen-smrd-loads.ll
+++ b/llvm/test/CodeGen/AMDGPU/widen-smrd-loads.ll
@@ -234,9 +234,8 @@
; VI-NEXT: s_waitcnt lgkmcnt(0)
; VI-NEXT: s_load_dword s0, s[0:1], 0x0
; VI-NEXT: s_waitcnt lgkmcnt(0)
-; VI-NEXT: s_and_b32 s1, s0, 0xffff
+; VI-NEXT: s_add_i32 s1, s0, 12
; VI-NEXT: v_mov_b32_e32 v2, s0
-; VI-NEXT: s_add_i32 s1, s1, 12
; VI-NEXT: v_add_u32_sdwa v0, vcc, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1
; VI-NEXT: s_or_b32 s0, s1, 4
; VI-NEXT: v_or_b32_sdwa v2, v0, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
diff --git a/llvm/test/CodeGen/ARM/addsubcarry-promotion.ll b/llvm/test/CodeGen/ARM/addsubcarry-promotion.ll
index 9d07ed6..a385164 100644
--- a/llvm/test/CodeGen/ARM/addsubcarry-promotion.ll
+++ b/llvm/test/CodeGen/ARM/addsubcarry-promotion.ll
@@ -14,8 +14,9 @@
; ARM-NEXT: adds r0, r1, r0
; ARM-NEXT: movw r1, #65535
; ARM-NEXT: sxth r2, r2
-; ARM-NEXT: adc r0, r2, #1
-; ARM-NEXT: tst r0, r1
+; ARM-NEXT: adc r0, r2, #0
+; ARM-NEXT: uxth r0, r0
+; ARM-NEXT: cmp r0, r1
; ARM-NEXT: bxeq lr
; ARM-NEXT: .LBB0_1: @ %for.cond
; ARM-NEXT: @ =>This Inner Loop Header: Depth=1
@@ -25,25 +26,33 @@
; THUMBV6M: @ %bb.0: @ %entry
; THUMBV6M-NEXT: rsbs r2, r2, #0
; THUMBV6M-NEXT: sxth r2, r2
-; THUMBV6M-NEXT: movs r3, #1
+; THUMBV6M-NEXT: movs r3, #0
; THUMBV6M-NEXT: adds r0, r1, r0
; THUMBV6M-NEXT: adcs r3, r2
-; THUMBV6M-NEXT: lsls r0, r3, #16
+; THUMBV6M-NEXT: uxth r0, r3
+; THUMBV6M-NEXT: ldr r1, .LCPI0_0
+; THUMBV6M-NEXT: cmp r0, r1
; THUMBV6M-NEXT: beq .LBB0_2
; THUMBV6M-NEXT: .LBB0_1: @ %for.cond
; THUMBV6M-NEXT: @ =>This Inner Loop Header: Depth=1
; THUMBV6M-NEXT: b .LBB0_1
; THUMBV6M-NEXT: .LBB0_2: @ %if.end
; THUMBV6M-NEXT: bx lr
+; THUMBV6M-NEXT: .p2align 2
+; THUMBV6M-NEXT: @ %bb.3:
+; THUMBV6M-NEXT: .LCPI0_0:
+; THUMBV6M-NEXT: .long 65535 @ 0xffff
;
; THUMBV8M-BASE-LABEL: fn1:
; THUMBV8M-BASE: @ %bb.0: @ %entry
; THUMBV8M-BASE-NEXT: rsbs r2, r2, #0
; THUMBV8M-BASE-NEXT: sxth r2, r2
-; THUMBV8M-BASE-NEXT: movs r3, #1
+; THUMBV8M-BASE-NEXT: movs r3, #0
; THUMBV8M-BASE-NEXT: adds r0, r1, r0
; THUMBV8M-BASE-NEXT: adcs r3, r2
-; THUMBV8M-BASE-NEXT: lsls r0, r3, #16
+; THUMBV8M-BASE-NEXT: uxth r0, r3
+; THUMBV8M-BASE-NEXT: movw r1, #65535
+; THUMBV8M-BASE-NEXT: cmp r0, r1
; THUMBV8M-BASE-NEXT: beq .LBB0_2
; THUMBV8M-BASE-NEXT: .LBB0_1: @ %for.cond
; THUMBV8M-BASE-NEXT: @ =>This Inner Loop Header: Depth=1
@@ -55,9 +64,11 @@
; THUMB: @ %bb.0: @ %entry
; THUMB-NEXT: rsbs r2, r2, #0
; THUMB-NEXT: adds r0, r0, r1
+; THUMB-NEXT: movw r1, #65535
; THUMB-NEXT: sxth r2, r2
-; THUMB-NEXT: adc r0, r2, #1
-; THUMB-NEXT: lsls r0, r0, #16
+; THUMB-NEXT: adc r0, r2, #0
+; THUMB-NEXT: uxth r0, r0
+; THUMB-NEXT: cmp r0, r1
; THUMB-NEXT: it eq
; THUMB-NEXT: bxeq lr
; THUMB-NEXT: .LBB0_1: @ %for.cond
diff --git a/llvm/test/CodeGen/ARM/icmp-shift-opt.ll b/llvm/test/CodeGen/ARM/icmp-shift-opt.ll
index d0021db..c5fd16a 100644
--- a/llvm/test/CodeGen/ARM/icmp-shift-opt.ll
+++ b/llvm/test/CodeGen/ARM/icmp-shift-opt.ll
@@ -136,11 +136,12 @@
define i1 @opt_setcc_shl_ne_zero_i128(i128 %a) nounwind {
; CHECK-LABEL: opt_setcc_shl_ne_zero_i128:
; CHECK: @ %bb.0:
+; CHECK-NEXT: orr r3, r1, r3
; CHECK-NEXT: orr r0, r2, r0
-; CHECK-NEXT: orr r0, r1, r0
-; CHECK-NEXT: orr r1, r0, r3
-; CHECK-NEXT: lsl r1, r1, #17
-; CHECK-NEXT: orrs r0, r1, r0, lsr #15
+; CHECK-NEXT: orr r2, r0, r3
+; CHECK-NEXT: orr r0, r0, r1
+; CHECK-NEXT: lsr r0, r0, #15
+; CHECK-NEXT: orrs r0, r0, r2, lsl #17
; CHECK-NEXT: movwne r0, #1
; CHECK-NEXT: bx lr
%shl = shl i128 %a, 17
diff --git a/llvm/test/CodeGen/ARM/reg_sequence.ll b/llvm/test/CodeGen/ARM/reg_sequence.ll
index 102ec39..2f49862 100644
--- a/llvm/test/CodeGen/ARM/reg_sequence.ll
+++ b/llvm/test/CodeGen/ARM/reg_sequence.ll
@@ -259,8 +259,9 @@
; PR7162
define arm_aapcs_vfpcc i32 @t10(float %x) nounwind {
; CHECK-LABEL: t10:
+; CHECK: vdup.32 [[Q0:q[0-9]+]], d0[0]
; CHECK: vmov.i32 [[Q9:q[0-9]+]], #0x3f000000
-; CHECK: vmul.f32 [[Q8:q[0-9]+]], [[Q0:q[0-9]+]], [[Q0]]
+; CHECK: vmul.f32 [[Q8:q[0-9]+]], [[Q0]], [[Q0]]
; CHECK-NEXT: vadd.f32 [[Q8]], [[Q8]], [[Q8]]
; CHECK-NEXT: vadd.f32 [[Q1:q[0-9]+]], [[Q8]], [[Q8]]
; CHECK-NEXT: vmul.f32 [[Q8]], [[Q9]], d1[0]
diff --git a/llvm/test/CodeGen/Hexagon/autohvx/isel-vpackew.ll b/llvm/test/CodeGen/Hexagon/autohvx/isel-vpackew.ll
index 3319a76..04b925d 100644
--- a/llvm/test/CodeGen/Hexagon/autohvx/isel-vpackew.ll
+++ b/llvm/test/CodeGen/Hexagon/autohvx/isel-vpackew.ll
@@ -5,7 +5,7 @@
; CHECK-LABEL: f0:
; CHECK: // %bb.0: // %b0
; CHECK-NEXT: {
-; CHECK-NEXT: r7 = #124
+; CHECK-NEXT: r7 = #-4
; CHECK-NEXT: v0 = vmem(r0+#0)
; CHECK-NEXT: }
; CHECK-NEXT: {
diff --git a/llvm/test/CodeGen/Hexagon/autohvx/mulh.ll b/llvm/test/CodeGen/Hexagon/autohvx/mulh.ll
index b304465..8cf7cb2 100644
--- a/llvm/test/CodeGen/Hexagon/autohvx/mulh.ll
+++ b/llvm/test/CodeGen/Hexagon/autohvx/mulh.ll
@@ -10,13 +10,7 @@
; V60-NEXT: v1:0.w = vmpy(v1.h,v0.h)
; V60-NEXT: }
; V60-NEXT: {
-; V60-NEXT: r7 = #124
-; V60-NEXT: }
-; V60-NEXT: {
-; V60-NEXT: v1:0 = vshuff(v1,v0,r7)
-; V60-NEXT: }
-; V60-NEXT: {
-; V60-NEXT: v0.h = vpacko(v1.w,v0.w)
+; V60-NEXT: v0.h = vshuffo(v1.h,v0.h)
; V60-NEXT: }
; V60-NEXT: {
; V60-NEXT: jumpr r31
@@ -28,13 +22,7 @@
; V65-NEXT: v1:0.w = vmpy(v1.h,v0.h)
; V65-NEXT: }
; V65-NEXT: {
-; V65-NEXT: r7 = #124
-; V65-NEXT: }
-; V65-NEXT: {
-; V65-NEXT: v1:0 = vshuff(v1,v0,r7)
-; V65-NEXT: }
-; V65-NEXT: {
-; V65-NEXT: v0.h = vpacko(v1.w,v0.w)
+; V65-NEXT: v0.h = vshuffo(v1.h,v0.h)
; V65-NEXT: }
; V65-NEXT: {
; V65-NEXT: jumpr r31
@@ -46,13 +34,7 @@
; V69-NEXT: v1:0.w = vmpy(v1.h,v0.h)
; V69-NEXT: }
; V69-NEXT: {
-; V69-NEXT: r7 = #124
-; V69-NEXT: }
-; V69-NEXT: {
-; V69-NEXT: v1:0 = vshuff(v1,v0,r7)
-; V69-NEXT: }
-; V69-NEXT: {
-; V69-NEXT: v0.h = vpacko(v1.w,v0.w)
+; V69-NEXT: v0.h = vshuffo(v1.h,v0.h)
; V69-NEXT: }
; V69-NEXT: {
; V69-NEXT: jumpr r31
@@ -72,13 +54,7 @@
; V60-NEXT: v1:0.uw = vmpy(v1.uh,v0.uh)
; V60-NEXT: }
; V60-NEXT: {
-; V60-NEXT: r7 = #124
-; V60-NEXT: }
-; V60-NEXT: {
-; V60-NEXT: v1:0 = vshuff(v1,v0,r7)
-; V60-NEXT: }
-; V60-NEXT: {
-; V60-NEXT: v0.h = vpacko(v1.w,v0.w)
+; V60-NEXT: v0.h = vshuffo(v1.h,v0.h)
; V60-NEXT: }
; V60-NEXT: {
; V60-NEXT: jumpr r31
@@ -90,13 +66,7 @@
; V65-NEXT: v1:0.uw = vmpy(v1.uh,v0.uh)
; V65-NEXT: }
; V65-NEXT: {
-; V65-NEXT: r7 = #124
-; V65-NEXT: }
-; V65-NEXT: {
-; V65-NEXT: v1:0 = vshuff(v1,v0,r7)
-; V65-NEXT: }
-; V65-NEXT: {
-; V65-NEXT: v0.h = vpacko(v1.w,v0.w)
+; V65-NEXT: v0.h = vshuffo(v1.h,v0.h)
; V65-NEXT: }
; V65-NEXT: {
; V65-NEXT: jumpr r31
diff --git a/llvm/test/CodeGen/PowerPC/aix32-cc-abi-vaarg.ll b/llvm/test/CodeGen/PowerPC/aix32-cc-abi-vaarg.ll
index bf66a1e..fc38fdb 100644
--- a/llvm/test/CodeGen/PowerPC/aix32-cc-abi-vaarg.ll
+++ b/llvm/test/CodeGen/PowerPC/aix32-cc-abi-vaarg.ll
@@ -247,7 +247,7 @@
; 32BIT-LABEL: body: |
; 32BIT-DAG: liveins: $f1, $r5, $r6, $r7, $r8, $r9, $r10
; 32BIT-DAG: renamable $r3 = ADDI %fixed-stack.0, 0
-; 32BIT-DAG: STW killed renamable $r7, 8, %fixed-stack.0 :: (store (s32), align 8)
+; 32BIT-DAG: STW killed renamable $r7, 8, %fixed-stack.0 :: (store (s32) into %fixed-stack.0 + 8, align 8)
; 32BIT-DAG: STW renamable $r5, 0, %fixed-stack.0 :: (store (s32) into %fixed-stack.0, align 16)
; 32BIT-DAG: STW renamable $r6, 4, %fixed-stack.0 :: (store (s32) into %fixed-stack.0 + 4)
; 32BIT-DAG: STW killed renamable $r8, 12, %fixed-stack.0 :: (store (s32))
@@ -270,22 +270,22 @@
; ASM32-LABEL: double_stack_va_arg:
; ASM32: # %bb.0: # %entry
; ASM32-NEXT: fadd 0, 1, 2
-; ASM32-NEXT: addi 3, 1, 128
-; ASM32-NEXT: lwz 4, 132(1)
+; ASM32-NEXT: addi 4, 1, 128
+; ASM32-NEXT: lwz 3, 132(1)
; ASM32-NEXT: fadd 0, 0, 3
-; ASM32-NEXT: stw 3, -4(1)
+; ASM32-NEXT: stw 4, -4(1)
; ASM32-NEXT: fadd 0, 0, 4
-; ASM32-NEXT: lwz 3, 128(1)
+; ASM32-NEXT: lwz 4, 128(1)
; ASM32-NEXT: fadd 0, 0, 5
-; ASM32-NEXT: stw 3, -16(1)
+; ASM32-NEXT: stw 3, -12(1)
; ASM32-NEXT: fadd 0, 0, 6
-; ASM32-NEXT: stw 4, -12(1)
+; ASM32-NEXT: stw 4, -16(1)
; ASM32-NEXT: fadd 0, 0, 7
; ASM32-NEXT: lfd 1, -16(1)
; ASM32-NEXT: fadd 0, 0, 8
-; ASM32-NEXT: stw 3, -24(1)
+; ASM32-NEXT: stw 3, -20(1)
; ASM32-NEXT: fadd 0, 0, 9
-; ASM32-NEXT: stw 4, -20(1)
+; ASM32-NEXT: stw 4, -24(1)
; ASM32-NEXT: fadd 0, 0, 10
; ASM32-NEXT: fadd 0, 0, 11
; ASM32-NEXT: fadd 0, 0, 12
@@ -353,24 +353,24 @@
; 32BIT-LABEL: body: |
; 32BIT-DAG: liveins: $f1, $f2, $f3, $f4, $f5, $f6, $f7, $f8, $f9, $f10, $f11, $f12, $f13
-; 32BIT-DAG: renamable $r3 = ADDI %fixed-stack.0, 0
-; 32BIT-DAG: STW killed renamable $r3, 0, %stack.0.arg1 :: (store (s32) into %ir.arg1)
-; 32BIT-DAG: renamable $r3 = LWZ 0, %fixed-stack.0 :: (load (s32) from %ir.argp.cur142, align 16)
+; 32BIT-DAG: renamable $r3 = LWZ 4, %fixed-stack.0 :: (load (s32) from %ir.argp.cur142 + 4)
+; 32BIT-DAG: renamable $r4 = ADDI %fixed-stack.0, 0
+; 32BIT-DAG: STW killed renamable $r4, 0, %stack.0.arg1 :: (store (s32) into %ir.arg1)
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f1, killed renamable $f2, implicit $rm
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f0, killed renamable $f3, implicit $rm
-; 32BIT-DAG: STW renamable $r3, 0, %stack.2 :: (store (s32) into %stack.2, align 8)
+; 32BIT-DAG: STW renamable $r3, 4, %stack.2 :: (store (s32) into %stack.2 + 4)
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f0, killed renamable $f4, implicit $rm
-; 32BIT-DAG: renamable $r4 = LWZ 4, %fixed-stack.0 :: (load (s32) from %ir.argp.cur142 + 4)
+; 32BIT-DAG: renamable $r4 = LWZ 0, %fixed-stack.0 :: (load (s32) from %ir.argp.cur142, align 16)
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f0, killed renamable $f5, implicit $rm
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f0, killed renamable $f6, implicit $rm
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f0, killed renamable $f7, implicit $rm
-; 32BIT-DAG: STW renamable $r4, 4, %stack.2 :: (store (s32) into %stack.2 + 4)
+; 32BIT-DAG: STW renamable $r4, 0, %stack.2 :: (store (s32) into %stack.2, align 8)
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f0, killed renamable $f8, implicit $rm
; 32BIT-DAG: renamable $f1 = LFD 0, %stack.2 :: (load (s64) from %stack.2)
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f0, killed renamable $f9, implicit $rm
-; 32BIT-DAG: STW killed renamable $r3, 0, %stack.3 :: (store (s32) into %stack.3, align 8)
+; 32BIT-DAG: STW killed renamable $r3, 4, %stack.3 :: (store (s32) into %stack.3 + 4)
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f0, killed renamable $f10, implicit $rm
-; 32BIT-DAG: STW killed renamable $r4, 4, %stack.3 :: (store (s32) into %stack.3 + 4)
+; 32BIT-DAG: STW killed renamable $r4, 0, %stack.3 :: (store (s32) into %stack.3, align 8)
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f0, killed renamable $f11, implicit $rm
; 32BIT-DAG: renamable $f2 = LFD 0, %stack.3 :: (load (s64) from %stack.3)
; 32BIT-DAG: renamable $f0 = nofpexcept FADD killed renamable $f0, killed renamable $f12, implicit $rm
diff --git a/llvm/test/CodeGen/PowerPC/combine-fneg.ll b/llvm/test/CodeGen/PowerPC/combine-fneg.ll
index 3015e68..8c9e1c2 100644
--- a/llvm/test/CodeGen/PowerPC/combine-fneg.ll
+++ b/llvm/test/CodeGen/PowerPC/combine-fneg.ll
@@ -12,10 +12,10 @@
; CHECK-NEXT: lxvd2x 2, 0, 3
; CHECK-NEXT: xvredp 1, 0
; CHECK-NEXT: xxlor 3, 2, 2
-; CHECK-NEXT: xvmaddadp 3, 0, 1
-; CHECK-NEXT: xvnmsubadp 1, 1, 3
-; CHECK-NEXT: xvmaddadp 2, 0, 1
-; CHECK-NEXT: xvmsubadp 1, 1, 2
+; CHECK-NEXT: xvnmsubadp 3, 0, 1
+; CHECK-NEXT: xvmaddadp 1, 1, 3
+; CHECK-NEXT: xvnmsubadp 2, 0, 1
+; CHECK-NEXT: xvnmaddadp 1, 1, 2
; CHECK-NEXT: xvmuldp 34, 34, 1
; CHECK-NEXT: xvmuldp 35, 35, 1
; CHECK-NEXT: blr
diff --git a/llvm/test/CodeGen/PowerPC/select_const.ll b/llvm/test/CodeGen/PowerPC/select_const.ll
index da2f73d..1e2de35 100644
--- a/llvm/test/CodeGen/PowerPC/select_const.ll
+++ b/llvm/test/CodeGen/PowerPC/select_const.ll
@@ -610,13 +610,24 @@
}
define i8 @shl_constant_sel_constants(i1 %cond) {
-; ALL-LABEL: shl_constant_sel_constants:
-; ALL: # %bb.0:
-; ALL-NEXT: clrlwi 3, 3, 31
-; ALL-NEXT: li 4, 1
-; ALL-NEXT: xori 3, 3, 3
-; ALL-NEXT: slw 3, 4, 3
-; ALL-NEXT: blr
+; ISEL-LABEL: shl_constant_sel_constants:
+; ISEL: # %bb.0:
+; ISEL-NEXT: andi. 3, 3, 1
+; ISEL-NEXT: li 4, 4
+; ISEL-NEXT: li 3, 8
+; ISEL-NEXT: iselgt 3, 4, 3
+; ISEL-NEXT: blr
+;
+; NO_ISEL-LABEL: shl_constant_sel_constants:
+; NO_ISEL: # %bb.0:
+; NO_ISEL-NEXT: andi. 3, 3, 1
+; NO_ISEL-NEXT: li 4, 4
+; NO_ISEL-NEXT: li 3, 8
+; NO_ISEL-NEXT: bc 12, 1, .LBB37_1
+; NO_ISEL-NEXT: blr
+; NO_ISEL-NEXT: .LBB37_1:
+; NO_ISEL-NEXT: addi 3, 4, 0
+; NO_ISEL-NEXT: blr
%sel = select i1 %cond, i8 2, i8 3
%bo = shl i8 1, %sel
ret i8 %bo
@@ -647,13 +658,24 @@
}
define i8 @lshr_constant_sel_constants(i1 %cond) {
-; ALL-LABEL: lshr_constant_sel_constants:
-; ALL: # %bb.0:
-; ALL-NEXT: clrlwi 3, 3, 31
-; ALL-NEXT: li 4, 64
-; ALL-NEXT: xori 3, 3, 3
-; ALL-NEXT: srw 3, 4, 3
-; ALL-NEXT: blr
+; ISEL-LABEL: lshr_constant_sel_constants:
+; ISEL: # %bb.0:
+; ISEL-NEXT: andi. 3, 3, 1
+; ISEL-NEXT: li 4, 16
+; ISEL-NEXT: li 3, 8
+; ISEL-NEXT: iselgt 3, 4, 3
+; ISEL-NEXT: blr
+;
+; NO_ISEL-LABEL: lshr_constant_sel_constants:
+; NO_ISEL: # %bb.0:
+; NO_ISEL-NEXT: andi. 3, 3, 1
+; NO_ISEL-NEXT: li 4, 16
+; NO_ISEL-NEXT: li 3, 8
+; NO_ISEL-NEXT: bc 12, 1, .LBB39_1
+; NO_ISEL-NEXT: blr
+; NO_ISEL-NEXT: .LBB39_1:
+; NO_ISEL-NEXT: addi 3, 4, 0
+; NO_ISEL-NEXT: blr
%sel = select i1 %cond, i8 2, i8 3
%bo = lshr i8 64, %sel
ret i8 %bo
@@ -663,7 +685,7 @@
define i8 @sel_constants_ashr_constant(i1 %cond) {
; ALL-LABEL: sel_constants_ashr_constant:
; ALL: # %bb.0:
-; ALL-NEXT: clrlwi 3, 3, 31
+; ALL-NEXT: clrldi 3, 3, 63
; ALL-NEXT: neg 3, 3
; ALL-NEXT: blr
%sel = select i1 %cond, i8 -4, i8 23
@@ -672,13 +694,24 @@
}
define i8 @ashr_constant_sel_constants(i1 %cond) {
-; ALL-LABEL: ashr_constant_sel_constants:
-; ALL: # %bb.0:
-; ALL-NEXT: clrlwi 3, 3, 31
-; ALL-NEXT: li 4, -128
-; ALL-NEXT: xori 3, 3, 3
-; ALL-NEXT: sraw 3, 4, 3
-; ALL-NEXT: blr
+; ISEL-LABEL: ashr_constant_sel_constants:
+; ISEL: # %bb.0:
+; ISEL-NEXT: andi. 3, 3, 1
+; ISEL-NEXT: li 4, -32
+; ISEL-NEXT: li 3, -16
+; ISEL-NEXT: iselgt 3, 4, 3
+; ISEL-NEXT: blr
+;
+; NO_ISEL-LABEL: ashr_constant_sel_constants:
+; NO_ISEL: # %bb.0:
+; NO_ISEL-NEXT: andi. 3, 3, 1
+; NO_ISEL-NEXT: li 4, -32
+; NO_ISEL-NEXT: li 3, -16
+; NO_ISEL-NEXT: bc 12, 1, .LBB41_1
+; NO_ISEL-NEXT: blr
+; NO_ISEL-NEXT: .LBB41_1:
+; NO_ISEL-NEXT: addi 3, 4, 0
+; NO_ISEL-NEXT: blr
%sel = select i1 %cond, i8 2, i8 3
%bo = ashr i8 128, %sel
ret i8 %bo
diff --git a/llvm/test/CodeGen/RISCV/mul.ll b/llvm/test/CodeGen/RISCV/mul.ll
index 341db9a..49c8d16 100644
--- a/llvm/test/CodeGen/RISCV/mul.ll
+++ b/llvm/test/CodeGen/RISCV/mul.ll
@@ -1299,8 +1299,9 @@
; RV32IM-NEXT: li a5, -63
; RV32IM-NEXT: mulhu a6, a3, a5
; RV32IM-NEXT: slli a7, a4, 6
-; RV32IM-NEXT: sub a7, a4, a7
-; RV32IM-NEXT: add a6, a7, a6
+; RV32IM-NEXT: sub a7, a7, a4
+; RV32IM-NEXT: sub a6, a6, a7
+; RV32IM-NEXT: neg a7, a7
; RV32IM-NEXT: sltu a7, a6, a7
; RV32IM-NEXT: mulhu t0, a4, a5
; RV32IM-NEXT: add a7, t0, a7
@@ -1314,8 +1315,8 @@
; RV32IM-NEXT: sub t4, t1, a4
; RV32IM-NEXT: slli t5, a1, 6
; RV32IM-NEXT: sub t6, a1, a3
-; RV32IM-NEXT: sub t5, t6, t5
-; RV32IM-NEXT: add t6, t4, t5
+; RV32IM-NEXT: sub t5, t5, t6
+; RV32IM-NEXT: sub t6, t4, t5
; RV32IM-NEXT: sltu s0, t6, t4
; RV32IM-NEXT: neg s1, a4
; RV32IM-NEXT: sltu t4, t4, s1
@@ -1323,7 +1324,6 @@
; RV32IM-NEXT: mulhu t1, a4, t2
; RV32IM-NEXT: add a7, t1, a7
; RV32IM-NEXT: add a7, a7, t4
-; RV32IM-NEXT: sltu t0, t5, t0
; RV32IM-NEXT: slli t1, a2, 6
; RV32IM-NEXT: sub a2, a2, t1
; RV32IM-NEXT: mulhu a5, a1, a5
@@ -1332,7 +1332,9 @@
; RV32IM-NEXT: add a4, a3, a4
; RV32IM-NEXT: sub a1, t3, a4
; RV32IM-NEXT: add a1, a1, a2
-; RV32IM-NEXT: add a1, a1, t0
+; RV32IM-NEXT: neg a2, t5
+; RV32IM-NEXT: sltu a2, a2, t0
+; RV32IM-NEXT: add a1, a1, a2
; RV32IM-NEXT: add a1, a7, a1
; RV32IM-NEXT: add a1, a1, s0
; RV32IM-NEXT: slli a2, a3, 6
diff --git a/llvm/test/CodeGen/RISCV/pr58511.ll b/llvm/test/CodeGen/RISCV/pr58511.ll
index df02d77..6280903 100644
--- a/llvm/test/CodeGen/RISCV/pr58511.ll
+++ b/llvm/test/CodeGen/RISCV/pr58511.ll
@@ -4,12 +4,12 @@
define i32 @f(i1 %0, i32 %1, ptr %2) {
; CHECK-LABEL: f:
; CHECK: # %bb.0: # %BB
-; CHECK-NEXT: slli a0, a0, 63
-; CHECK-NEXT: srai a0, a0, 63
; CHECK-NEXT: lui a3, 4097
; CHECK-NEXT: addiw a3, a3, -2047
-; CHECK-NEXT: or a0, a0, a3
; CHECK-NEXT: mul a1, a1, a3
+; CHECK-NEXT: slli a0, a0, 63
+; CHECK-NEXT: srai a0, a0, 63
+; CHECK-NEXT: or a0, a0, a3
; CHECK-NEXT: sw a1, 0(a2)
; CHECK-NEXT: ret
BB:
@@ -23,12 +23,12 @@
define i32 @g(i1 %0, i32 %1, ptr %2) {
; CHECK-LABEL: g:
; CHECK: # %bb.0: # %BB
-; CHECK-NEXT: andi a0, a0, 1
-; CHECK-NEXT: addi a0, a0, -1
; CHECK-NEXT: lui a3, 4097
; CHECK-NEXT: addiw a3, a3, -2047
-; CHECK-NEXT: or a0, a0, a3
; CHECK-NEXT: mul a1, a1, a3
+; CHECK-NEXT: andi a0, a0, 1
+; CHECK-NEXT: addi a0, a0, -1
+; CHECK-NEXT: or a0, a0, a3
; CHECK-NEXT: sw a1, 0(a2)
; CHECK-NEXT: ret
BB:
diff --git a/llvm/test/CodeGen/SystemZ/pr36164.ll b/llvm/test/CodeGen/SystemZ/pr36164.ll
index 4ed07c8..e4ac2f5e 100644
--- a/llvm/test/CodeGen/SystemZ/pr36164.ll
+++ b/llvm/test/CodeGen/SystemZ/pr36164.ll
@@ -17,17 +17,11 @@
; CHECK: # %bb.0:
; CHECK-NEXT: lhi %r0, 1
; CHECK-NEXT: larl %r1, g_938
-; CHECK-NEXT: lhi %r2, 3
+; CHECK-NEXT: lhi %r2, 0
; CHECK-NEXT: lhi %r3, 4
; CHECK-NEXT: larl %r4, g_11
; CHECK-NEXT: .LBB0_1: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: strl %r0, g_73
-; CHECK-NEXT: lrl %r5, g_832
-; CHECK-NEXT: lrl %r5, g_832
-; CHECK-NEXT: lrl %r5, g_832
-; CHECK-NEXT: lrl %r5, g_832
-; CHECK-NEXT: lrl %r5, g_832
-; CHECK-NEXT: lrl %r5, g_832
; CHECK-NEXT: strl %r2, g_69
; CHECK-NEXT: lrl %r5, g_832
; CHECK-NEXT: lrl %r5, g_832
@@ -44,6 +38,12 @@
; CHECK-NEXT: lrl %r5, g_832
; CHECK-NEXT: lrl %r5, g_832
; CHECK-NEXT: lrl %r5, g_832
+; CHECK-NEXT: lrl %r5, g_832
+; CHECK-NEXT: lrl %r5, g_832
+; CHECK-NEXT: lrl %r5, g_832
+; CHECK-NEXT: lrl %r5, g_832
+; CHECK-NEXT: lrl %r5, g_832
+; CHECK-NEXT: lrl %r5, g_832
; CHECK-NEXT: agsi 0(%r1), 24
; CHECK-NEXT: lrl %r5, g_832
; CHECK-NEXT: strl %r3, g_69
diff --git a/llvm/test/CodeGen/Thumb2/mve-vst3.ll b/llvm/test/CodeGen/Thumb2/mve-vst3.ll
index 959a0bf..65fb570 100644
--- a/llvm/test/CodeGen/Thumb2/mve-vst3.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-vst3.ll
@@ -8,17 +8,15 @@
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: .save {r4, lr}
; CHECK-NEXT: push {r4, lr}
-; CHECK-NEXT: ldrd lr, r12, [r0]
-; CHECK-NEXT: ldrd r3, r2, [r0, #8]
+; CHECK-NEXT: ldrd r12, r3, [r0]
+; CHECK-NEXT: ldrd lr, r2, [r0, #8]
; CHECK-NEXT: ldrd r4, r0, [r0, #16]
-; CHECK-NEXT: vmov q1[2], q1[0], lr, r3
-; CHECK-NEXT: str r2, [r1, #16]
-; CHECK-NEXT: vmov.32 q0[0], r4
-; CHECK-NEXT: vmov q1[3], q1[1], r12, r2
-; CHECK-NEXT: vmov.32 q0[1], r0
+; CHECK-NEXT: vmov.32 q1[1], r3
+; CHECK-NEXT: vmov q1[2], q1[0], r12, lr
+; CHECK-NEXT: strd r2, r0, [r1, #16]
+; CHECK-NEXT: vmov q0[2], q0[0], r4, r0
; CHECK-NEXT: vmov.f32 s8, s4
; CHECK-NEXT: vmov.f32 s9, s6
-; CHECK-NEXT: str r0, [r1, #20]
; CHECK-NEXT: vmov.f32 s10, s0
; CHECK-NEXT: vmov.f32 s11, s5
; CHECK-NEXT: vstrw.32 q2, [r1]
diff --git a/llvm/test/CodeGen/X86/2011-10-19-LegelizeLoad.ll b/llvm/test/CodeGen/X86/2011-10-19-LegelizeLoad.ll
index 65ff22f..9910295 100644
--- a/llvm/test/CodeGen/X86/2011-10-19-LegelizeLoad.ll
+++ b/llvm/test/CodeGen/X86/2011-10-19-LegelizeLoad.ll
@@ -17,12 +17,12 @@
define dso_local i32 @main() nounwind uwtable {
; CHECK-LABEL: main:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: movl i(%rip), %esi
-; CHECK-NEXT: movl j(%rip), %eax
-; CHECK-NEXT: movl %esi, %edx
-; CHECK-NEXT: shrl $8, %edx
+; CHECK-NEXT: movq i(%rip), %rsi
+; CHECK-NEXT: movq j(%rip), %rax
+; CHECK-NEXT: movq %rsi, %rdx
+; CHECK-NEXT: shrq $8, %rdx
; CHECK-NEXT: movsbl %al, %ecx
-; CHECK-NEXT: shrl $8, %eax
+; CHECK-NEXT: shrq $8, %rax
; CHECK-NEXT: cbtw
; CHECK-NEXT: idivb %dl
; CHECK-NEXT: movl %eax, %edx
diff --git a/llvm/test/CodeGen/X86/2012-08-07-CmpISelBug.ll b/llvm/test/CodeGen/X86/2012-08-07-CmpISelBug.ll
index 418cbb0..c00e863 100644
--- a/llvm/test/CodeGen/X86/2012-08-07-CmpISelBug.ll
+++ b/llvm/test/CodeGen/X86/2012-08-07-CmpISelBug.ll
@@ -8,11 +8,13 @@
define void @foo(i8 %arg4, i32 %arg5, ptr %arg14) nounwind {
; CHECK-LABEL: foo:
; CHECK: ## %bb.0: ## %bb
+; CHECK-NEXT: ## kill: def $edi killed $edi def $rdi
; CHECK-NEXT: andl $32, %edi
-; CHECK-NEXT: movb $81, %al
-; CHECK-NEXT: subb %dil, %al
-; CHECK-NEXT: testl %esi, %edi
+; CHECK-NEXT: leal 13(%rdi), %eax
+; CHECK-NEXT: xorb $-14, %al
+; CHECK-NEXT: addb $82, %al
; CHECK-NEXT: movzbl %al, %eax
+; CHECK-NEXT: testl %esi, %edi
; CHECK-NEXT: movl $1, %ecx
; CHECK-NEXT: cmovnel %eax, %ecx
; CHECK-NEXT: xorb $81, %cl
diff --git a/llvm/test/CodeGen/X86/addcarry.ll b/llvm/test/CodeGen/X86/addcarry.ll
index 0b7ba92..0410b7e 100644
--- a/llvm/test/CodeGen/X86/addcarry.ll
+++ b/llvm/test/CodeGen/X86/addcarry.ll
@@ -742,32 +742,38 @@
define i32 @add_U320_without_i128_add(ptr nocapture dereferenceable(40) %0, i64 %1, i64 %2, i64 %3, i64 %4, i64 %5) nounwind {
; CHECK-LABEL: add_U320_without_i128_add:
; CHECK: # %bb.0:
+; CHECK-NEXT: pushq %r14
+; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: movq 16(%rdi), %rax
-; CHECK-NEXT: movq 24(%rdi), %r10
-; CHECK-NEXT: movq 32(%rdi), %r11
+; CHECK-NEXT: leaq (%rax,%rcx), %r10
; CHECK-NEXT: addq %rsi, (%rdi)
; CHECK-NEXT: adcq %rdx, 8(%rdi)
; CHECK-NEXT: movq %rax, %rdx
; CHECK-NEXT: adcq %rcx, %rdx
+; CHECK-NEXT: movq 24(%rdi), %rsi
+; CHECK-NEXT: leaq (%r8,%rsi), %r11
+; CHECK-NEXT: xorl %ebx, %ebx
+; CHECK-NEXT: cmpq %r10, %rdx
+; CHECK-NEXT: setb %bl
; CHECK-NEXT: addq %rcx, %rax
-; CHECK-NEXT: movq %r10, %rcx
-; CHECK-NEXT: adcq %r8, %rcx
-; CHECK-NEXT: cmpq %rax, %rdx
-; CHECK-NEXT: adcq $0, %rcx
-; CHECK-NEXT: leaq (%r11,%r9), %rsi
-; CHECK-NEXT: addq %r8, %r10
-; CHECK-NEXT: movq %r11, %r8
-; CHECK-NEXT: adcq %r9, %r8
-; CHECK-NEXT: cmpq %r10, %rcx
-; CHECK-NEXT: adcq $0, %r8
+; CHECK-NEXT: adcq %r11, %rbx
+; CHECK-NEXT: movq 32(%rdi), %rcx
+; CHECK-NEXT: leaq (%r9,%rcx), %r10
+; CHECK-NEXT: xorl %r14d, %r14d
+; CHECK-NEXT: cmpq %r11, %rbx
+; CHECK-NEXT: setb %r14b
+; CHECK-NEXT: addq %rsi, %r8
+; CHECK-NEXT: adcq %r10, %r14
; CHECK-NEXT: xorl %eax, %eax
-; CHECK-NEXT: cmpq %rsi, %r8
+; CHECK-NEXT: cmpq %r10, %r14
; CHECK-NEXT: setb %al
-; CHECK-NEXT: addq %r9, %r11
+; CHECK-NEXT: addq %rcx, %r9
; CHECK-NEXT: movq %rdx, 16(%rdi)
-; CHECK-NEXT: movq %rcx, 24(%rdi)
-; CHECK-NEXT: movq %r8, 32(%rdi)
+; CHECK-NEXT: movq %rbx, 24(%rdi)
+; CHECK-NEXT: movq %r14, 32(%rdi)
; CHECK-NEXT: adcl $0, %eax
+; CHECK-NEXT: popq %rbx
+; CHECK-NEXT: popq %r14
; CHECK-NEXT: retq
%7 = load i64, ptr %0, align 8
%8 = getelementptr inbounds %struct.U320, ptr %0, i64 0, i32 0, i64 1
diff --git a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
index d61e33c..9ecc89b 100644
--- a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast.ll
@@ -3479,26 +3479,26 @@
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vmovdqa 48(%rdi), %xmm1
-; AVX-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX-NEXT: vpaddb (%rsi), %xmm0, %xmm0
-; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,0,0,0]
+; AVX-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[2,3,6,7,10,11,14,15,u,u,u,u,u,u,u,u]
-; AVX-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
-; AVX-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX-NEXT: vpaddb (%rdx), %xmm0, %xmm0
-; AVX-NEXT: vpaddb 32(%rdx), %xmm2, %xmm1
-; AVX-NEXT: vpaddb 16(%rdx), %xmm2, %xmm2
-; AVX-NEXT: vmovdqa %xmm2, 16(%rcx)
-; AVX-NEXT: vmovdqa %xmm1, 32(%rcx)
-; AVX-NEXT: vmovdqa %xmm0, (%rcx)
+; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm0[0,0,0,0,4,5,6,7]
+; AVX-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; AVX-NEXT: vpaddb (%rdx), %xmm1, %xmm1
+; AVX-NEXT: vpaddb 32(%rdx), %xmm0, %xmm2
+; AVX-NEXT: vpaddb 16(%rdx), %xmm0, %xmm0
+; AVX-NEXT: vmovdqa %xmm0, 16(%rcx)
+; AVX-NEXT: vmovdqa %xmm2, 32(%rcx)
+; AVX-NEXT: vmovdqa %xmm1, (%rcx)
; AVX-NEXT: retq
;
; AVX2-LABEL: vec384_i16_widen_to_i32_factor2_broadcast_to_v12i32_factor12:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm1
-; AVX2-NEXT: vpaddb (%rsi), %xmm0, %xmm0
; AVX2-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
+; AVX2-NEXT: vpaddb (%rsi), %xmm0, %xmm0
; AVX2-NEXT: vpbroadcastw %xmm0, %ymm0
; AVX2-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7],ymm0[8],ymm1[9],ymm0[10],ymm1[11],ymm0[12],ymm1[13],ymm0[14],ymm1[15]
; AVX2-NEXT: vpaddb (%rdx), %ymm1, %ymm1
@@ -3627,9 +3627,9 @@
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm1
+; AVX2-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX2-NEXT: vpaddb (%rsi), %xmm0, %xmm0
; AVX2-NEXT: vpbroadcastw %xmm0, %xmm0
-; AVX2-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2],xmm0[3],xmm1[4,5],xmm0[6],xmm1[7]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm1
; AVX2-NEXT: vpaddb (%rdx), %ymm1, %ymm1
@@ -3757,15 +3757,15 @@
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm1
-; AVX2-NEXT: vpaddb (%rsi), %xmm0, %xmm0
-; AVX2-NEXT: vpbroadcastw %xmm0, %xmm2
; AVX2-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX2-NEXT: vpbroadcastq %xmm0, %ymm0
-; AVX2-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7],ymm0[8],ymm1[9,10,11],ymm0[12],ymm1[13,14,15]
-; AVX2-NEXT: vpaddb (%rdx), %ymm0, %ymm0
-; AVX2-NEXT: vpaddb 32(%rdx), %ymm2, %ymm1
-; AVX2-NEXT: vmovdqa %ymm0, (%rcx)
-; AVX2-NEXT: vmovdqa %ymm1, 32(%rcx)
+; AVX2-NEXT: vpaddb (%rsi), %xmm0, %xmm0
+; AVX2-NEXT: vpbroadcastq %xmm0, %ymm2
+; AVX2-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3],ymm2[4],ymm1[5,6,7],ymm2[8],ymm1[9,10,11],ymm2[12],ymm1[13,14,15]
+; AVX2-NEXT: vpbroadcastw %xmm0, %xmm0
+; AVX2-NEXT: vpaddb (%rdx), %ymm1, %ymm1
+; AVX2-NEXT: vpaddb 32(%rdx), %ymm0, %ymm0
+; AVX2-NEXT: vmovdqa %ymm1, (%rcx)
+; AVX2-NEXT: vmovdqa %ymm0, 32(%rcx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -3874,15 +3874,15 @@
; AVX-NEXT: vmovdqa 48(%rdi), %xmm1
; AVX-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX-NEXT: vpaddb (%rsi), %xmm0, %xmm0
-; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,0,1,1]
-; AVX-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[0,0,0,0]
-; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0],xmm1[1,2,3,4,5],xmm3[6],xmm1[7]
-; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,0,0,0]
+; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3,4,5],xmm2[6],xmm1[7]
+; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,1]
+; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; AVX-NEXT: vpaddb (%rdx), %xmm1, %xmm1
-; AVX-NEXT: vpaddb 32(%rdx), %xmm2, %xmm2
-; AVX-NEXT: vpaddb 16(%rdx), %xmm0, %xmm0
-; AVX-NEXT: vmovdqa %xmm0, 16(%rcx)
-; AVX-NEXT: vmovdqa %xmm2, 32(%rcx)
+; AVX-NEXT: vpaddb 32(%rdx), %xmm0, %xmm0
+; AVX-NEXT: vpaddb 16(%rdx), %xmm2, %xmm2
+; AVX-NEXT: vmovdqa %xmm2, 16(%rcx)
+; AVX-NEXT: vmovdqa %xmm0, 32(%rcx)
; AVX-NEXT: vmovdqa %xmm1, (%rcx)
; AVX-NEXT: retq
;
@@ -3890,9 +3890,9 @@
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm1
+; AVX2-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX2-NEXT: vpaddb (%rsi), %xmm0, %xmm0
; AVX2-NEXT: vpbroadcastw %xmm0, %xmm0
-; AVX2-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX2-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3,4,5],xmm0[6],xmm1[7]
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm1
; AVX2-NEXT: vpaddb (%rdx), %ymm1, %ymm1
@@ -4243,17 +4243,17 @@
; AVX-NEXT: vmovdqa 48(%rdi), %xmm1
; AVX-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX-NEXT: vpaddb (%rsi), %xmm0, %xmm0
-; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,1]
-; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[1,3],ymm0[4,4],ymm1[5,7]
-; AVX-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,2,1,3,4,6,5,7]
-; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT: vpaddb 16(%rdx), %xmm1, %xmm1
-; AVX-NEXT: vpaddb (%rdx), %xmm0, %xmm0
-; AVX-NEXT: vpaddb 32(%rdx), %xmm2, %xmm2
-; AVX-NEXT: vmovdqa %xmm2, 32(%rcx)
-; AVX-NEXT: vmovdqa %xmm0, (%rcx)
-; AVX-NEXT: vmovdqa %xmm1, 16(%rcx)
+; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm2
+; AVX-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[1,3],ymm2[4,4],ymm1[5,7]
+; AVX-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,2,1,3,4,6,5,7]
+; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX-NEXT: vpaddb 16(%rdx), %xmm2, %xmm2
+; AVX-NEXT: vpaddb (%rdx), %xmm1, %xmm1
+; AVX-NEXT: vpaddb 32(%rdx), %xmm0, %xmm0
+; AVX-NEXT: vmovdqa %xmm0, 32(%rcx)
+; AVX-NEXT: vmovdqa %xmm1, (%rcx)
+; AVX-NEXT: vmovdqa %xmm2, 16(%rcx)
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
@@ -4261,15 +4261,15 @@
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-NEXT: vmovdqa 48(%rdi), %xmm1
-; AVX2-NEXT: vpaddb (%rsi), %xmm0, %xmm0
-; AVX2-NEXT: vpbroadcastd %xmm0, %xmm2
; AVX2-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
-; AVX2-NEXT: vpbroadcastq %xmm0, %ymm0
-; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4,5,6,7]
-; AVX2-NEXT: vpaddb (%rdx), %ymm0, %ymm0
-; AVX2-NEXT: vpaddb 32(%rdx), %ymm2, %ymm1
-; AVX2-NEXT: vmovdqa %ymm0, (%rcx)
-; AVX2-NEXT: vmovdqa %ymm1, 32(%rcx)
+; AVX2-NEXT: vpaddb (%rsi), %xmm0, %xmm0
+; AVX2-NEXT: vpbroadcastq %xmm0, %ymm2
+; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[3],ymm2[4,5,6,7]
+; AVX2-NEXT: vpbroadcastd %xmm0, %xmm0
+; AVX2-NEXT: vpaddb (%rdx), %ymm1, %ymm1
+; AVX2-NEXT: vpaddb 32(%rdx), %ymm0, %ymm0
+; AVX2-NEXT: vmovdqa %ymm1, (%rcx)
+; AVX2-NEXT: vmovdqa %ymm0, 32(%rcx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -4371,17 +4371,17 @@
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa (%rdi), %xmm0
; AVX-NEXT: vmovdqa 48(%rdi), %xmm1
-; AVX-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX-NEXT: vpaddb (%rsi), %xmm0, %xmm0
-; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,0,1,1]
+; AVX-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,1]
+; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; AVX-NEXT: vpaddb (%rdx), %xmm1, %xmm1
-; AVX-NEXT: vpaddb 32(%rdx), %xmm2, %xmm2
-; AVX-NEXT: vpaddb 16(%rdx), %xmm0, %xmm0
-; AVX-NEXT: vmovdqa %xmm0, 16(%rcx)
-; AVX-NEXT: vmovdqa %xmm2, 32(%rcx)
+; AVX-NEXT: vpaddb 32(%rdx), %xmm0, %xmm0
+; AVX-NEXT: vpaddb 16(%rdx), %xmm2, %xmm2
+; AVX-NEXT: vmovdqa %xmm2, 16(%rcx)
+; AVX-NEXT: vmovdqa %xmm0, 32(%rcx)
; AVX-NEXT: vmovdqa %xmm1, (%rcx)
; AVX-NEXT: retq
;
@@ -4389,16 +4389,16 @@
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %ymm0
; AVX2-NEXT: vmovdqa 32(%rdi), %ymm1
-; AVX2-NEXT: vpaddb 32(%rsi), %ymm1, %ymm1
; AVX2-NEXT: vpaddb (%rsi), %ymm0, %ymm0
-; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,0,1,1]
-; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm1 = [0,5,6,0]
-; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0
-; AVX2-NEXT: vpaddb (%rdx), %ymm0, %ymm0
-; AVX2-NEXT: vpaddb 32(%rdx), %ymm2, %ymm1
-; AVX2-NEXT: vmovdqa %ymm0, (%rcx)
-; AVX2-NEXT: vmovdqa %ymm1, 32(%rcx)
+; AVX2-NEXT: vpaddb 32(%rsi), %ymm1, %ymm1
+; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [0,5,6,0]
+; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1
+; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX2-NEXT: vpaddb (%rdx), %ymm1, %ymm1
+; AVX2-NEXT: vpaddb 32(%rdx), %ymm0, %ymm0
+; AVX2-NEXT: vmovdqa %ymm1, (%rcx)
+; AVX2-NEXT: vmovdqa %ymm0, 32(%rcx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
index 64e2afc..5047041 100644
--- a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
+++ b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
@@ -3647,10 +3647,10 @@
; AVX-NEXT: vmovdqa (%rdi), %xmm2
; AVX-NEXT: vmovdqa 16(%rdi), %xmm3
; AVX-NEXT: vpaddb 48(%rsi), %xmm3, %xmm3
-; AVX-NEXT: vpaddb 32(%rsi), %xmm2, %xmm2
; AVX-NEXT: vpaddb 16(%rsi), %xmm0, %xmm0
-; AVX-NEXT: vmovdqa %xmm0, 16(%rdx)
+; AVX-NEXT: vpaddb 32(%rsi), %xmm2, %xmm2
; AVX-NEXT: vmovdqa %xmm2, 32(%rdx)
+; AVX-NEXT: vmovdqa %xmm0, 16(%rdx)
; AVX-NEXT: vmovdqa %xmm3, 48(%rdx)
; AVX-NEXT: vmovdqa %xmm1, (%rdx)
; AVX-NEXT: vzeroupper
@@ -3833,10 +3833,10 @@
; AVX-NEXT: vmovdqa (%rdi), %xmm2
; AVX-NEXT: vmovdqa 16(%rdi), %xmm3
; AVX-NEXT: vpaddb 48(%rsi), %xmm3, %xmm3
-; AVX-NEXT: vpaddb 32(%rsi), %xmm2, %xmm2
; AVX-NEXT: vpaddb 16(%rsi), %xmm0, %xmm0
-; AVX-NEXT: vmovdqa %xmm0, 16(%rdx)
+; AVX-NEXT: vpaddb 32(%rsi), %xmm2, %xmm2
; AVX-NEXT: vmovdqa %xmm2, 32(%rdx)
+; AVX-NEXT: vmovdqa %xmm0, 16(%rdx)
; AVX-NEXT: vmovdqa %xmm3, 48(%rdx)
; AVX-NEXT: vmovdqa %xmm1, (%rdx)
; AVX-NEXT: vzeroupper
diff --git a/llvm/test/CodeGen/X86/avx512-mask-op.ll b/llvm/test/CodeGen/X86/avx512-mask-op.ll
index d9f051c..1c39789 100644
--- a/llvm/test/CodeGen/X86/avx512-mask-op.ll
+++ b/llvm/test/CodeGen/X86/avx512-mask-op.ll
@@ -74,7 +74,8 @@
; X86-LABEL: mask8_zext:
; X86: ## %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: xorl $255, %eax
+; X86-NEXT: notb %al
+; X86-NEXT: movzbl %al, %eax
; X86-NEXT: retl
%m0 = bitcast i8 %x to <8 x i1>
%m1 = xor <8 x i1> %m0, <i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1, i1 -1>
diff --git a/llvm/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll b/llvm/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
index 9cb3cea..fa463d3 100644
--- a/llvm/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
+++ b/llvm/test/CodeGen/X86/avx512bw-intrinsics-upgrade.ll
@@ -1967,16 +1967,16 @@
; X86-NEXT: adcl %ecx, %edx # encoding: [0x11,0xca]
; X86-NEXT: vpcmpnltb %zmm1, %zmm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x3f,0xc1,0x05]
; X86-NEXT: kshiftrq $32, %k0, %k2 # encoding: [0xc4,0xe3,0xf9,0x31,0xd0,0x20]
-; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
-; X86-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
-; X86-NEXT: addl %eax, %esi # encoding: [0x01,0xc6]
-; X86-NEXT: adcl %edx, %ecx # encoding: [0x11,0xd1]
+; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
+; X86-NEXT: addl %eax, %ecx # encoding: [0x01,0xc1]
+; X86-NEXT: kmovd %k2, %esi # encoding: [0xc5,0xfb,0x93,0xf2]
+; X86-NEXT: adcl %edx, %esi # encoding: [0x11,0xd6]
; X86-NEXT: vpcmpgtb %zmm1, %zmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0x64,0xc1]
; X86-NEXT: kshiftrq $32, %k0, %k1 # encoding: [0xc4,0xe3,0xf9,0x31,0xc8,0x20]
-; X86-NEXT: kmovd %k1, %edx # encoding: [0xc5,0xfb,0x93,0xd1]
; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
-; X86-NEXT: addl %esi, %eax # encoding: [0x01,0xf0]
-; X86-NEXT: adcl %ecx, %edx # encoding: [0x11,0xca]
+; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
+; X86-NEXT: kmovd %k1, %edx # encoding: [0xc5,0xfb,0x93,0xd1]
+; X86-NEXT: adcl %esi, %edx # encoding: [0x11,0xf2]
; X86-NEXT: addl {{[0-9]+}}(%esp), %eax # encoding: [0x03,0x44,0x24,0x08]
; X86-NEXT: adcl {{[0-9]+}}(%esp), %edx # encoding: [0x13,0x54,0x24,0x0c]
; X86-NEXT: popl %esi # encoding: [0x5e]
@@ -2140,16 +2140,16 @@
; X86-NEXT: adcl %ecx, %edx # encoding: [0x11,0xca]
; X86-NEXT: vpcmpnltub %zmm1, %zmm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x3e,0xc1,0x05]
; X86-NEXT: kshiftrq $32, %k0, %k2 # encoding: [0xc4,0xe3,0xf9,0x31,0xd0,0x20]
-; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
-; X86-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
-; X86-NEXT: addl %eax, %esi # encoding: [0x01,0xc6]
-; X86-NEXT: adcl %edx, %ecx # encoding: [0x11,0xd1]
+; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
+; X86-NEXT: addl %eax, %ecx # encoding: [0x01,0xc1]
+; X86-NEXT: kmovd %k2, %esi # encoding: [0xc5,0xfb,0x93,0xf2]
+; X86-NEXT: adcl %edx, %esi # encoding: [0x11,0xd6]
; X86-NEXT: vpcmpnleub %zmm1, %zmm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x3e,0xc1,0x06]
; X86-NEXT: kshiftrq $32, %k0, %k1 # encoding: [0xc4,0xe3,0xf9,0x31,0xc8,0x20]
-; X86-NEXT: kmovd %k1, %edx # encoding: [0xc5,0xfb,0x93,0xd1]
; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
-; X86-NEXT: addl %esi, %eax # encoding: [0x01,0xf0]
-; X86-NEXT: adcl %ecx, %edx # encoding: [0x11,0xca]
+; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
+; X86-NEXT: kmovd %k1, %edx # encoding: [0xc5,0xfb,0x93,0xd1]
+; X86-NEXT: adcl %esi, %edx # encoding: [0x11,0xf2]
; X86-NEXT: addl {{[0-9]+}}(%esp), %eax # encoding: [0x03,0x44,0x24,0x08]
; X86-NEXT: adcl {{[0-9]+}}(%esp), %edx # encoding: [0x13,0x54,0x24,0x0c]
; X86-NEXT: popl %esi # encoding: [0x5e]
diff --git a/llvm/test/CodeGen/X86/avx512vl-vec-masked-cmp.ll b/llvm/test/CodeGen/X86/avx512vl-vec-masked-cmp.ll
index bed8d5f..42ea007 100644
--- a/llvm/test/CodeGen/X86/avx512vl-vec-masked-cmp.ll
+++ b/llvm/test/CodeGen/X86/avx512vl-vec-masked-cmp.ll
@@ -2698,8 +2698,9 @@
; NoVLX-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vpcmpeqq %zmm1, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -2723,8 +2724,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vmovdqa (%rdi), %xmm1
; NoVLX-NEXT: vpcmpeqq %zmm1, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -2751,8 +2753,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpeqq %zmm1, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -2781,8 +2784,9 @@
; NoVLX-NEXT: vmovdqa (%rsi), %xmm1
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpeqq %zmm1, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -2810,8 +2814,9 @@
; NoVLX: # %bb.0: # %entry
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vpcmpeqq (%rdi){1to8}, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -2838,8 +2843,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpeqq (%rsi){1to8}, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -7496,8 +7502,9 @@
; NoVLX-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vpcmpgtq %zmm1, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -7521,8 +7528,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vmovdqa (%rdi), %xmm1
; NoVLX-NEXT: vpcmpgtq %zmm1, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -7549,8 +7557,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpgtq %zmm1, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -7579,8 +7588,9 @@
; NoVLX-NEXT: vmovdqa (%rsi), %xmm1
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpgtq %zmm1, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -7608,8 +7618,9 @@
; NoVLX: # %bb.0: # %entry
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vpcmpgtq (%rdi){1to8}, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -7636,8 +7647,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpgtq (%rsi){1to8}, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -12354,8 +12366,9 @@
; NoVLX-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vpcmpnltq %zmm1, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -12379,8 +12392,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vmovdqa (%rdi), %xmm1
; NoVLX-NEXT: vpcmpnltq %zmm1, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -12407,8 +12421,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpnltq %zmm1, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -12437,8 +12452,9 @@
; NoVLX-NEXT: vmovdqa (%rsi), %xmm1
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpnltq %zmm1, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -12466,8 +12482,9 @@
; NoVLX: # %bb.0: # %entry
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vpcmpnltq (%rdi){1to8}, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -12494,8 +12511,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpnltq (%rsi){1to8}, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -17232,8 +17250,9 @@
; NoVLX-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vpcmpltuq %zmm1, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -17257,8 +17276,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vmovdqa (%rdi), %xmm1
; NoVLX-NEXT: vpcmpltuq %zmm1, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -17285,8 +17305,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpltuq %zmm1, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -17315,8 +17336,9 @@
; NoVLX-NEXT: vmovdqa (%rsi), %xmm1
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpltuq %zmm1, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -17344,8 +17366,9 @@
; NoVLX: # %bb.0: # %entry
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vpcmpltuq (%rdi){1to8}, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -17372,8 +17395,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vpcmpltuq (%rsi){1to8}, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -21057,8 +21081,9 @@
; NoVLX-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vcmpeqpd %zmm1, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -21082,8 +21107,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vmovapd (%rdi), %xmm1
; NoVLX-NEXT: vcmpeqpd %zmm1, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -21107,8 +21133,9 @@
; NoVLX: # %bb.0: # %entry
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: vcmpeqpd (%rdi){1to8}, %zmm0, %k0
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -21136,8 +21163,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vcmpeqpd %zmm1, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -21165,8 +21193,9 @@
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vmovapd (%rsi), %xmm1
; NoVLX-NEXT: vcmpeqpd %zmm1, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
@@ -21194,8 +21223,9 @@
; NoVLX-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; NoVLX-NEXT: kmovw %edi, %k1
; NoVLX-NEXT: vcmpeqpd (%rsi){1to8}, %zmm0, %k0 {%k1}
+; NoVLX-NEXT: kshiftlw $14, %k0, %k0
+; NoVLX-NEXT: kshiftrw $14, %k0, %k0
; NoVLX-NEXT: kmovw %k0, %eax
-; NoVLX-NEXT: andl $3, %eax
; NoVLX-NEXT: vzeroupper
; NoVLX-NEXT: retq
entry:
diff --git a/llvm/test/CodeGen/X86/bitcast-vector-bool.ll b/llvm/test/CodeGen/X86/bitcast-vector-bool.ll
index a6aa52d..8687dae 100644
--- a/llvm/test/CodeGen/X86/bitcast-vector-bool.ll
+++ b/llvm/test/CodeGen/X86/bitcast-vector-bool.ll
@@ -121,15 +121,22 @@
; SSE41-NEXT: setb %al
; SSE41-NEXT: retq
;
-; AVX12-LABEL: trunc_v4i32_cmp:
-; AVX12: # %bb.0:
-; AVX12-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX12-NEXT: setb %al
-; AVX12-NEXT: retq
+; AVX1-LABEL: trunc_v4i32_cmp:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX1-NEXT: setb %al
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: trunc_v4i32_cmp:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; AVX2-NEXT: vptest %xmm1, %xmm0
+; AVX2-NEXT: setb %al
+; AVX2-NEXT: retq
;
; AVX512-LABEL: trunc_v4i32_cmp:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} xmm1 = [4294967297,4294967297]
+; AVX512-NEXT: vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
; AVX512-NEXT: vptest %xmm1, %xmm0
; AVX512-NEXT: setb %al
; AVX512-NEXT: retq
@@ -195,18 +202,11 @@
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
-; AVX12-LABEL: trunc_v8i16_cmp:
-; AVX12: # %bb.0:
-; AVX12-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX12-NEXT: setne %al
-; AVX12-NEXT: retq
-;
-; AVX512-LABEL: trunc_v8i16_cmp:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} xmm1 = [281479271743489,281479271743489]
-; AVX512-NEXT: vptest %xmm1, %xmm0
-; AVX512-NEXT: setne %al
-; AVX512-NEXT: retq
+; AVX-LABEL: trunc_v8i16_cmp:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX-NEXT: setne %al
+; AVX-NEXT: retq
%1 = trunc <8 x i16> %a0 to <8 x i1>
%2 = bitcast <8 x i1> %1 to i8
%3 = icmp ne i8 %2, 0
@@ -273,18 +273,11 @@
; SSE41-NEXT: setae %al
; SSE41-NEXT: retq
;
-; AVX12-LABEL: trunc_v16i8_cmp:
-; AVX12: # %bb.0:
-; AVX12-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX12-NEXT: setae %al
-; AVX12-NEXT: retq
-;
-; AVX512-LABEL: trunc_v16i8_cmp:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} xmm1 = [72340172838076673,72340172838076673]
-; AVX512-NEXT: vptest %xmm1, %xmm0
-; AVX512-NEXT: setae %al
-; AVX512-NEXT: retq
+; AVX-LABEL: trunc_v16i8_cmp:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX-NEXT: setae %al
+; AVX-NEXT: retq
%1 = trunc <16 x i8> %a0 to <16 x i1>
%2 = bitcast <16 x i1> %1 to i16
%3 = icmp ne i16 %2, -1
@@ -427,7 +420,7 @@
;
; AVX2-LABEL: trunc_v8i132_cmp:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setae %al
; AVX2-NEXT: vzeroupper
@@ -435,7 +428,7 @@
;
; AVX512-LABEL: trunc_v8i132_cmp:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX512-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX512-NEXT: vptest %ymm1, %ymm0
; AVX512-NEXT: setae %al
; AVX512-NEXT: vzeroupper
@@ -526,28 +519,12 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1-LABEL: trunc_v16i16_cmp:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
-; AVX1-NEXT: sete %al
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: trunc_v16i16_cmp:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX2-NEXT: vptest %ymm1, %ymm0
-; AVX2-NEXT: sete %al
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
-;
-; AVX512-LABEL: trunc_v16i16_cmp:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX512-NEXT: vptest %ymm1, %ymm0
-; AVX512-NEXT: sete %al
-; AVX512-NEXT: vzeroupper
-; AVX512-NEXT: retq
+; AVX-LABEL: trunc_v16i16_cmp:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
+; AVX-NEXT: sete %al
+; AVX-NEXT: vzeroupper
+; AVX-NEXT: retq
%1 = trunc <16 x i16> %a0 to <16 x i1>
%2 = bitcast <16 x i1> %1 to i16
%3 = icmp eq i16 %2, 0
@@ -618,28 +595,12 @@
; SSE41-NEXT: setb %al
; SSE41-NEXT: retq
;
-; AVX1-LABEL: trunc_v32i8_cmp:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
-; AVX1-NEXT: setb %al
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: trunc_v32i8_cmp:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX2-NEXT: vptest %ymm1, %ymm0
-; AVX2-NEXT: setb %al
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
-;
-; AVX512-LABEL: trunc_v32i8_cmp:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX512-NEXT: vptest %ymm1, %ymm0
-; AVX512-NEXT: setb %al
-; AVX512-NEXT: vzeroupper
-; AVX512-NEXT: retq
+; AVX-LABEL: trunc_v32i8_cmp:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
+; AVX-NEXT: setb %al
+; AVX-NEXT: vzeroupper
+; AVX-NEXT: retq
%1 = trunc <32 x i8> %a0 to <32 x i1>
%2 = bitcast <32 x i1> %1 to i32
%3 = icmp eq i32 %2, -1
@@ -887,7 +848,7 @@
; AVX2-LABEL: trunc_v16i32_cmp:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: sete %al
; AVX2-NEXT: vzeroupper
@@ -993,8 +954,7 @@
; AVX2-LABEL: trunc_v32i16_cmp:
; AVX2: # %bb.0:
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX2-NEXT: vptest %ymm1, %ymm0
+; AVX2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX2-NEXT: setae %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
@@ -1119,8 +1079,7 @@
; AVX2-LABEL: trunc_v64i8_cmp:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX2-NEXT: vptest %ymm1, %ymm0
+; AVX2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/const-shift-of-constmasked.ll b/llvm/test/CodeGen/X86/const-shift-of-constmasked.ll
index 1fd1387..10787dc 100644
--- a/llvm/test/CodeGen/X86/const-shift-of-constmasked.ll
+++ b/llvm/test/CodeGen/X86/const-shift-of-constmasked.ll
@@ -574,7 +574,7 @@
define i16 @test_i16_2032_mask_lshr_4(i16 %a0) {
; X86-LABEL: test_i16_2032_mask_lshr_4:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: shrl $4, %eax
; X86-NEXT: andl $127, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
@@ -594,7 +594,7 @@
define i16 @test_i16_2032_mask_lshr_5(i16 %a0) {
; X86-LABEL: test_i16_2032_mask_lshr_5:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: shrl $5, %eax
; X86-NEXT: andl $63, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
@@ -614,7 +614,7 @@
define i16 @test_i16_2032_mask_lshr_6(i16 %a0) {
; X86-LABEL: test_i16_2032_mask_lshr_6:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: shrl $6, %eax
; X86-NEXT: andl $31, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
@@ -755,7 +755,7 @@
define i16 @test_i16_2032_mask_ashr_4(i16 %a0) {
; X86-LABEL: test_i16_2032_mask_ashr_4:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: shrl $4, %eax
; X86-NEXT: andl $127, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
@@ -775,7 +775,7 @@
define i16 @test_i16_2032_mask_ashr_5(i16 %a0) {
; X86-LABEL: test_i16_2032_mask_ashr_5:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: shrl $5, %eax
; X86-NEXT: andl $63, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
@@ -795,7 +795,7 @@
define i16 @test_i16_2032_mask_ashr_6(i16 %a0) {
; X86-LABEL: test_i16_2032_mask_ashr_6:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: shrl $6, %eax
; X86-NEXT: andl $31, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
diff --git a/llvm/test/CodeGen/X86/dagcombine-select.ll b/llvm/test/CodeGen/X86/dagcombine-select.ll
index ff022c3..9bebec6 100644
--- a/llvm/test/CodeGen/X86/dagcombine-select.ll
+++ b/llvm/test/CodeGen/X86/dagcombine-select.ll
@@ -194,12 +194,10 @@
define i32 @shl_constant_sel_constants(i1 %cond) {
; CHECK-LABEL: shl_constant_sel_constants:
; CHECK: # %bb.0:
-; CHECK-NEXT: movl %edi, %ecx
-; CHECK-NEXT: andb $1, %cl
-; CHECK-NEXT: xorb $3, %cl
-; CHECK-NEXT: movl $1, %eax
-; CHECK-NEXT: # kill: def $cl killed $cl killed $ecx
-; CHECK-NEXT: shll %cl, %eax
+; CHECK-NEXT: notb %dil
+; CHECK-NEXT: movzbl %dil, %eax
+; CHECK-NEXT: andl $1, %eax
+; CHECK-NEXT: leal 4(,%rax,4), %eax
; CHECK-NEXT: retq
%sel = select i1 %cond, i32 2, i32 3
%bo = shl i32 1, %sel
@@ -209,12 +207,9 @@
define i32 @lshr_constant_sel_constants(i1 %cond) {
; CHECK-LABEL: lshr_constant_sel_constants:
; CHECK: # %bb.0:
-; CHECK-NEXT: movl %edi, %ecx
-; CHECK-NEXT: andb $1, %cl
-; CHECK-NEXT: xorb $3, %cl
-; CHECK-NEXT: movl $64, %eax
-; CHECK-NEXT: # kill: def $cl killed $cl killed $ecx
-; CHECK-NEXT: shrl %cl, %eax
+; CHECK-NEXT: # kill: def $edi killed $edi def $rdi
+; CHECK-NEXT: andl $1, %edi
+; CHECK-NEXT: leal 8(,%rdi,8), %eax
; CHECK-NEXT: retq
%sel = select i1 %cond, i32 2, i32 3
%bo = lshr i32 64, %sel
@@ -224,12 +219,10 @@
define i32 @ashr_constant_sel_constants(i1 %cond) {
; CHECK-LABEL: ashr_constant_sel_constants:
; CHECK: # %bb.0:
-; CHECK-NEXT: movl %edi, %ecx
-; CHECK-NEXT: andb $1, %cl
-; CHECK-NEXT: xorb $3, %cl
-; CHECK-NEXT: movl $128, %eax
-; CHECK-NEXT: # kill: def $cl killed $cl killed $ecx
-; CHECK-NEXT: shrl %cl, %eax
+; CHECK-NEXT: # kill: def $edi killed $edi def $rdi
+; CHECK-NEXT: andl $1, %edi
+; CHECK-NEXT: shll $4, %edi
+; CHECK-NEXT: leal 16(%rdi), %eax
; CHECK-NEXT: retq
%sel = select i1 %cond, i32 2, i32 3
%bo = ashr i32 128, %sel
diff --git a/llvm/test/CodeGen/X86/field-extract-use-trunc.ll b/llvm/test/CodeGen/X86/field-extract-use-trunc.ll
index b9721d2..1a169a0 100644
--- a/llvm/test/CodeGen/X86/field-extract-use-trunc.ll
+++ b/llvm/test/CodeGen/X86/field-extract-use-trunc.ll
@@ -73,7 +73,7 @@
define i16 @test5(i16 %f12) nounwind {
; i686-LABEL: test5:
; i686: # %bb.0:
-; i686-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; i686-NEXT: movl {{[0-9]+}}(%esp), %eax
; i686-NEXT: shrl $6, %eax
; i686-NEXT: movsbl %al, %eax
; i686-NEXT: # kill: def $ax killed $ax killed $eax
diff --git a/llvm/test/CodeGen/X86/horizontal-sum.ll b/llvm/test/CodeGen/X86/horizontal-sum.ll
index e1d2f9c..ab9a1e9 100644
--- a/llvm/test/CodeGen/X86/horizontal-sum.ll
+++ b/llvm/test/CodeGen/X86/horizontal-sum.ll
@@ -32,17 +32,17 @@
; SSSE3-FAST-NEXT: haddps %xmm2, %xmm0
; SSSE3-FAST-NEXT: retq
;
-; AVX-SLOW-LABEL: pair_sum_v4f32_v4f32:
-; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vhaddps %xmm1, %xmm0, %xmm0
-; AVX-SLOW-NEXT: vhaddps %xmm2, %xmm2, %xmm1
-; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm0[0,2],xmm1[0,1]
-; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,1]
-; AVX-SLOW-NEXT: vhaddps %xmm3, %xmm3, %xmm1
-; AVX-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
-; AVX-SLOW-NEXT: vinsertps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[0]
-; AVX-SLOW-NEXT: vaddps %xmm0, %xmm1, %xmm0
-; AVX-SLOW-NEXT: retq
+; AVX1-SLOW-LABEL: pair_sum_v4f32_v4f32:
+; AVX1-SLOW: # %bb.0:
+; AVX1-SLOW-NEXT: vhaddps %xmm1, %xmm0, %xmm0
+; AVX1-SLOW-NEXT: vhaddps %xmm2, %xmm2, %xmm1
+; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm0[0,2],xmm1[0,1]
+; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,1]
+; AVX1-SLOW-NEXT: vhaddps %xmm3, %xmm3, %xmm1
+; AVX1-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
+; AVX1-SLOW-NEXT: vinsertps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[0]
+; AVX1-SLOW-NEXT: vaddps %xmm0, %xmm1, %xmm0
+; AVX1-SLOW-NEXT: retq
;
; AVX-FAST-LABEL: pair_sum_v4f32_v4f32:
; AVX-FAST: # %bb.0:
@@ -50,6 +50,18 @@
; AVX-FAST-NEXT: vhaddps %xmm3, %xmm2, %xmm1
; AVX-FAST-NEXT: vhaddps %xmm1, %xmm0, %xmm0
; AVX-FAST-NEXT: retq
+;
+; AVX2-SLOW-LABEL: pair_sum_v4f32_v4f32:
+; AVX2-SLOW: # %bb.0:
+; AVX2-SLOW-NEXT: vhaddps %xmm1, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vhaddps %xmm2, %xmm2, %xmm1
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm0[0,2],xmm1[0,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
+; AVX2-SLOW-NEXT: vhaddps %xmm3, %xmm3, %xmm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
+; AVX2-SLOW-NEXT: vinsertps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[0]
+; AVX2-SLOW-NEXT: vaddps %xmm0, %xmm1, %xmm0
+; AVX2-SLOW-NEXT: retq
%5 = shufflevector <4 x float> %0, <4 x float> poison, <2 x i32> <i32 0, i32 2>
%6 = shufflevector <4 x float> %0, <4 x float> poison, <2 x i32> <i32 1, i32 3>
%7 = fadd <2 x float> %5, %6
@@ -114,28 +126,34 @@
; AVX1-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
; AVX1-SLOW-NEXT: retq
;
-; AVX-FAST-LABEL: pair_sum_v4i32_v4i32:
-; AVX-FAST: # %bb.0:
-; AVX-FAST-NEXT: vphaddd %xmm1, %xmm0, %xmm0
-; AVX-FAST-NEXT: vphaddd %xmm3, %xmm2, %xmm1
-; AVX-FAST-NEXT: vphaddd %xmm1, %xmm0, %xmm0
-; AVX-FAST-NEXT: retq
+; AVX1-FAST-LABEL: pair_sum_v4i32_v4i32:
+; AVX1-FAST: # %bb.0:
+; AVX1-FAST-NEXT: vphaddd %xmm1, %xmm0, %xmm0
+; AVX1-FAST-NEXT: vphaddd %xmm3, %xmm2, %xmm1
+; AVX1-FAST-NEXT: vphaddd %xmm1, %xmm0, %xmm0
+; AVX1-FAST-NEXT: retq
;
; AVX2-SLOW-LABEL: pair_sum_v4i32_v4i32:
; AVX2-SLOW: # %bb.0:
; AVX2-SLOW-NEXT: vphaddd %xmm1, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,3,1,3]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vphaddd %xmm2, %xmm2, %xmm1
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
; AVX2-SLOW-NEXT: vpaddd %xmm2, %xmm1, %xmm1
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,3]
; AVX2-SLOW-NEXT: vphaddd %xmm3, %xmm3, %xmm1
-; AVX2-SLOW-NEXT: vpbroadcastd %xmm1, %xmm2
-; AVX2-SLOW-NEXT: vpaddd %xmm1, %xmm2, %xmm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
+; AVX2-SLOW-NEXT: vpaddd %xmm2, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[0]
; AVX2-SLOW-NEXT: retq
+;
+; AVX2-FAST-LABEL: pair_sum_v4i32_v4i32:
+; AVX2-FAST: # %bb.0:
+; AVX2-FAST-NEXT: vphaddd %xmm3, %xmm2, %xmm2
+; AVX2-FAST-NEXT: vphaddd %xmm1, %xmm0, %xmm0
+; AVX2-FAST-NEXT: vphaddd %xmm2, %xmm0, %xmm0
+; AVX2-FAST-NEXT: retq
%5 = shufflevector <4 x i32> %0, <4 x i32> poison, <2 x i32> <i32 0, i32 2>
%6 = shufflevector <4 x i32> %0, <4 x i32> poison, <2 x i32> <i32 1, i32 3>
%7 = add <2 x i32> %5, %6
@@ -173,14 +191,15 @@
; SSSE3-SLOW-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
; SSSE3-SLOW-NEXT: addps %xmm1, %xmm0
; SSSE3-SLOW-NEXT: haddps %xmm3, %xmm2
-; SSSE3-SLOW-NEXT: haddps %xmm4, %xmm5
-; SSSE3-SLOW-NEXT: haddps %xmm5, %xmm2
+; SSSE3-SLOW-NEXT: movaps %xmm5, %xmm1
+; SSSE3-SLOW-NEXT: haddps %xmm4, %xmm1
+; SSSE3-SLOW-NEXT: haddps %xmm1, %xmm2
; SSSE3-SLOW-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1,3,2]
; SSSE3-SLOW-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; SSSE3-SLOW-NEXT: haddps %xmm7, %xmm6
-; SSSE3-SLOW-NEXT: haddps %xmm6, %xmm6
-; SSSE3-SLOW-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,3],xmm6[0,1]
-; SSSE3-SLOW-NEXT: movaps %xmm2, %xmm1
+; SSSE3-SLOW-NEXT: haddps %xmm5, %xmm4
+; SSSE3-SLOW-NEXT: haddps %xmm6, %xmm4
+; SSSE3-SLOW-NEXT: movaps %xmm4, %xmm1
; SSSE3-SLOW-NEXT: retq
;
; SSSE3-FAST-LABEL: pair_sum_v8f32_v4f32:
@@ -247,13 +266,13 @@
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
; AVX2-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vhaddps %xmm4, %xmm4, %xmm1
-; AVX2-SLOW-NEXT: vhaddps %xmm5, %xmm5, %xmm8
+; AVX2-SLOW-NEXT: vhaddps %xmm5, %xmm5, %xmm4
; AVX2-SLOW-NEXT: vhaddps %xmm3, %xmm2, %xmm2
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm2[0,2],xmm1[0,1]
-; AVX2-SLOW-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0,1,2],xmm8[0]
-; AVX2-SLOW-NEXT: vhaddps %xmm4, %xmm5, %xmm3
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,3],xmm3[3,1]
-; AVX2-SLOW-NEXT: vaddps %xmm2, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm2[0,2],xmm1[0,3]
+; AVX2-SLOW-NEXT: vinsertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[0]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
+; AVX2-SLOW-NEXT: vaddps %xmm1, %xmm3, %xmm1
; AVX2-SLOW-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX2-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -268,13 +287,13 @@
; AVX2-FAST-NEXT: vhaddps %xmm1, %xmm0, %xmm0
; AVX2-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX2-FAST-NEXT: vhaddps %xmm4, %xmm4, %xmm1
-; AVX2-FAST-NEXT: vhaddps %xmm5, %xmm5, %xmm8
+; AVX2-FAST-NEXT: vhaddps %xmm5, %xmm5, %xmm4
; AVX2-FAST-NEXT: vhaddps %xmm3, %xmm2, %xmm2
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm2[0,2],xmm1[0,1]
-; AVX2-FAST-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0,1,2],xmm8[0]
-; AVX2-FAST-NEXT: vhaddps %xmm4, %xmm5, %xmm3
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,3],xmm3[3,1]
-; AVX2-FAST-NEXT: vaddps %xmm2, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm2[0,2],xmm1[0,3]
+; AVX2-FAST-NEXT: vinsertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[0]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
+; AVX2-FAST-NEXT: vaddps %xmm1, %xmm3, %xmm1
; AVX2-FAST-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX2-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX2-FAST-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -421,11 +440,9 @@
; AVX2-SLOW-NEXT: vphaddd %xmm4, %xmm4, %xmm1
; AVX2-SLOW-NEXT: vphaddd %xmm5, %xmm5, %xmm4
; AVX2-SLOW-NEXT: vphaddd %xmm3, %xmm2, %xmm2
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm1[0]
-; AVX2-SLOW-NEXT: vpbroadcastd %xmm4, %xmm5
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],xmm5[3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,1]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm2[0,2],xmm1[0,3]
+; AVX2-SLOW-NEXT: vinsertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[0]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
; AVX2-SLOW-NEXT: vpaddd %xmm1, %xmm3, %xmm1
; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
@@ -444,11 +461,9 @@
; AVX2-FAST-NEXT: vphaddd %xmm4, %xmm4, %xmm1
; AVX2-FAST-NEXT: vphaddd %xmm5, %xmm5, %xmm4
; AVX2-FAST-NEXT: vphaddd %xmm3, %xmm2, %xmm2
-; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
-; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm1[0]
-; AVX2-FAST-NEXT: vpbroadcastd %xmm4, %xmm5
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],xmm5[3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,1]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm2[0,2],xmm1[0,3]
+; AVX2-FAST-NEXT: vinsertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[0]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
; AVX2-FAST-NEXT: vpaddd %xmm1, %xmm3, %xmm1
; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
@@ -737,15 +752,15 @@
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[3,3,3,3]
; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm3[1,1,1,1]
+; AVX2-SLOW-NEXT: vpbroadcastd %xmm3, %xmm5
+; AVX2-SLOW-NEXT: vpaddd %xmm5, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm3[2,2,2,2]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm4[3]
; AVX2-SLOW-NEXT: vpaddd %xmm1, %xmm2, %xmm1
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
; AVX2-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vpbroadcastq %xmm3, %xmm1
-; AVX2-SLOW-NEXT: vpbroadcastd %xmm3, %xmm2
-; AVX2-SLOW-NEXT: vpaddd %xmm2, %xmm1, %xmm1
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[2,2,2,2]
-; AVX2-SLOW-NEXT: vpaddd %xmm2, %xmm3, %xmm2
-; AVX2-SLOW-NEXT: vpaddd %xmm1, %xmm2, %xmm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: sequential_sum_v4i32_v4i32:
@@ -761,14 +776,14 @@
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[3,3,3,3]
; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
+; AVX2-FAST-NEXT: vphaddd %xmm3, %xmm3, %xmm4
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm5 = xmm3[2,2,2,2]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm5[3]
+; AVX2-FAST-NEXT: vpbroadcastd %xmm4, %xmm4
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
; AVX2-FAST-NEXT: vpaddd %xmm1, %xmm2, %xmm1
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
; AVX2-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vphaddd %xmm3, %xmm3, %xmm1
-; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[2,2,2,2]
-; AVX2-FAST-NEXT: vpbroadcastd %xmm1, %xmm1
-; AVX2-FAST-NEXT: vpaddd %xmm2, %xmm3, %xmm2
-; AVX2-FAST-NEXT: vpaddd %xmm1, %xmm2, %xmm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
; AVX2-FAST-NEXT: retq
%5 = shufflevector <4 x i32> %0, <4 x i32> %1, <2 x i32> <i32 0, i32 4>
%6 = shufflevector <4 x i32> %0, <4 x i32> %1, <2 x i32> <i32 1, i32 5>
diff --git a/llvm/test/CodeGen/X86/icmp-shift-opt.ll b/llvm/test/CodeGen/X86/icmp-shift-opt.ll
index 1673649..7482de0 100644
--- a/llvm/test/CodeGen/X86/icmp-shift-opt.ll
+++ b/llvm/test/CodeGen/X86/icmp-shift-opt.ll
@@ -223,11 +223,11 @@
; X86-LABEL: opt_setcc_expanded_shl_correct_shifts:
; X86: # %bb.0:
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: orl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: orl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: orl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: shll $17, %ecx
-; X86-NEXT: orl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: orl %ecx, %eax
+; X86-NEXT: orl %eax, %ecx
+; X86-NEXT: shldl $17, %eax, %ecx
; X86-NEXT: sete %al
; X86-NEXT: retl
;
diff --git a/llvm/test/CodeGen/X86/insertelement-var-index.ll b/llvm/test/CodeGen/X86/insertelement-var-index.ll
index 8ed8495..1d90791 100644
--- a/llvm/test/CodeGen/X86/insertelement-var-index.ll
+++ b/llvm/test/CodeGen/X86/insertelement-var-index.ll
@@ -2288,58 +2288,31 @@
; SSE-NEXT: divl %ecx
; SSE-NEXT: retq
;
-; AVX1-LABEL: PR44139:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vbroadcastsd (%rdi), %ymm0
-; AVX1-NEXT: vpinsrq $1, (%rdi), %xmm0, %xmm1
-; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-NEXT: vmovaps %ymm0, 64(%rdi)
-; AVX1-NEXT: vmovaps %ymm0, 96(%rdi)
-; AVX1-NEXT: vmovaps %ymm0, 32(%rdi)
-; AVX1-NEXT: movl (%rdi), %eax
-; AVX1-NEXT: vmovaps %ymm1, (%rdi)
-; AVX1-NEXT: leal 2147483647(%rax), %ecx
-; AVX1-NEXT: testl %eax, %eax
-; AVX1-NEXT: cmovnsl %eax, %ecx
-; AVX1-NEXT: andl $-2147483648, %ecx # imm = 0x80000000
-; AVX1-NEXT: addl %eax, %ecx
-; AVX1-NEXT: # kill: def $eax killed $eax killed $rax
-; AVX1-NEXT: xorl %edx, %edx
-; AVX1-NEXT: divl %ecx
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: PR44139:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq (%rdi), %ymm0
-; AVX2-NEXT: vpinsrq $1, (%rdi), %xmm0, %xmm1
-; AVX2-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-NEXT: vmovdqa %ymm0, 64(%rdi)
-; AVX2-NEXT: vmovdqa %ymm0, 96(%rdi)
-; AVX2-NEXT: vmovdqa %ymm0, 32(%rdi)
-; AVX2-NEXT: movl (%rdi), %eax
-; AVX2-NEXT: vmovdqa %ymm1, (%rdi)
-; AVX2-NEXT: leal 2147483647(%rax), %ecx
-; AVX2-NEXT: testl %eax, %eax
-; AVX2-NEXT: cmovnsl %eax, %ecx
-; AVX2-NEXT: andl $-2147483648, %ecx # imm = 0x80000000
-; AVX2-NEXT: addl %eax, %ecx
-; AVX2-NEXT: # kill: def $eax killed $eax killed $rax
-; AVX2-NEXT: xorl %edx, %edx
-; AVX2-NEXT: divl %ecx
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
+; AVX1OR2-LABEL: PR44139:
+; AVX1OR2: # %bb.0:
+; AVX1OR2-NEXT: vbroadcastsd (%rdi), %ymm0
+; AVX1OR2-NEXT: movl (%rdi), %eax
+; AVX1OR2-NEXT: vmovaps %ymm0, 64(%rdi)
+; AVX1OR2-NEXT: vmovaps %ymm0, 96(%rdi)
+; AVX1OR2-NEXT: vmovaps %ymm0, (%rdi)
+; AVX1OR2-NEXT: vmovaps %ymm0, 32(%rdi)
+; AVX1OR2-NEXT: leal 2147483647(%rax), %ecx
+; AVX1OR2-NEXT: testl %eax, %eax
+; AVX1OR2-NEXT: cmovnsl %eax, %ecx
+; AVX1OR2-NEXT: andl $-2147483648, %ecx # imm = 0x80000000
+; AVX1OR2-NEXT: addl %eax, %ecx
+; AVX1OR2-NEXT: # kill: def $eax killed $eax killed $rax
+; AVX1OR2-NEXT: xorl %edx, %edx
+; AVX1OR2-NEXT: divl %ecx
+; AVX1OR2-NEXT: vzeroupper
+; AVX1OR2-NEXT: retq
;
; AVX512-LABEL: PR44139:
; AVX512: # %bb.0:
-; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512-NEXT: vpbroadcastq (%rdi), %zmm1
-; AVX512-NEXT: vpmovqd %zmm0, %ymm0
-; AVX512-NEXT: vpinsrq $1, (%rdi), %xmm1, %xmm2
-; AVX512-NEXT: vinserti32x4 $0, %xmm2, %zmm1, %zmm2
-; AVX512-NEXT: vmovdqa64 %zmm1, 64(%rdi)
-; AVX512-NEXT: vmovdqa64 %zmm2, (%rdi)
-; AVX512-NEXT: vmovd %xmm0, %eax
+; AVX512-NEXT: vbroadcastsd (%rdi), %zmm0
+; AVX512-NEXT: movl (%rdi), %eax
+; AVX512-NEXT: vmovaps %zmm0, (%rdi)
+; AVX512-NEXT: vmovaps %zmm0, 64(%rdi)
; AVX512-NEXT: leal 2147483647(%rax), %ecx
; AVX512-NEXT: testl %eax, %eax
; AVX512-NEXT: cmovnsl %eax, %ecx
@@ -2354,14 +2327,12 @@
; X86AVX2-LABEL: PR44139:
; X86AVX2: # %bb.0:
; X86AVX2-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86AVX2-NEXT: movl (%ecx), %eax
; X86AVX2-NEXT: vbroadcastsd (%ecx), %ymm0
-; X86AVX2-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; X86AVX2-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; X86AVX2-NEXT: vmovaps %ymm0, 64(%ecx)
; X86AVX2-NEXT: vmovaps %ymm0, 96(%ecx)
+; X86AVX2-NEXT: vmovaps %ymm0, (%ecx)
; X86AVX2-NEXT: vmovaps %ymm0, 32(%ecx)
-; X86AVX2-NEXT: movl (%ecx), %eax
-; X86AVX2-NEXT: vmovaps %ymm1, (%ecx)
; X86AVX2-NEXT: leal 2147483647(%eax), %ecx
; X86AVX2-NEXT: testl %eax, %eax
; X86AVX2-NEXT: cmovnsl %eax, %ecx
diff --git a/llvm/test/CodeGen/X86/is_fpclass-fp80.ll b/llvm/test/CodeGen/X86/is_fpclass-fp80.ll
index 4464d89..4ebcce5 100644
--- a/llvm/test/CodeGen/X86/is_fpclass-fp80.ll
+++ b/llvm/test/CodeGen/X86/is_fpclass-fp80.ll
@@ -319,10 +319,10 @@
; CHECK-64-LABEL: is_neginf_f80:
; CHECK-64: # %bb.0: # %entry
; CHECK-64-NEXT: movzwl {{[0-9]+}}(%rsp), %eax
-; CHECK-64-NEXT: xorq $65535, %rax # imm = 0xFFFF
+; CHECK-64-NEXT: xorl $65535, %eax # imm = 0xFFFF
; CHECK-64-NEXT: movabsq $-9223372036854775808, %rcx # imm = 0x8000000000000000
; CHECK-64-NEXT: xorq {{[0-9]+}}(%rsp), %rcx
-; CHECK-64-NEXT: orq %rax, %rcx
+; CHECK-64-NEXT: orq %rcx, %rax
; CHECK-64-NEXT: sete %al
; CHECK-64-NEXT: retq
entry:
diff --git a/llvm/test/CodeGen/X86/isel-blendi-gettargetconstant.ll b/llvm/test/CodeGen/X86/isel-blendi-gettargetconstant.ll
index fb7efc2..7777a57 100644
--- a/llvm/test/CodeGen/X86/isel-blendi-gettargetconstant.ll
+++ b/llvm/test/CodeGen/X86/isel-blendi-gettargetconstant.ll
@@ -5,9 +5,10 @@
; CHECK-LABEL: csrot_:
; CHECK: # %bb.0:
; CHECK-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; CHECK-NEXT: xorps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; CHECK-NEXT: blendps {{.*#+}} xmm0 = xmm0[0],mem[1,2,3]
-; CHECK-NEXT: movlps %xmm0, (%rax)
+; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; CHECK-NEXT: xorps %xmm0, %xmm1
+; CHECK-NEXT: blendps {{.*#+}} xmm1 = xmm1[0],mem[1,2,3]
+; CHECK-NEXT: movlps %xmm1, (%rax)
; CHECK-NEXT: retq
1:
%2 = load float, ptr %0, align 4
diff --git a/llvm/test/CodeGen/X86/masked_store.ll b/llvm/test/CodeGen/X86/masked_store.ll
index ac538e0..d831a1b 100644
--- a/llvm/test/CodeGen/X86/masked_store.ll
+++ b/llvm/test/CodeGen/X86/masked_store.ll
@@ -6185,20 +6185,19 @@
; AVX2-NEXT: vpcmpgtd 64(%rdi), %ymm3, %ymm3
; AVX2-NEXT: vextracti128 $1, %ymm3, %xmm5
; AVX2-NEXT: vpackssdw %xmm5, %xmm3, %xmm3
-; AVX2-NEXT: vpacksswb %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm5, %xmm5, %xmm5
-; AVX2-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero
-; AVX2-NEXT: vpslld $31, %ymm5, %ymm5
-; AVX2-NEXT: vpmaskmovd %ymm1, %ymm5, 32(%rdx)
+; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
+; AVX2-NEXT: vpslld $31, %ymm3, %ymm3
+; AVX2-NEXT: vpmaskmovd %ymm2, %ymm3, 64(%rdx)
+; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
+; AVX2-NEXT: vpslld $31, %ymm2, %ymm2
+; AVX2-NEXT: vpmaskmovd %ymm1, %ymm2, 32(%rdx)
; AVX2-NEXT: vpmovzxwd {{.*#+}} xmm1 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT: vpslld $31, %ymm1, %ymm1
; AVX2-NEXT: vpmaskmovd %ymm0, %ymm1, (%rdx)
-; AVX2-NEXT: vpmovzxwd {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
-; AVX2-NEXT: vpslld $31, %ymm0, %ymm0
-; AVX2-NEXT: vpmaskmovd %ymm2, %ymm0, 64(%rdx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -6278,158 +6277,77 @@
; From https://reviews.llvm.org/rGf8d9097168b7#1165311
define void @undefshuffle(<8 x i1> %i0, ptr %src, ptr %dst) #0 {
-; SSE2-LABEL: undefshuffle:
-; SSE2: ## %bb.0:
-; SSE2-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)
-; SSE2-NEXT: movb $1, %al
-; SSE2-NEXT: testb %al, %al
-; SSE2-NEXT: testb %al, %al
-; SSE2-NEXT: testb %al, %al
-; SSE2-NEXT: movzbl -{{[0-9]+}}(%rsp), %ecx
-; SSE2-NEXT: testb %al, %al
-; SSE2-NEXT: movd %ecx, %xmm0
-; SSE2-NEXT: movzbl -{{[0-9]+}}(%rsp), %ecx
-; SSE2-NEXT: testb %al, %al
-; SSE2-NEXT: pinsrw $1, %ecx, %xmm0
-; SSE2-NEXT: movl -{{[0-9]+}}(%rsp), %ecx
-; SSE2-NEXT: testb %al, %al
-; SSE2-NEXT: pinsrw $2, %ecx, %xmm0
-; SSE2-NEXT: movzbl -{{[0-9]+}}(%rsp), %eax
-; SSE2-NEXT: movb $1, %cl
-; SSE2-NEXT: testb %cl, %cl
-; SSE2-NEXT: pinsrw $3, %eax, %xmm0
-; SSE2-NEXT: testb %cl, %cl
-; SSE2-NEXT: psllw $15, %xmm0
-; SSE2-NEXT: packsswb %xmm0, %xmm0
-; SSE2-NEXT: pmovmskb %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: jne LBB32_1
-; SSE2-NEXT: ## %bb.2: ## %else23
-; SSE2-NEXT: testb $2, %al
-; SSE2-NEXT: jne LBB32_3
-; SSE2-NEXT: LBB32_4: ## %else25
-; SSE2-NEXT: testb $4, %al
-; SSE2-NEXT: jne LBB32_5
-; SSE2-NEXT: LBB32_6: ## %else27
-; SSE2-NEXT: testb $8, %al
-; SSE2-NEXT: jne LBB32_7
-; SSE2-NEXT: LBB32_8: ## %else29
-; SSE2-NEXT: testb $16, %al
-; SSE2-NEXT: jne LBB32_9
-; SSE2-NEXT: LBB32_10: ## %else31
-; SSE2-NEXT: testb $32, %al
-; SSE2-NEXT: jne LBB32_11
-; SSE2-NEXT: LBB32_12: ## %else33
-; SSE2-NEXT: testb $64, %al
-; SSE2-NEXT: jne LBB32_13
-; SSE2-NEXT: LBB32_14: ## %else35
-; SSE2-NEXT: testb $-128, %al
-; SSE2-NEXT: jne LBB32_15
-; SSE2-NEXT: LBB32_16: ## %else37
-; SSE2-NEXT: retq
-; SSE2-NEXT: LBB32_1: ## %cond.store
-; SSE2-NEXT: movl $0, (%rsi)
-; SSE2-NEXT: testb $2, %al
-; SSE2-NEXT: je LBB32_4
-; SSE2-NEXT: LBB32_3: ## %cond.store24
-; SSE2-NEXT: movl $0, 4(%rsi)
-; SSE2-NEXT: testb $4, %al
-; SSE2-NEXT: je LBB32_6
-; SSE2-NEXT: LBB32_5: ## %cond.store26
-; SSE2-NEXT: movl $0, 8(%rsi)
-; SSE2-NEXT: testb $8, %al
-; SSE2-NEXT: je LBB32_8
-; SSE2-NEXT: LBB32_7: ## %cond.store28
-; SSE2-NEXT: movl $0, 12(%rsi)
-; SSE2-NEXT: testb $16, %al
-; SSE2-NEXT: je LBB32_10
-; SSE2-NEXT: LBB32_9: ## %cond.store30
-; SSE2-NEXT: movl $0, 16(%rsi)
-; SSE2-NEXT: testb $32, %al
-; SSE2-NEXT: je LBB32_12
-; SSE2-NEXT: LBB32_11: ## %cond.store32
-; SSE2-NEXT: movl $0, 20(%rsi)
-; SSE2-NEXT: testb $64, %al
-; SSE2-NEXT: je LBB32_14
-; SSE2-NEXT: LBB32_13: ## %cond.store34
-; SSE2-NEXT: movl $0, 24(%rsi)
-; SSE2-NEXT: testb $-128, %al
-; SSE2-NEXT: je LBB32_16
-; SSE2-NEXT: LBB32_15: ## %cond.store36
-; SSE2-NEXT: movl $0, 28(%rsi)
-; SSE2-NEXT: retq
-;
-; SSE4-LABEL: undefshuffle:
-; SSE4: ## %bb.0:
-; SSE4-NEXT: movb $1, %al
-; SSE4-NEXT: testb %al, %al
-; SSE4-NEXT: testb %al, %al
-; SSE4-NEXT: testb %al, %al
-; SSE4-NEXT: testb %al, %al
-; SSE4-NEXT: testb %al, %al
-; SSE4-NEXT: testb %al, %al
-; SSE4-NEXT: testb %al, %al
-; SSE4-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; SSE4-NEXT: testb %al, %al
-; SSE4-NEXT: psllw $15, %xmm0
-; SSE4-NEXT: packsswb %xmm0, %xmm0
-; SSE4-NEXT: pmovmskb %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: jne LBB32_1
-; SSE4-NEXT: ## %bb.2: ## %else23
-; SSE4-NEXT: testb $2, %al
-; SSE4-NEXT: jne LBB32_3
-; SSE4-NEXT: LBB32_4: ## %else25
-; SSE4-NEXT: testb $4, %al
-; SSE4-NEXT: jne LBB32_5
-; SSE4-NEXT: LBB32_6: ## %else27
-; SSE4-NEXT: testb $8, %al
-; SSE4-NEXT: jne LBB32_7
-; SSE4-NEXT: LBB32_8: ## %else29
-; SSE4-NEXT: testb $16, %al
-; SSE4-NEXT: jne LBB32_9
-; SSE4-NEXT: LBB32_10: ## %else31
-; SSE4-NEXT: testb $32, %al
-; SSE4-NEXT: jne LBB32_11
-; SSE4-NEXT: LBB32_12: ## %else33
-; SSE4-NEXT: testb $64, %al
-; SSE4-NEXT: jne LBB32_13
-; SSE4-NEXT: LBB32_14: ## %else35
-; SSE4-NEXT: testb $-128, %al
-; SSE4-NEXT: jne LBB32_15
-; SSE4-NEXT: LBB32_16: ## %else37
-; SSE4-NEXT: retq
-; SSE4-NEXT: LBB32_1: ## %cond.store
-; SSE4-NEXT: movl $0, (%rsi)
-; SSE4-NEXT: testb $2, %al
-; SSE4-NEXT: je LBB32_4
-; SSE4-NEXT: LBB32_3: ## %cond.store24
-; SSE4-NEXT: movl $0, 4(%rsi)
-; SSE4-NEXT: testb $4, %al
-; SSE4-NEXT: je LBB32_6
-; SSE4-NEXT: LBB32_5: ## %cond.store26
-; SSE4-NEXT: movl $0, 8(%rsi)
-; SSE4-NEXT: testb $8, %al
-; SSE4-NEXT: je LBB32_8
-; SSE4-NEXT: LBB32_7: ## %cond.store28
-; SSE4-NEXT: movl $0, 12(%rsi)
-; SSE4-NEXT: testb $16, %al
-; SSE4-NEXT: je LBB32_10
-; SSE4-NEXT: LBB32_9: ## %cond.store30
-; SSE4-NEXT: movl $0, 16(%rsi)
-; SSE4-NEXT: testb $32, %al
-; SSE4-NEXT: je LBB32_12
-; SSE4-NEXT: LBB32_11: ## %cond.store32
-; SSE4-NEXT: movl $0, 20(%rsi)
-; SSE4-NEXT: testb $64, %al
-; SSE4-NEXT: je LBB32_14
-; SSE4-NEXT: LBB32_13: ## %cond.store34
-; SSE4-NEXT: movl $0, 24(%rsi)
-; SSE4-NEXT: testb $-128, %al
-; SSE4-NEXT: je LBB32_16
-; SSE4-NEXT: LBB32_15: ## %cond.store36
-; SSE4-NEXT: movl $0, 28(%rsi)
-; SSE4-NEXT: retq
+; SSE-LABEL: undefshuffle:
+; SSE: ## %bb.0:
+; SSE-NEXT: movb $1, %al
+; SSE-NEXT: testb %al, %al
+; SSE-NEXT: testb %al, %al
+; SSE-NEXT: testb %al, %al
+; SSE-NEXT: testb %al, %al
+; SSE-NEXT: testb %al, %al
+; SSE-NEXT: testb %al, %al
+; SSE-NEXT: testb %al, %al
+; SSE-NEXT: movq {{.*#+}} xmm0 = xmm0[0],zero
+; SSE-NEXT: testb %al, %al
+; SSE-NEXT: psllw $15, %xmm0
+; SSE-NEXT: packsswb %xmm0, %xmm0
+; SSE-NEXT: pmovmskb %xmm0, %eax
+; SSE-NEXT: testb $1, %al
+; SSE-NEXT: jne LBB32_1
+; SSE-NEXT: ## %bb.2: ## %else23
+; SSE-NEXT: testb $2, %al
+; SSE-NEXT: jne LBB32_3
+; SSE-NEXT: LBB32_4: ## %else25
+; SSE-NEXT: testb $4, %al
+; SSE-NEXT: jne LBB32_5
+; SSE-NEXT: LBB32_6: ## %else27
+; SSE-NEXT: testb $8, %al
+; SSE-NEXT: jne LBB32_7
+; SSE-NEXT: LBB32_8: ## %else29
+; SSE-NEXT: testb $16, %al
+; SSE-NEXT: jne LBB32_9
+; SSE-NEXT: LBB32_10: ## %else31
+; SSE-NEXT: testb $32, %al
+; SSE-NEXT: jne LBB32_11
+; SSE-NEXT: LBB32_12: ## %else33
+; SSE-NEXT: testb $64, %al
+; SSE-NEXT: jne LBB32_13
+; SSE-NEXT: LBB32_14: ## %else35
+; SSE-NEXT: testb $-128, %al
+; SSE-NEXT: jne LBB32_15
+; SSE-NEXT: LBB32_16: ## %else37
+; SSE-NEXT: retq
+; SSE-NEXT: LBB32_1: ## %cond.store
+; SSE-NEXT: movl $0, (%rsi)
+; SSE-NEXT: testb $2, %al
+; SSE-NEXT: je LBB32_4
+; SSE-NEXT: LBB32_3: ## %cond.store24
+; SSE-NEXT: movl $0, 4(%rsi)
+; SSE-NEXT: testb $4, %al
+; SSE-NEXT: je LBB32_6
+; SSE-NEXT: LBB32_5: ## %cond.store26
+; SSE-NEXT: movl $0, 8(%rsi)
+; SSE-NEXT: testb $8, %al
+; SSE-NEXT: je LBB32_8
+; SSE-NEXT: LBB32_7: ## %cond.store28
+; SSE-NEXT: movl $0, 12(%rsi)
+; SSE-NEXT: testb $16, %al
+; SSE-NEXT: je LBB32_10
+; SSE-NEXT: LBB32_9: ## %cond.store30
+; SSE-NEXT: movl $0, 16(%rsi)
+; SSE-NEXT: testb $32, %al
+; SSE-NEXT: je LBB32_12
+; SSE-NEXT: LBB32_11: ## %cond.store32
+; SSE-NEXT: movl $0, 20(%rsi)
+; SSE-NEXT: testb $64, %al
+; SSE-NEXT: je LBB32_14
+; SSE-NEXT: LBB32_13: ## %cond.store34
+; SSE-NEXT: movl $0, 24(%rsi)
+; SSE-NEXT: testb $-128, %al
+; SSE-NEXT: je LBB32_16
+; SSE-NEXT: LBB32_15: ## %cond.store36
+; SSE-NEXT: movl $0, 28(%rsi)
+; SSE-NEXT: retq
;
; AVX1-LABEL: undefshuffle:
; AVX1: ## %bb.0:
diff --git a/llvm/test/CodeGen/X86/midpoint-int-vec-512.ll b/llvm/test/CodeGen/X86/midpoint-int-vec-512.ll
index 0f10742..c2f8cbf 100644
--- a/llvm/test/CodeGen/X86/midpoint-int-vec-512.ll
+++ b/llvm/test/CodeGen/X86/midpoint-int-vec-512.ll
@@ -692,20 +692,18 @@
; AVX512F-NEXT: vpsubb %ymm5, %ymm1, %ymm1
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512F-NEXT: vpsrlw $1, %ymm2, %ymm2
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm5
-; AVX512F-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm5, %zmm5
-; AVX512F-NEXT: vpbroadcastb {{.*#+}} ymm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512F-NEXT: vpand %ymm6, %ymm2, %ymm2
-; AVX512F-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX512F-NEXT: vpsubb %ymm2, %ymm7, %ymm2
-; AVX512F-NEXT: vpand %ymm6, %ymm1, %ymm1
-; AVX512F-NEXT: vpsubb %ymm1, %ymm7, %ymm1
; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512F-NEXT: vpternlogq $226, %zmm5, %zmm4, %zmm1
+; AVX512F-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm1, %zmm1
; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; AVX512F-NEXT: vpaddb %ymm3, %ymm2, %ymm2
-; AVX512F-NEXT: vpaddb %ymm0, %ymm1, %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512F-NEXT: vpxor %xmm5, %xmm5, %xmm5
+; AVX512F-NEXT: vpsubb %ymm2, %ymm5, %ymm2
+; AVX512F-NEXT: vpsubb %ymm1, %ymm5, %ymm5
+; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT: vpternlogq $226, %zmm1, %zmm4, %zmm2
+; AVX512F-NEXT: vextracti64x4 $1, %zmm2, %ymm1
+; AVX512F-NEXT: vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT: vpaddb %ymm0, %ymm2, %ymm0
+; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-FALLBACK-LABEL: vec512_i8_signed_reg_reg:
@@ -723,20 +721,18 @@
; AVX512VL-FALLBACK-NEXT: vpsubb %ymm5, %ymm1, %ymm1
; AVX512VL-FALLBACK-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VL-FALLBACK-NEXT: vpsrlw $1, %ymm2, %ymm2
-; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm5
-; AVX512VL-FALLBACK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm5, %zmm5
-; AVX512VL-FALLBACK-NEXT: vpbroadcastd {{.*#+}} ymm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512VL-FALLBACK-NEXT: vpand %ymm6, %ymm2, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX512VL-FALLBACK-NEXT: vpsubb %ymm2, %ymm7, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpand %ymm6, %ymm1, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpsubb %ymm1, %ymm7, %ymm1
; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512VL-FALLBACK-NEXT: vpternlogq $226, %zmm5, %zmm4, %zmm1
+; AVX512VL-FALLBACK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm1, %zmm1
; AVX512VL-FALLBACK-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpaddb %ymm3, %ymm2, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpaddb %ymm0, %ymm1, %ymm0
-; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512VL-FALLBACK-NEXT: vpxor %xmm5, %xmm5, %xmm5
+; AVX512VL-FALLBACK-NEXT: vpsubb %ymm2, %ymm5, %ymm2
+; AVX512VL-FALLBACK-NEXT: vpsubb %ymm1, %ymm5, %ymm5
+; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm5, %zmm2
+; AVX512VL-FALLBACK-NEXT: vpternlogq $226, %zmm1, %zmm4, %zmm2
+; AVX512VL-FALLBACK-NEXT: vextracti64x4 $1, %zmm2, %ymm1
+; AVX512VL-FALLBACK-NEXT: vpaddb %ymm3, %ymm1, %ymm1
+; AVX512VL-FALLBACK-NEXT: vpaddb %ymm0, %ymm2, %ymm0
+; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512VL-FALLBACK-NEXT: retq
;
; AVX512BW-LABEL: vec512_i8_signed_reg_reg:
@@ -778,20 +774,18 @@
; AVX512F-NEXT: vpsubb %ymm6, %ymm1, %ymm1
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512F-NEXT: vpsrlw $1, %ymm2, %ymm2
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm4
-; AVX512F-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm4, %zmm4
-; AVX512F-NEXT: vpbroadcastb {{.*#+}} ymm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512F-NEXT: vpand %ymm6, %ymm2, %ymm2
-; AVX512F-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX512F-NEXT: vpsubb %ymm2, %ymm7, %ymm2
-; AVX512F-NEXT: vpand %ymm6, %ymm1, %ymm1
-; AVX512F-NEXT: vpsubb %ymm1, %ymm7, %ymm1
; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512F-NEXT: vpternlogq $184, %zmm4, %zmm5, %zmm1
+; AVX512F-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm1, %zmm1
; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; AVX512F-NEXT: vpaddb %ymm3, %ymm2, %ymm2
-; AVX512F-NEXT: vpaddb %ymm0, %ymm1, %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512F-NEXT: vpxor %xmm4, %xmm4, %xmm4
+; AVX512F-NEXT: vpsubb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT: vpsubb %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm4, %zmm2
+; AVX512F-NEXT: vpternlogq $216, %zmm5, %zmm1, %zmm2
+; AVX512F-NEXT: vextracti64x4 $1, %zmm2, %ymm1
+; AVX512F-NEXT: vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT: vpaddb %ymm0, %ymm2, %ymm0
+; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-FALLBACK-LABEL: vec512_i8_unsigned_reg_reg:
@@ -809,20 +803,18 @@
; AVX512VL-FALLBACK-NEXT: vpsubb %ymm6, %ymm1, %ymm1
; AVX512VL-FALLBACK-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VL-FALLBACK-NEXT: vpsrlw $1, %ymm2, %ymm2
-; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm4
-; AVX512VL-FALLBACK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm4, %zmm4
-; AVX512VL-FALLBACK-NEXT: vpbroadcastd {{.*#+}} ymm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512VL-FALLBACK-NEXT: vpand %ymm6, %ymm2, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX512VL-FALLBACK-NEXT: vpsubb %ymm2, %ymm7, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpand %ymm6, %ymm1, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpsubb %ymm1, %ymm7, %ymm1
; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512VL-FALLBACK-NEXT: vpternlogq $184, %zmm4, %zmm5, %zmm1
+; AVX512VL-FALLBACK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm1, %zmm1
; AVX512VL-FALLBACK-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpaddb %ymm3, %ymm2, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpaddb %ymm0, %ymm1, %ymm0
-; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512VL-FALLBACK-NEXT: vpxor %xmm4, %xmm4, %xmm4
+; AVX512VL-FALLBACK-NEXT: vpsubb %ymm2, %ymm4, %ymm2
+; AVX512VL-FALLBACK-NEXT: vpsubb %ymm1, %ymm4, %ymm4
+; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm4, %zmm2
+; AVX512VL-FALLBACK-NEXT: vpternlogq $216, %zmm5, %zmm1, %zmm2
+; AVX512VL-FALLBACK-NEXT: vextracti64x4 $1, %zmm2, %ymm1
+; AVX512VL-FALLBACK-NEXT: vpaddb %ymm3, %ymm1, %ymm1
+; AVX512VL-FALLBACK-NEXT: vpaddb %ymm0, %ymm2, %ymm0
+; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512VL-FALLBACK-NEXT: retq
;
; AVX512BW-LABEL: vec512_i8_unsigned_reg_reg:
@@ -867,20 +859,18 @@
; AVX512F-NEXT: vpsubb %ymm5, %ymm0, %ymm0
; AVX512F-NEXT: vpsrlw $1, %ymm0, %ymm0
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm5
-; AVX512F-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm5, %zmm5
-; AVX512F-NEXT: vpbroadcastb {{.*#+}} ymm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512F-NEXT: vpand %ymm6, %ymm1, %ymm1
-; AVX512F-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX512F-NEXT: vpsubb %ymm1, %ymm7, %ymm1
-; AVX512F-NEXT: vpand %ymm6, %ymm0, %ymm0
-; AVX512F-NEXT: vpsubb %ymm0, %ymm7, %ymm0
; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512F-NEXT: vpternlogq $226, %zmm5, %zmm4, %zmm0
+; AVX512F-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm0
; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; AVX512F-NEXT: vpaddb %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT: vpaddb %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT: vpxor %xmm5, %xmm5, %xmm5
+; AVX512F-NEXT: vpsubb %ymm1, %ymm5, %ymm1
+; AVX512F-NEXT: vpsubb %ymm0, %ymm5, %ymm5
+; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm5, %zmm1
+; AVX512F-NEXT: vpternlogq $226, %zmm0, %zmm4, %zmm1
+; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm0
+; AVX512F-NEXT: vpaddb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT: vpaddb %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-FALLBACK-LABEL: vec512_i8_signed_mem_reg:
@@ -899,20 +889,18 @@
; AVX512VL-FALLBACK-NEXT: vpsubb %ymm5, %ymm0, %ymm0
; AVX512VL-FALLBACK-NEXT: vpsrlw $1, %ymm0, %ymm0
; AVX512VL-FALLBACK-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm5
-; AVX512VL-FALLBACK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm5, %zmm5
-; AVX512VL-FALLBACK-NEXT: vpbroadcastd {{.*#+}} ymm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512VL-FALLBACK-NEXT: vpand %ymm6, %ymm1, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX512VL-FALLBACK-NEXT: vpsubb %ymm1, %ymm7, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpand %ymm6, %ymm0, %ymm0
-; AVX512VL-FALLBACK-NEXT: vpsubb %ymm0, %ymm7, %ymm0
; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512VL-FALLBACK-NEXT: vpternlogq $226, %zmm5, %zmm4, %zmm0
+; AVX512VL-FALLBACK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm0
; AVX512VL-FALLBACK-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpaddb %ymm3, %ymm1, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpaddb %ymm2, %ymm0, %ymm0
-; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512VL-FALLBACK-NEXT: vpxor %xmm5, %xmm5, %xmm5
+; AVX512VL-FALLBACK-NEXT: vpsubb %ymm1, %ymm5, %ymm1
+; AVX512VL-FALLBACK-NEXT: vpsubb %ymm0, %ymm5, %ymm5
+; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm5, %zmm1
+; AVX512VL-FALLBACK-NEXT: vpternlogq $226, %zmm0, %zmm4, %zmm1
+; AVX512VL-FALLBACK-NEXT: vextracti64x4 $1, %zmm1, %ymm0
+; AVX512VL-FALLBACK-NEXT: vpaddb %ymm3, %ymm0, %ymm0
+; AVX512VL-FALLBACK-NEXT: vpaddb %ymm2, %ymm1, %ymm1
+; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
; AVX512VL-FALLBACK-NEXT: retq
;
; AVX512BW-LABEL: vec512_i8_signed_mem_reg:
@@ -957,20 +945,18 @@
; AVX512F-NEXT: vpsubb %ymm5, %ymm1, %ymm1
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512F-NEXT: vpsrlw $1, %ymm2, %ymm2
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm5
-; AVX512F-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm5, %zmm5
-; AVX512F-NEXT: vpbroadcastb {{.*#+}} ymm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512F-NEXT: vpand %ymm6, %ymm2, %ymm2
-; AVX512F-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX512F-NEXT: vpsubb %ymm2, %ymm7, %ymm2
-; AVX512F-NEXT: vpand %ymm6, %ymm1, %ymm1
-; AVX512F-NEXT: vpsubb %ymm1, %ymm7, %ymm1
; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512F-NEXT: vpternlogq $226, %zmm5, %zmm4, %zmm1
+; AVX512F-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm1, %zmm1
; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; AVX512F-NEXT: vpaddb %ymm3, %ymm2, %ymm2
-; AVX512F-NEXT: vpaddb %ymm0, %ymm1, %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512F-NEXT: vpxor %xmm5, %xmm5, %xmm5
+; AVX512F-NEXT: vpsubb %ymm2, %ymm5, %ymm2
+; AVX512F-NEXT: vpsubb %ymm1, %ymm5, %ymm5
+; AVX512F-NEXT: vinserti64x4 $1, %ymm2, %zmm5, %zmm2
+; AVX512F-NEXT: vpternlogq $226, %zmm1, %zmm4, %zmm2
+; AVX512F-NEXT: vextracti64x4 $1, %zmm2, %ymm1
+; AVX512F-NEXT: vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT: vpaddb %ymm0, %ymm2, %ymm0
+; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-FALLBACK-LABEL: vec512_i8_signed_reg_mem:
@@ -989,20 +975,18 @@
; AVX512VL-FALLBACK-NEXT: vpsubb %ymm5, %ymm1, %ymm1
; AVX512VL-FALLBACK-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VL-FALLBACK-NEXT: vpsrlw $1, %ymm2, %ymm2
-; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm5
-; AVX512VL-FALLBACK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm5, %zmm5
-; AVX512VL-FALLBACK-NEXT: vpbroadcastd {{.*#+}} ymm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512VL-FALLBACK-NEXT: vpand %ymm6, %ymm2, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX512VL-FALLBACK-NEXT: vpsubb %ymm2, %ymm7, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpand %ymm6, %ymm1, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpsubb %ymm1, %ymm7, %ymm1
; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
-; AVX512VL-FALLBACK-NEXT: vpternlogq $226, %zmm5, %zmm4, %zmm1
+; AVX512VL-FALLBACK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm1, %zmm1
; AVX512VL-FALLBACK-NEXT: vextracti64x4 $1, %zmm1, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpaddb %ymm3, %ymm2, %ymm2
-; AVX512VL-FALLBACK-NEXT: vpaddb %ymm0, %ymm1, %ymm0
-; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512VL-FALLBACK-NEXT: vpxor %xmm5, %xmm5, %xmm5
+; AVX512VL-FALLBACK-NEXT: vpsubb %ymm2, %ymm5, %ymm2
+; AVX512VL-FALLBACK-NEXT: vpsubb %ymm1, %ymm5, %ymm5
+; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm2, %zmm5, %zmm2
+; AVX512VL-FALLBACK-NEXT: vpternlogq $226, %zmm1, %zmm4, %zmm2
+; AVX512VL-FALLBACK-NEXT: vextracti64x4 $1, %zmm2, %ymm1
+; AVX512VL-FALLBACK-NEXT: vpaddb %ymm3, %ymm1, %ymm1
+; AVX512VL-FALLBACK-NEXT: vpaddb %ymm0, %ymm2, %ymm0
+; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512VL-FALLBACK-NEXT: retq
;
; AVX512BW-LABEL: vec512_i8_signed_reg_mem:
@@ -1048,20 +1032,18 @@
; AVX512F-NEXT: vpsubb %ymm5, %ymm0, %ymm0
; AVX512F-NEXT: vpsrlw $1, %ymm0, %ymm0
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm5
-; AVX512F-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm5, %zmm5
-; AVX512F-NEXT: vpbroadcastb {{.*#+}} ymm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512F-NEXT: vpand %ymm6, %ymm1, %ymm1
-; AVX512F-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX512F-NEXT: vpsubb %ymm1, %ymm7, %ymm1
-; AVX512F-NEXT: vpand %ymm6, %ymm0, %ymm0
-; AVX512F-NEXT: vpsubb %ymm0, %ymm7, %ymm0
; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512F-NEXT: vpternlogq $226, %zmm5, %zmm4, %zmm0
+; AVX512F-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm0
; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; AVX512F-NEXT: vpaddb %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT: vpaddb %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT: vpxor %xmm5, %xmm5, %xmm5
+; AVX512F-NEXT: vpsubb %ymm1, %ymm5, %ymm1
+; AVX512F-NEXT: vpsubb %ymm0, %ymm5, %ymm5
+; AVX512F-NEXT: vinserti64x4 $1, %ymm1, %zmm5, %zmm1
+; AVX512F-NEXT: vpternlogq $226, %zmm0, %zmm4, %zmm1
+; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm0
+; AVX512F-NEXT: vpaddb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT: vpaddb %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-FALLBACK-LABEL: vec512_i8_signed_mem_mem:
@@ -1081,20 +1063,18 @@
; AVX512VL-FALLBACK-NEXT: vpsubb %ymm5, %ymm0, %ymm0
; AVX512VL-FALLBACK-NEXT: vpsrlw $1, %ymm0, %ymm0
; AVX512VL-FALLBACK-NEXT: vpsrlw $1, %ymm1, %ymm1
-; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm5
-; AVX512VL-FALLBACK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm5, %zmm5
-; AVX512VL-FALLBACK-NEXT: vpbroadcastd {{.*#+}} ymm6 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
-; AVX512VL-FALLBACK-NEXT: vpand %ymm6, %ymm1, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX512VL-FALLBACK-NEXT: vpsubb %ymm1, %ymm7, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpand %ymm6, %ymm0, %ymm0
-; AVX512VL-FALLBACK-NEXT: vpsubb %ymm0, %ymm7, %ymm0
; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512VL-FALLBACK-NEXT: vpternlogq $226, %zmm5, %zmm4, %zmm0
+; AVX512VL-FALLBACK-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm0
; AVX512VL-FALLBACK-NEXT: vextracti64x4 $1, %zmm0, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpaddb %ymm3, %ymm1, %ymm1
-; AVX512VL-FALLBACK-NEXT: vpaddb %ymm2, %ymm0, %ymm0
-; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512VL-FALLBACK-NEXT: vpxor %xmm5, %xmm5, %xmm5
+; AVX512VL-FALLBACK-NEXT: vpsubb %ymm1, %ymm5, %ymm1
+; AVX512VL-FALLBACK-NEXT: vpsubb %ymm0, %ymm5, %ymm5
+; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm1, %zmm5, %zmm1
+; AVX512VL-FALLBACK-NEXT: vpternlogq $226, %zmm0, %zmm4, %zmm1
+; AVX512VL-FALLBACK-NEXT: vextracti64x4 $1, %zmm1, %ymm0
+; AVX512VL-FALLBACK-NEXT: vpaddb %ymm3, %ymm0, %ymm0
+; AVX512VL-FALLBACK-NEXT: vpaddb %ymm2, %ymm1, %ymm1
+; AVX512VL-FALLBACK-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
; AVX512VL-FALLBACK-NEXT: retq
;
; AVX512BW-LABEL: vec512_i8_signed_mem_mem:
diff --git a/llvm/test/CodeGen/X86/movmsk-cmp.ll b/llvm/test/CodeGen/X86/movmsk-cmp.ll
index dffff76..38b77c6 100644
--- a/llvm/test/CodeGen/X86/movmsk-cmp.ll
+++ b/llvm/test/CodeGen/X86/movmsk-cmp.ll
@@ -1120,24 +1120,11 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: allzeros_v16i8_and1:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: sete %al
-; AVX1OR2-NEXT: retq
-;
-; KNL-LABEL: allzeros_v16i8_and1:
-; KNL: # %bb.0:
-; KNL-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; KNL-NEXT: sete %al
-; KNL-NEXT: retq
-;
-; SKX-LABEL: allzeros_v16i8_and1:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastq {{.*#+}} xmm1 = [72340172838076673,72340172838076673]
-; SKX-NEXT: vptest %xmm1, %xmm0
-; SKX-NEXT: sete %al
-; SKX-NEXT: retq
+; AVX-LABEL: allzeros_v16i8_and1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX-NEXT: sete %al
+; AVX-NEXT: retq
%tmp = and <16 x i8> %arg, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
%tmp1 = icmp ne <16 x i8> %tmp, zeroinitializer
%tmp2 = bitcast <16 x i1> %tmp1 to i16
@@ -1215,28 +1202,12 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1-LABEL: allzeros_v32i8_and1:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
-; AVX1-NEXT: sete %al
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: allzeros_v32i8_and1:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX2-NEXT: vptest %ymm1, %ymm0
-; AVX2-NEXT: sete %al
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
-;
-; AVX512-LABEL: allzeros_v32i8_and1:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX512-NEXT: vptest %ymm1, %ymm0
-; AVX512-NEXT: sete %al
-; AVX512-NEXT: vzeroupper
-; AVX512-NEXT: retq
+; AVX-LABEL: allzeros_v32i8_and1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
+; AVX-NEXT: sete %al
+; AVX-NEXT: vzeroupper
+; AVX-NEXT: retq
%tmp = and <32 x i8> %arg, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
%tmp1 = icmp ne <32 x i8> %tmp, zeroinitializer
%tmp2 = bitcast <32 x i1> %tmp1 to i32
@@ -1337,8 +1308,7 @@
; AVX2-LABEL: allzeros_v64i8_and1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX2-NEXT: vptest %ymm1, %ymm0
+; AVX2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX2-NEXT: sete %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
@@ -1416,24 +1386,11 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: allzeros_v8i16_and1:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: sete %al
-; AVX1OR2-NEXT: retq
-;
-; KNL-LABEL: allzeros_v8i16_and1:
-; KNL: # %bb.0:
-; KNL-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; KNL-NEXT: sete %al
-; KNL-NEXT: retq
-;
-; SKX-LABEL: allzeros_v8i16_and1:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastq {{.*#+}} xmm1 = [281479271743489,281479271743489]
-; SKX-NEXT: vptest %xmm1, %xmm0
-; SKX-NEXT: sete %al
-; SKX-NEXT: retq
+; AVX-LABEL: allzeros_v8i16_and1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX-NEXT: sete %al
+; AVX-NEXT: retq
%tmp = and <8 x i16> %arg, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
%tmp1 = icmp ne <8 x i16> %tmp, zeroinitializer
%tmp2 = bitcast <8 x i1> %tmp1 to i8
@@ -1608,8 +1565,7 @@
; AVX2-LABEL: allzeros_v32i16_and1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX2-NEXT: vptest %ymm1, %ymm0
+; AVX2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX2-NEXT: sete %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
@@ -1645,28 +1601,12 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1-LABEL: allzeros_v16i16_and1:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
-; AVX1-NEXT: sete %al
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: allzeros_v16i16_and1:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX2-NEXT: vptest %ymm1, %ymm0
-; AVX2-NEXT: sete %al
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
-;
-; AVX512-LABEL: allzeros_v16i16_and1:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX512-NEXT: vptest %ymm1, %ymm0
-; AVX512-NEXT: sete %al
-; AVX512-NEXT: vzeroupper
-; AVX512-NEXT: retq
+; AVX-LABEL: allzeros_v16i16_and1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
+; AVX-NEXT: sete %al
+; AVX-NEXT: vzeroupper
+; AVX-NEXT: retq
%tmp = and <16 x i16> %arg, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
%tmp1 = icmp ne <16 x i16> %tmp, zeroinitializer
%tmp2 = bitcast <16 x i1> %tmp1 to i16
@@ -1730,24 +1670,25 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: allzeros_v4i32_and1:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: sete %al
-; AVX1OR2-NEXT: retq
+; AVX1-LABEL: allzeros_v4i32_and1:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX1-NEXT: sete %al
+; AVX1-NEXT: retq
;
-; KNL-LABEL: allzeros_v4i32_and1:
-; KNL: # %bb.0:
-; KNL-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; KNL-NEXT: sete %al
-; KNL-NEXT: retq
+; AVX2-LABEL: allzeros_v4i32_and1:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; AVX2-NEXT: vptest %xmm1, %xmm0
+; AVX2-NEXT: sete %al
+; AVX2-NEXT: retq
;
-; SKX-LABEL: allzeros_v4i32_and1:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastq {{.*#+}} xmm1 = [4294967297,4294967297]
-; SKX-NEXT: vptest %xmm1, %xmm0
-; SKX-NEXT: sete %al
-; SKX-NEXT: retq
+; AVX512-LABEL: allzeros_v4i32_and1:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; AVX512-NEXT: vptest %xmm1, %xmm0
+; AVX512-NEXT: sete %al
+; AVX512-NEXT: retq
%tmp = and <4 x i32> %arg, <i32 1, i32 1, i32 1, i32 1>
%tmp1 = icmp ne <4 x i32> %tmp, zeroinitializer
%tmp2 = bitcast <4 x i1> %tmp1 to i4
@@ -1837,7 +1778,7 @@
;
; AVX2-LABEL: allzeros_v8i32_and1:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: sete %al
; AVX2-NEXT: vzeroupper
@@ -1845,7 +1786,7 @@
;
; AVX512-LABEL: allzeros_v8i32_and1:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX512-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX512-NEXT: vptest %ymm1, %ymm0
; AVX512-NEXT: sete %al
; AVX512-NEXT: vzeroupper
@@ -1948,7 +1889,7 @@
; AVX2-LABEL: allzeros_v16i32_and1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: sete %al
; AVX2-NEXT: vzeroupper
@@ -2325,24 +2266,11 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: allzeros_v16i8_and4:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: sete %al
-; AVX1OR2-NEXT: retq
-;
-; KNL-LABEL: allzeros_v16i8_and4:
-; KNL: # %bb.0:
-; KNL-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; KNL-NEXT: sete %al
-; KNL-NEXT: retq
-;
-; SKX-LABEL: allzeros_v16i8_and4:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastq {{.*#+}} xmm1 = [289360691352306692,289360691352306692]
-; SKX-NEXT: vptest %xmm1, %xmm0
-; SKX-NEXT: sete %al
-; SKX-NEXT: retq
+; AVX-LABEL: allzeros_v16i8_and4:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX-NEXT: sete %al
+; AVX-NEXT: retq
%tmp = and <16 x i8> %arg, <i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4>
%tmp1 = icmp ne <16 x i8> %tmp, zeroinitializer
%tmp2 = bitcast <16 x i1> %tmp1 to i16
@@ -2420,28 +2348,12 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1-LABEL: allzeros_v32i8_and4:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
-; AVX1-NEXT: sete %al
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: allzeros_v32i8_and4:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [289360691352306692,289360691352306692,289360691352306692,289360691352306692]
-; AVX2-NEXT: vptest %ymm1, %ymm0
-; AVX2-NEXT: sete %al
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
-;
-; AVX512-LABEL: allzeros_v32i8_and4:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [289360691352306692,289360691352306692,289360691352306692,289360691352306692]
-; AVX512-NEXT: vptest %ymm1, %ymm0
-; AVX512-NEXT: sete %al
-; AVX512-NEXT: vzeroupper
-; AVX512-NEXT: retq
+; AVX-LABEL: allzeros_v32i8_and4:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
+; AVX-NEXT: sete %al
+; AVX-NEXT: vzeroupper
+; AVX-NEXT: retq
%tmp = and <32 x i8> %arg, <i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4>
%tmp1 = icmp ne <32 x i8> %tmp, zeroinitializer
%tmp2 = bitcast <32 x i1> %tmp1 to i32
@@ -2542,8 +2454,7 @@
; AVX2-LABEL: allzeros_v64i8_and4:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [289360691352306692,289360691352306692,289360691352306692,289360691352306692]
-; AVX2-NEXT: vptest %ymm1, %ymm0
+; AVX2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX2-NEXT: sete %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
@@ -2621,24 +2532,11 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: allzeros_v8i16_and4:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: sete %al
-; AVX1OR2-NEXT: retq
-;
-; KNL-LABEL: allzeros_v8i16_and4:
-; KNL: # %bb.0:
-; KNL-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; KNL-NEXT: sete %al
-; KNL-NEXT: retq
-;
-; SKX-LABEL: allzeros_v8i16_and4:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastq {{.*#+}} xmm1 = [1125917086973956,1125917086973956]
-; SKX-NEXT: vptest %xmm1, %xmm0
-; SKX-NEXT: sete %al
-; SKX-NEXT: retq
+; AVX-LABEL: allzeros_v8i16_and4:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX-NEXT: sete %al
+; AVX-NEXT: retq
%tmp = and <8 x i16> %arg, <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4>
%tmp1 = icmp ne <8 x i16> %tmp, zeroinitializer
%tmp2 = bitcast <8 x i1> %tmp1 to i8
@@ -2813,8 +2711,7 @@
; AVX2-LABEL: allzeros_v32i16_and4:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [1125917086973956,1125917086973956,1125917086973956,1125917086973956]
-; AVX2-NEXT: vptest %ymm1, %ymm0
+; AVX2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX2-NEXT: sete %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
@@ -2850,28 +2747,12 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1-LABEL: allzeros_v16i16_and4:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
-; AVX1-NEXT: sete %al
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: allzeros_v16i16_and4:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [1125917086973956,1125917086973956,1125917086973956,1125917086973956]
-; AVX2-NEXT: vptest %ymm1, %ymm0
-; AVX2-NEXT: sete %al
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
-;
-; AVX512-LABEL: allzeros_v16i16_and4:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [1125917086973956,1125917086973956,1125917086973956,1125917086973956]
-; AVX512-NEXT: vptest %ymm1, %ymm0
-; AVX512-NEXT: sete %al
-; AVX512-NEXT: vzeroupper
-; AVX512-NEXT: retq
+; AVX-LABEL: allzeros_v16i16_and4:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
+; AVX-NEXT: sete %al
+; AVX-NEXT: vzeroupper
+; AVX-NEXT: retq
%tmp = and <16 x i16> %arg, <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4>
%tmp1 = icmp ne <16 x i16> %tmp, zeroinitializer
%tmp2 = bitcast <16 x i1> %tmp1 to i16
@@ -2935,24 +2816,25 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: allzeros_v4i32_and4:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: sete %al
-; AVX1OR2-NEXT: retq
+; AVX1-LABEL: allzeros_v4i32_and4:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX1-NEXT: sete %al
+; AVX1-NEXT: retq
;
-; KNL-LABEL: allzeros_v4i32_and4:
-; KNL: # %bb.0:
-; KNL-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; KNL-NEXT: sete %al
-; KNL-NEXT: retq
+; AVX2-LABEL: allzeros_v4i32_and4:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [4,4,4,4]
+; AVX2-NEXT: vptest %xmm1, %xmm0
+; AVX2-NEXT: sete %al
+; AVX2-NEXT: retq
;
-; SKX-LABEL: allzeros_v4i32_and4:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastq {{.*#+}} xmm1 = [17179869188,17179869188]
-; SKX-NEXT: vptest %xmm1, %xmm0
-; SKX-NEXT: sete %al
-; SKX-NEXT: retq
+; AVX512-LABEL: allzeros_v4i32_and4:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpbroadcastd {{.*#+}} xmm1 = [4,4,4,4]
+; AVX512-NEXT: vptest %xmm1, %xmm0
+; AVX512-NEXT: sete %al
+; AVX512-NEXT: retq
%tmp = and <4 x i32> %arg, <i32 4, i32 4, i32 4, i32 4>
%tmp1 = icmp ne <4 x i32> %tmp, zeroinitializer
%tmp2 = bitcast <4 x i1> %tmp1 to i4
@@ -3042,7 +2924,7 @@
;
; AVX2-LABEL: allzeros_v8i32_and4:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [17179869188,17179869188,17179869188,17179869188]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: sete %al
; AVX2-NEXT: vzeroupper
@@ -3050,7 +2932,7 @@
;
; AVX512-LABEL: allzeros_v8i32_and4:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [17179869188,17179869188,17179869188,17179869188]
+; AVX512-NEXT: vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
; AVX512-NEXT: vptest %ymm1, %ymm0
; AVX512-NEXT: sete %al
; AVX512-NEXT: vzeroupper
@@ -3153,7 +3035,7 @@
; AVX2-LABEL: allzeros_v16i32_and4:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [17179869188,17179869188,17179869188,17179869188]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: sete %al
; AVX2-NEXT: vzeroupper
diff --git a/llvm/test/CodeGen/X86/mulvi32.ll b/llvm/test/CodeGen/X86/mulvi32.ll
index 46d5da5..88bc356 100644
--- a/llvm/test/CodeGen/X86/mulvi32.ll
+++ b/llvm/test/CodeGen/X86/mulvi32.ll
@@ -134,31 +134,31 @@
define <4 x i64> @_mul4xi32toi64a(<4 x i32>, <4 x i32>) {
; SSE2-LABEL: _mul4xi32toi64a:
; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[2,1,3,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,1,3,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,1,1,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,1,1,3]
; SSE2-NEXT: pmuludq %xmm3, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; SSE2-NEXT: pmuludq %xmm1, %xmm0
-; SSE2-NEXT: movdqa %xmm2, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm1[2,1,3,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,1,3,3]
+; SSE2-NEXT: pmuludq %xmm3, %xmm1
+; SSE2-NEXT: movdqa %xmm2, %xmm0
; SSE2-NEXT: retq
;
; SSE42-LABEL: _mul4xi32toi64a:
; SSE42: # %bb.0:
-; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm0[2,1,3,3]
-; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,1,3,3]
+; SSE42-NEXT: pmovzxdq {{.*#+}} xmm3 = xmm1[0],zero,xmm1[1],zero
+; SSE42-NEXT: pmovzxdq {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero
; SSE42-NEXT: pmuludq %xmm3, %xmm2
-; SSE42-NEXT: pmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
-; SSE42-NEXT: pmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
-; SSE42-NEXT: pmuludq %xmm1, %xmm0
-; SSE42-NEXT: movdqa %xmm2, %xmm1
+; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm1[2,1,3,3]
+; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,1,3,3]
+; SSE42-NEXT: pmuludq %xmm3, %xmm1
+; SSE42-NEXT: movdqa %xmm2, %xmm0
; SSE42-NEXT: retq
;
; AVX1-LABEL: _mul4xi32toi64a:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[2,2,3,3]
-; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm1[2,2,3,3]
-; AVX1-NEXT: vpmuludq %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[2,1,3,3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[2,1,3,3]
+; AVX1-NEXT: vpmuludq %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
; AVX1-NEXT: vpmuludq %xmm1, %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/nontemporal-3.ll b/llvm/test/CodeGen/X86/nontemporal-3.ll
index a2d2c5c..3799bd6 100644
--- a/llvm/test/CodeGen/X86/nontemporal-3.ll
+++ b/llvm/test/CodeGen/X86/nontemporal-3.ll
@@ -494,14 +494,14 @@
; CHECK-LABEL: test_zero_v8f64_align1:
; CHECK: # %bb.0:
; CHECK-NEXT: xorl %eax, %eax
-; CHECK-NEXT: movntiq %rax, 8(%rdi)
-; CHECK-NEXT: movntiq %rax, (%rdi)
; CHECK-NEXT: movntiq %rax, 24(%rdi)
; CHECK-NEXT: movntiq %rax, 16(%rdi)
-; CHECK-NEXT: movntiq %rax, 40(%rdi)
-; CHECK-NEXT: movntiq %rax, 32(%rdi)
+; CHECK-NEXT: movntiq %rax, 8(%rdi)
+; CHECK-NEXT: movntiq %rax, (%rdi)
; CHECK-NEXT: movntiq %rax, 56(%rdi)
; CHECK-NEXT: movntiq %rax, 48(%rdi)
+; CHECK-NEXT: movntiq %rax, 40(%rdi)
+; CHECK-NEXT: movntiq %rax, 32(%rdi)
; CHECK-NEXT: retq
store <8 x double> zeroinitializer, ptr %dst, align 1, !nontemporal !1
ret void
@@ -511,67 +511,67 @@
; SSE2-LABEL: test_zero_v16f32_align1:
; SSE2: # %bb.0:
; SSE2-NEXT: xorl %eax, %eax
-; SSE2-NEXT: movntiq %rax, 8(%rdi)
-; SSE2-NEXT: movntiq %rax, (%rdi)
; SSE2-NEXT: movntiq %rax, 24(%rdi)
; SSE2-NEXT: movntiq %rax, 16(%rdi)
-; SSE2-NEXT: movntiq %rax, 40(%rdi)
-; SSE2-NEXT: movntiq %rax, 32(%rdi)
+; SSE2-NEXT: movntiq %rax, 8(%rdi)
+; SSE2-NEXT: movntiq %rax, (%rdi)
; SSE2-NEXT: movntiq %rax, 56(%rdi)
; SSE2-NEXT: movntiq %rax, 48(%rdi)
+; SSE2-NEXT: movntiq %rax, 40(%rdi)
+; SSE2-NEXT: movntiq %rax, 32(%rdi)
; SSE2-NEXT: retq
;
; SSE4A-LABEL: test_zero_v16f32_align1:
; SSE4A: # %bb.0:
; SSE4A-NEXT: xorl %eax, %eax
-; SSE4A-NEXT: movntiq %rax, 8(%rdi)
; SSE4A-NEXT: movntiq %rax, 24(%rdi)
-; SSE4A-NEXT: movntiq %rax, 40(%rdi)
+; SSE4A-NEXT: movntiq %rax, 8(%rdi)
; SSE4A-NEXT: movntiq %rax, 56(%rdi)
+; SSE4A-NEXT: movntiq %rax, 40(%rdi)
; SSE4A-NEXT: xorps %xmm0, %xmm0
-; SSE4A-NEXT: movntsd %xmm0, (%rdi)
; SSE4A-NEXT: movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, 32(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, (%rdi)
; SSE4A-NEXT: movntsd %xmm0, 48(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 32(%rdi)
; SSE4A-NEXT: retq
;
; SSE41-LABEL: test_zero_v16f32_align1:
; SSE41: # %bb.0:
; SSE41-NEXT: xorl %eax, %eax
-; SSE41-NEXT: movntiq %rax, 8(%rdi)
-; SSE41-NEXT: movntiq %rax, (%rdi)
; SSE41-NEXT: movntiq %rax, 24(%rdi)
; SSE41-NEXT: movntiq %rax, 16(%rdi)
-; SSE41-NEXT: movntiq %rax, 40(%rdi)
-; SSE41-NEXT: movntiq %rax, 32(%rdi)
+; SSE41-NEXT: movntiq %rax, 8(%rdi)
+; SSE41-NEXT: movntiq %rax, (%rdi)
; SSE41-NEXT: movntiq %rax, 56(%rdi)
; SSE41-NEXT: movntiq %rax, 48(%rdi)
+; SSE41-NEXT: movntiq %rax, 40(%rdi)
+; SSE41-NEXT: movntiq %rax, 32(%rdi)
; SSE41-NEXT: retq
;
; AVX-LABEL: test_zero_v16f32_align1:
; AVX: # %bb.0:
; AVX-NEXT: xorl %eax, %eax
-; AVX-NEXT: movntiq %rax, 8(%rdi)
-; AVX-NEXT: movntiq %rax, (%rdi)
; AVX-NEXT: movntiq %rax, 24(%rdi)
; AVX-NEXT: movntiq %rax, 16(%rdi)
-; AVX-NEXT: movntiq %rax, 40(%rdi)
-; AVX-NEXT: movntiq %rax, 32(%rdi)
+; AVX-NEXT: movntiq %rax, 8(%rdi)
+; AVX-NEXT: movntiq %rax, (%rdi)
; AVX-NEXT: movntiq %rax, 56(%rdi)
; AVX-NEXT: movntiq %rax, 48(%rdi)
+; AVX-NEXT: movntiq %rax, 40(%rdi)
+; AVX-NEXT: movntiq %rax, 32(%rdi)
; AVX-NEXT: retq
;
; AVX512-LABEL: test_zero_v16f32_align1:
; AVX512: # %bb.0:
; AVX512-NEXT: xorl %eax, %eax
-; AVX512-NEXT: movntiq %rax, 8(%rdi)
-; AVX512-NEXT: movntiq %rax, (%rdi)
; AVX512-NEXT: movntiq %rax, 24(%rdi)
; AVX512-NEXT: movntiq %rax, 16(%rdi)
-; AVX512-NEXT: movntiq %rax, 40(%rdi)
-; AVX512-NEXT: movntiq %rax, 32(%rdi)
+; AVX512-NEXT: movntiq %rax, 8(%rdi)
+; AVX512-NEXT: movntiq %rax, (%rdi)
; AVX512-NEXT: movntiq %rax, 56(%rdi)
; AVX512-NEXT: movntiq %rax, 48(%rdi)
+; AVX512-NEXT: movntiq %rax, 40(%rdi)
+; AVX512-NEXT: movntiq %rax, 32(%rdi)
; AVX512-NEXT: retq
store <16 x float> zeroinitializer, ptr %dst, align 1, !nontemporal !1
ret void
@@ -581,66 +581,66 @@
; SSE2-LABEL: test_zero_v8i64_align1:
; SSE2: # %bb.0:
; SSE2-NEXT: xorl %eax, %eax
-; SSE2-NEXT: movntiq %rax, 8(%rdi)
-; SSE2-NEXT: movntiq %rax, (%rdi)
; SSE2-NEXT: movntiq %rax, 24(%rdi)
; SSE2-NEXT: movntiq %rax, 16(%rdi)
-; SSE2-NEXT: movntiq %rax, 40(%rdi)
-; SSE2-NEXT: movntiq %rax, 32(%rdi)
+; SSE2-NEXT: movntiq %rax, 8(%rdi)
+; SSE2-NEXT: movntiq %rax, (%rdi)
; SSE2-NEXT: movntiq %rax, 56(%rdi)
; SSE2-NEXT: movntiq %rax, 48(%rdi)
+; SSE2-NEXT: movntiq %rax, 40(%rdi)
+; SSE2-NEXT: movntiq %rax, 32(%rdi)
; SSE2-NEXT: retq
;
; SSE4A-LABEL: test_zero_v8i64_align1:
; SSE4A: # %bb.0:
; SSE4A-NEXT: xorps %xmm0, %xmm0
-; SSE4A-NEXT: movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, (%rdi)
; SSE4A-NEXT: movntsd %xmm0, 24(%rdi)
; SSE4A-NEXT: movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, 40(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, 32(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 8(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, (%rdi)
; SSE4A-NEXT: movntsd %xmm0, 56(%rdi)
; SSE4A-NEXT: movntsd %xmm0, 48(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 40(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 32(%rdi)
; SSE4A-NEXT: retq
;
; SSE41-LABEL: test_zero_v8i64_align1:
; SSE41: # %bb.0:
; SSE41-NEXT: xorl %eax, %eax
-; SSE41-NEXT: movntiq %rax, 8(%rdi)
-; SSE41-NEXT: movntiq %rax, (%rdi)
; SSE41-NEXT: movntiq %rax, 24(%rdi)
; SSE41-NEXT: movntiq %rax, 16(%rdi)
-; SSE41-NEXT: movntiq %rax, 40(%rdi)
-; SSE41-NEXT: movntiq %rax, 32(%rdi)
+; SSE41-NEXT: movntiq %rax, 8(%rdi)
+; SSE41-NEXT: movntiq %rax, (%rdi)
; SSE41-NEXT: movntiq %rax, 56(%rdi)
; SSE41-NEXT: movntiq %rax, 48(%rdi)
+; SSE41-NEXT: movntiq %rax, 40(%rdi)
+; SSE41-NEXT: movntiq %rax, 32(%rdi)
; SSE41-NEXT: retq
;
; AVX-LABEL: test_zero_v8i64_align1:
; AVX: # %bb.0:
; AVX-NEXT: xorl %eax, %eax
-; AVX-NEXT: movntiq %rax, 8(%rdi)
-; AVX-NEXT: movntiq %rax, (%rdi)
; AVX-NEXT: movntiq %rax, 24(%rdi)
; AVX-NEXT: movntiq %rax, 16(%rdi)
-; AVX-NEXT: movntiq %rax, 40(%rdi)
-; AVX-NEXT: movntiq %rax, 32(%rdi)
+; AVX-NEXT: movntiq %rax, 8(%rdi)
+; AVX-NEXT: movntiq %rax, (%rdi)
; AVX-NEXT: movntiq %rax, 56(%rdi)
; AVX-NEXT: movntiq %rax, 48(%rdi)
+; AVX-NEXT: movntiq %rax, 40(%rdi)
+; AVX-NEXT: movntiq %rax, 32(%rdi)
; AVX-NEXT: retq
;
; AVX512-LABEL: test_zero_v8i64_align1:
; AVX512: # %bb.0:
; AVX512-NEXT: xorl %eax, %eax
-; AVX512-NEXT: movntiq %rax, 8(%rdi)
-; AVX512-NEXT: movntiq %rax, (%rdi)
; AVX512-NEXT: movntiq %rax, 24(%rdi)
; AVX512-NEXT: movntiq %rax, 16(%rdi)
-; AVX512-NEXT: movntiq %rax, 40(%rdi)
-; AVX512-NEXT: movntiq %rax, 32(%rdi)
+; AVX512-NEXT: movntiq %rax, 8(%rdi)
+; AVX512-NEXT: movntiq %rax, (%rdi)
; AVX512-NEXT: movntiq %rax, 56(%rdi)
; AVX512-NEXT: movntiq %rax, 48(%rdi)
+; AVX512-NEXT: movntiq %rax, 40(%rdi)
+; AVX512-NEXT: movntiq %rax, 32(%rdi)
; AVX512-NEXT: retq
store <8 x i64> zeroinitializer, ptr %dst, align 1, !nontemporal !1
ret void
@@ -650,66 +650,66 @@
; SSE2-LABEL: test_zero_v16i32_align1:
; SSE2: # %bb.0:
; SSE2-NEXT: xorl %eax, %eax
-; SSE2-NEXT: movntiq %rax, 8(%rdi)
-; SSE2-NEXT: movntiq %rax, (%rdi)
; SSE2-NEXT: movntiq %rax, 24(%rdi)
; SSE2-NEXT: movntiq %rax, 16(%rdi)
-; SSE2-NEXT: movntiq %rax, 40(%rdi)
-; SSE2-NEXT: movntiq %rax, 32(%rdi)
+; SSE2-NEXT: movntiq %rax, 8(%rdi)
+; SSE2-NEXT: movntiq %rax, (%rdi)
; SSE2-NEXT: movntiq %rax, 56(%rdi)
; SSE2-NEXT: movntiq %rax, 48(%rdi)
+; SSE2-NEXT: movntiq %rax, 40(%rdi)
+; SSE2-NEXT: movntiq %rax, 32(%rdi)
; SSE2-NEXT: retq
;
; SSE4A-LABEL: test_zero_v16i32_align1:
; SSE4A: # %bb.0:
; SSE4A-NEXT: xorps %xmm0, %xmm0
-; SSE4A-NEXT: movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, (%rdi)
; SSE4A-NEXT: movntsd %xmm0, 24(%rdi)
; SSE4A-NEXT: movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, 40(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, 32(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 8(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, (%rdi)
; SSE4A-NEXT: movntsd %xmm0, 56(%rdi)
; SSE4A-NEXT: movntsd %xmm0, 48(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 40(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 32(%rdi)
; SSE4A-NEXT: retq
;
; SSE41-LABEL: test_zero_v16i32_align1:
; SSE41: # %bb.0:
; SSE41-NEXT: xorl %eax, %eax
-; SSE41-NEXT: movntiq %rax, 8(%rdi)
-; SSE41-NEXT: movntiq %rax, (%rdi)
; SSE41-NEXT: movntiq %rax, 24(%rdi)
; SSE41-NEXT: movntiq %rax, 16(%rdi)
-; SSE41-NEXT: movntiq %rax, 40(%rdi)
-; SSE41-NEXT: movntiq %rax, 32(%rdi)
+; SSE41-NEXT: movntiq %rax, 8(%rdi)
+; SSE41-NEXT: movntiq %rax, (%rdi)
; SSE41-NEXT: movntiq %rax, 56(%rdi)
; SSE41-NEXT: movntiq %rax, 48(%rdi)
+; SSE41-NEXT: movntiq %rax, 40(%rdi)
+; SSE41-NEXT: movntiq %rax, 32(%rdi)
; SSE41-NEXT: retq
;
; AVX-LABEL: test_zero_v16i32_align1:
; AVX: # %bb.0:
; AVX-NEXT: xorl %eax, %eax
-; AVX-NEXT: movntiq %rax, 8(%rdi)
-; AVX-NEXT: movntiq %rax, (%rdi)
; AVX-NEXT: movntiq %rax, 24(%rdi)
; AVX-NEXT: movntiq %rax, 16(%rdi)
-; AVX-NEXT: movntiq %rax, 40(%rdi)
-; AVX-NEXT: movntiq %rax, 32(%rdi)
+; AVX-NEXT: movntiq %rax, 8(%rdi)
+; AVX-NEXT: movntiq %rax, (%rdi)
; AVX-NEXT: movntiq %rax, 56(%rdi)
; AVX-NEXT: movntiq %rax, 48(%rdi)
+; AVX-NEXT: movntiq %rax, 40(%rdi)
+; AVX-NEXT: movntiq %rax, 32(%rdi)
; AVX-NEXT: retq
;
; AVX512-LABEL: test_zero_v16i32_align1:
; AVX512: # %bb.0:
; AVX512-NEXT: xorl %eax, %eax
-; AVX512-NEXT: movntiq %rax, 8(%rdi)
-; AVX512-NEXT: movntiq %rax, (%rdi)
; AVX512-NEXT: movntiq %rax, 24(%rdi)
; AVX512-NEXT: movntiq %rax, 16(%rdi)
-; AVX512-NEXT: movntiq %rax, 40(%rdi)
-; AVX512-NEXT: movntiq %rax, 32(%rdi)
+; AVX512-NEXT: movntiq %rax, 8(%rdi)
+; AVX512-NEXT: movntiq %rax, (%rdi)
; AVX512-NEXT: movntiq %rax, 56(%rdi)
; AVX512-NEXT: movntiq %rax, 48(%rdi)
+; AVX512-NEXT: movntiq %rax, 40(%rdi)
+; AVX512-NEXT: movntiq %rax, 32(%rdi)
; AVX512-NEXT: retq
store <16 x i32> zeroinitializer, ptr %dst, align 1, !nontemporal !1
ret void
@@ -719,66 +719,66 @@
; SSE2-LABEL: test_zero_v32i16_align1:
; SSE2: # %bb.0:
; SSE2-NEXT: xorl %eax, %eax
-; SSE2-NEXT: movntiq %rax, 8(%rdi)
-; SSE2-NEXT: movntiq %rax, (%rdi)
; SSE2-NEXT: movntiq %rax, 24(%rdi)
; SSE2-NEXT: movntiq %rax, 16(%rdi)
-; SSE2-NEXT: movntiq %rax, 40(%rdi)
-; SSE2-NEXT: movntiq %rax, 32(%rdi)
+; SSE2-NEXT: movntiq %rax, 8(%rdi)
+; SSE2-NEXT: movntiq %rax, (%rdi)
; SSE2-NEXT: movntiq %rax, 56(%rdi)
; SSE2-NEXT: movntiq %rax, 48(%rdi)
+; SSE2-NEXT: movntiq %rax, 40(%rdi)
+; SSE2-NEXT: movntiq %rax, 32(%rdi)
; SSE2-NEXT: retq
;
; SSE4A-LABEL: test_zero_v32i16_align1:
; SSE4A: # %bb.0:
; SSE4A-NEXT: xorps %xmm0, %xmm0
-; SSE4A-NEXT: movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, (%rdi)
; SSE4A-NEXT: movntsd %xmm0, 24(%rdi)
; SSE4A-NEXT: movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, 40(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, 32(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 8(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, (%rdi)
; SSE4A-NEXT: movntsd %xmm0, 56(%rdi)
; SSE4A-NEXT: movntsd %xmm0, 48(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 40(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 32(%rdi)
; SSE4A-NEXT: retq
;
; SSE41-LABEL: test_zero_v32i16_align1:
; SSE41: # %bb.0:
; SSE41-NEXT: xorl %eax, %eax
-; SSE41-NEXT: movntiq %rax, 8(%rdi)
-; SSE41-NEXT: movntiq %rax, (%rdi)
; SSE41-NEXT: movntiq %rax, 24(%rdi)
; SSE41-NEXT: movntiq %rax, 16(%rdi)
-; SSE41-NEXT: movntiq %rax, 40(%rdi)
-; SSE41-NEXT: movntiq %rax, 32(%rdi)
+; SSE41-NEXT: movntiq %rax, 8(%rdi)
+; SSE41-NEXT: movntiq %rax, (%rdi)
; SSE41-NEXT: movntiq %rax, 56(%rdi)
; SSE41-NEXT: movntiq %rax, 48(%rdi)
+; SSE41-NEXT: movntiq %rax, 40(%rdi)
+; SSE41-NEXT: movntiq %rax, 32(%rdi)
; SSE41-NEXT: retq
;
; AVX-LABEL: test_zero_v32i16_align1:
; AVX: # %bb.0:
; AVX-NEXT: xorl %eax, %eax
-; AVX-NEXT: movntiq %rax, 8(%rdi)
-; AVX-NEXT: movntiq %rax, (%rdi)
; AVX-NEXT: movntiq %rax, 24(%rdi)
; AVX-NEXT: movntiq %rax, 16(%rdi)
-; AVX-NEXT: movntiq %rax, 40(%rdi)
-; AVX-NEXT: movntiq %rax, 32(%rdi)
+; AVX-NEXT: movntiq %rax, 8(%rdi)
+; AVX-NEXT: movntiq %rax, (%rdi)
; AVX-NEXT: movntiq %rax, 56(%rdi)
; AVX-NEXT: movntiq %rax, 48(%rdi)
+; AVX-NEXT: movntiq %rax, 40(%rdi)
+; AVX-NEXT: movntiq %rax, 32(%rdi)
; AVX-NEXT: retq
;
; AVX512-LABEL: test_zero_v32i16_align1:
; AVX512: # %bb.0:
; AVX512-NEXT: xorl %eax, %eax
-; AVX512-NEXT: movntiq %rax, 8(%rdi)
-; AVX512-NEXT: movntiq %rax, (%rdi)
; AVX512-NEXT: movntiq %rax, 24(%rdi)
; AVX512-NEXT: movntiq %rax, 16(%rdi)
-; AVX512-NEXT: movntiq %rax, 40(%rdi)
-; AVX512-NEXT: movntiq %rax, 32(%rdi)
+; AVX512-NEXT: movntiq %rax, 8(%rdi)
+; AVX512-NEXT: movntiq %rax, (%rdi)
; AVX512-NEXT: movntiq %rax, 56(%rdi)
; AVX512-NEXT: movntiq %rax, 48(%rdi)
+; AVX512-NEXT: movntiq %rax, 40(%rdi)
+; AVX512-NEXT: movntiq %rax, 32(%rdi)
; AVX512-NEXT: retq
store <32 x i16> zeroinitializer, ptr %dst, align 1, !nontemporal !1
ret void
@@ -788,66 +788,66 @@
; SSE2-LABEL: test_zero_v64i8_align1:
; SSE2: # %bb.0:
; SSE2-NEXT: xorl %eax, %eax
-; SSE2-NEXT: movntiq %rax, 8(%rdi)
-; SSE2-NEXT: movntiq %rax, (%rdi)
; SSE2-NEXT: movntiq %rax, 24(%rdi)
; SSE2-NEXT: movntiq %rax, 16(%rdi)
-; SSE2-NEXT: movntiq %rax, 40(%rdi)
-; SSE2-NEXT: movntiq %rax, 32(%rdi)
+; SSE2-NEXT: movntiq %rax, 8(%rdi)
+; SSE2-NEXT: movntiq %rax, (%rdi)
; SSE2-NEXT: movntiq %rax, 56(%rdi)
; SSE2-NEXT: movntiq %rax, 48(%rdi)
+; SSE2-NEXT: movntiq %rax, 40(%rdi)
+; SSE2-NEXT: movntiq %rax, 32(%rdi)
; SSE2-NEXT: retq
;
; SSE4A-LABEL: test_zero_v64i8_align1:
; SSE4A: # %bb.0:
; SSE4A-NEXT: xorps %xmm0, %xmm0
-; SSE4A-NEXT: movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, (%rdi)
; SSE4A-NEXT: movntsd %xmm0, 24(%rdi)
; SSE4A-NEXT: movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, 40(%rdi)
-; SSE4A-NEXT: movntsd %xmm0, 32(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 8(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, (%rdi)
; SSE4A-NEXT: movntsd %xmm0, 56(%rdi)
; SSE4A-NEXT: movntsd %xmm0, 48(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 40(%rdi)
+; SSE4A-NEXT: movntsd %xmm0, 32(%rdi)
; SSE4A-NEXT: retq
;
; SSE41-LABEL: test_zero_v64i8_align1:
; SSE41: # %bb.0:
; SSE41-NEXT: xorl %eax, %eax
-; SSE41-NEXT: movntiq %rax, 8(%rdi)
-; SSE41-NEXT: movntiq %rax, (%rdi)
; SSE41-NEXT: movntiq %rax, 24(%rdi)
; SSE41-NEXT: movntiq %rax, 16(%rdi)
-; SSE41-NEXT: movntiq %rax, 40(%rdi)
-; SSE41-NEXT: movntiq %rax, 32(%rdi)
+; SSE41-NEXT: movntiq %rax, 8(%rdi)
+; SSE41-NEXT: movntiq %rax, (%rdi)
; SSE41-NEXT: movntiq %rax, 56(%rdi)
; SSE41-NEXT: movntiq %rax, 48(%rdi)
+; SSE41-NEXT: movntiq %rax, 40(%rdi)
+; SSE41-NEXT: movntiq %rax, 32(%rdi)
; SSE41-NEXT: retq
;
; AVX-LABEL: test_zero_v64i8_align1:
; AVX: # %bb.0:
; AVX-NEXT: xorl %eax, %eax
-; AVX-NEXT: movntiq %rax, 8(%rdi)
-; AVX-NEXT: movntiq %rax, (%rdi)
; AVX-NEXT: movntiq %rax, 24(%rdi)
; AVX-NEXT: movntiq %rax, 16(%rdi)
-; AVX-NEXT: movntiq %rax, 40(%rdi)
-; AVX-NEXT: movntiq %rax, 32(%rdi)
+; AVX-NEXT: movntiq %rax, 8(%rdi)
+; AVX-NEXT: movntiq %rax, (%rdi)
; AVX-NEXT: movntiq %rax, 56(%rdi)
; AVX-NEXT: movntiq %rax, 48(%rdi)
+; AVX-NEXT: movntiq %rax, 40(%rdi)
+; AVX-NEXT: movntiq %rax, 32(%rdi)
; AVX-NEXT: retq
;
; AVX512-LABEL: test_zero_v64i8_align1:
; AVX512: # %bb.0:
; AVX512-NEXT: xorl %eax, %eax
-; AVX512-NEXT: movntiq %rax, 8(%rdi)
-; AVX512-NEXT: movntiq %rax, (%rdi)
; AVX512-NEXT: movntiq %rax, 24(%rdi)
; AVX512-NEXT: movntiq %rax, 16(%rdi)
-; AVX512-NEXT: movntiq %rax, 40(%rdi)
-; AVX512-NEXT: movntiq %rax, 32(%rdi)
+; AVX512-NEXT: movntiq %rax, 8(%rdi)
+; AVX512-NEXT: movntiq %rax, (%rdi)
; AVX512-NEXT: movntiq %rax, 56(%rdi)
; AVX512-NEXT: movntiq %rax, 48(%rdi)
+; AVX512-NEXT: movntiq %rax, 40(%rdi)
+; AVX512-NEXT: movntiq %rax, 32(%rdi)
; AVX512-NEXT: retq
store <64 x i8> zeroinitializer, ptr %dst, align 1, !nontemporal !1
ret void
diff --git a/llvm/test/CodeGen/X86/pmulh.ll b/llvm/test/CodeGen/X86/pmulh.ll
index 96613b6..f28f07e 100644
--- a/llvm/test/CodeGen/X86/pmulh.ll
+++ b/llvm/test/CodeGen/X86/pmulh.ll
@@ -319,41 +319,41 @@
; SSE2-LABEL: and_mulhuw_v16i16:
; SSE2: # %bb.0:
; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [32767,32767,32767,32767]
-; SSE2-NEXT: pand %xmm8, %xmm1
-; SSE2-NEXT: pand %xmm8, %xmm0
-; SSE2-NEXT: packssdw %xmm1, %xmm0
; SSE2-NEXT: pand %xmm8, %xmm3
; SSE2-NEXT: pand %xmm8, %xmm2
; SSE2-NEXT: packssdw %xmm3, %xmm2
-; SSE2-NEXT: pand %xmm8, %xmm5
-; SSE2-NEXT: pand %xmm8, %xmm4
-; SSE2-NEXT: packssdw %xmm5, %xmm4
-; SSE2-NEXT: pmulhw %xmm4, %xmm0
+; SSE2-NEXT: pand %xmm8, %xmm1
+; SSE2-NEXT: pand %xmm8, %xmm0
+; SSE2-NEXT: packssdw %xmm1, %xmm0
; SSE2-NEXT: pand %xmm8, %xmm7
-; SSE2-NEXT: pand %xmm6, %xmm8
-; SSE2-NEXT: packssdw %xmm7, %xmm8
-; SSE2-NEXT: pmulhw %xmm2, %xmm8
-; SSE2-NEXT: movdqa %xmm8, %xmm1
+; SSE2-NEXT: pand %xmm8, %xmm6
+; SSE2-NEXT: packssdw %xmm7, %xmm6
+; SSE2-NEXT: pmulhw %xmm2, %xmm6
+; SSE2-NEXT: pand %xmm8, %xmm5
+; SSE2-NEXT: pand %xmm4, %xmm8
+; SSE2-NEXT: packssdw %xmm5, %xmm8
+; SSE2-NEXT: pmulhw %xmm8, %xmm0
+; SSE2-NEXT: movdqa %xmm6, %xmm1
; SSE2-NEXT: retq
;
; SSE41-LABEL: and_mulhuw_v16i16:
; SSE41: # %bb.0:
; SSE41-NEXT: movdqa {{.*#+}} xmm8 = [32767,32767,32767,32767]
-; SSE41-NEXT: pand %xmm8, %xmm1
-; SSE41-NEXT: pand %xmm8, %xmm0
-; SSE41-NEXT: packusdw %xmm1, %xmm0
; SSE41-NEXT: pand %xmm8, %xmm3
; SSE41-NEXT: pand %xmm8, %xmm2
; SSE41-NEXT: packusdw %xmm3, %xmm2
-; SSE41-NEXT: pand %xmm8, %xmm5
-; SSE41-NEXT: pand %xmm8, %xmm4
-; SSE41-NEXT: packusdw %xmm5, %xmm4
-; SSE41-NEXT: pmulhw %xmm4, %xmm0
+; SSE41-NEXT: pand %xmm8, %xmm1
+; SSE41-NEXT: pand %xmm8, %xmm0
+; SSE41-NEXT: packusdw %xmm1, %xmm0
; SSE41-NEXT: pand %xmm8, %xmm7
-; SSE41-NEXT: pand %xmm6, %xmm8
-; SSE41-NEXT: packusdw %xmm7, %xmm8
-; SSE41-NEXT: pmulhw %xmm2, %xmm8
-; SSE41-NEXT: movdqa %xmm8, %xmm1
+; SSE41-NEXT: pand %xmm8, %xmm6
+; SSE41-NEXT: packusdw %xmm7, %xmm6
+; SSE41-NEXT: pmulhw %xmm2, %xmm6
+; SSE41-NEXT: pand %xmm8, %xmm5
+; SSE41-NEXT: pand %xmm4, %xmm8
+; SSE41-NEXT: packusdw %xmm5, %xmm8
+; SSE41-NEXT: pmulhw %xmm8, %xmm0
+; SSE41-NEXT: movdqa %xmm6, %xmm1
; SSE41-NEXT: retq
;
; AVX2-LABEL: and_mulhuw_v16i16:
@@ -417,13 +417,6 @@
define <16 x i16> @ashr_mulhuw_v16i16(<16 x i32> %a, <16 x i32> %b) {
; SSE2-LABEL: ashr_mulhuw_v16i16:
; SSE2: # %bb.0:
-; SSE2-NEXT: psrad $16, %xmm7
-; SSE2-NEXT: psrad $16, %xmm6
-; SSE2-NEXT: packssdw %xmm7, %xmm6
-; SSE2-NEXT: psrad $16, %xmm3
-; SSE2-NEXT: psrad $16, %xmm2
-; SSE2-NEXT: packssdw %xmm3, %xmm2
-; SSE2-NEXT: pmulhw %xmm6, %xmm2
; SSE2-NEXT: psrad $16, %xmm5
; SSE2-NEXT: psrad $16, %xmm4
; SSE2-NEXT: packssdw %xmm5, %xmm4
@@ -431,25 +424,32 @@
; SSE2-NEXT: psrad $16, %xmm0
; SSE2-NEXT: packssdw %xmm1, %xmm0
; SSE2-NEXT: pmulhw %xmm4, %xmm0
+; SSE2-NEXT: psrad $16, %xmm7
+; SSE2-NEXT: psrad $16, %xmm6
+; SSE2-NEXT: packssdw %xmm7, %xmm6
+; SSE2-NEXT: psrad $16, %xmm3
+; SSE2-NEXT: psrad $16, %xmm2
+; SSE2-NEXT: packssdw %xmm3, %xmm2
+; SSE2-NEXT: pmulhw %xmm6, %xmm2
; SSE2-NEXT: movdqa %xmm2, %xmm1
; SSE2-NEXT: retq
;
; SSE41-LABEL: ashr_mulhuw_v16i16:
; SSE41: # %bb.0:
-; SSE41-NEXT: psrld $16, %xmm1
-; SSE41-NEXT: psrld $16, %xmm0
-; SSE41-NEXT: packusdw %xmm1, %xmm0
; SSE41-NEXT: psrld $16, %xmm3
; SSE41-NEXT: psrld $16, %xmm2
; SSE41-NEXT: packusdw %xmm3, %xmm2
-; SSE41-NEXT: psrld $16, %xmm5
-; SSE41-NEXT: psrld $16, %xmm4
-; SSE41-NEXT: packusdw %xmm5, %xmm4
-; SSE41-NEXT: pmulhw %xmm4, %xmm0
+; SSE41-NEXT: psrld $16, %xmm1
+; SSE41-NEXT: psrld $16, %xmm0
+; SSE41-NEXT: packusdw %xmm1, %xmm0
; SSE41-NEXT: psrld $16, %xmm7
; SSE41-NEXT: psrld $16, %xmm6
; SSE41-NEXT: packusdw %xmm7, %xmm6
; SSE41-NEXT: pmulhw %xmm2, %xmm6
+; SSE41-NEXT: psrld $16, %xmm5
+; SSE41-NEXT: psrld $16, %xmm4
+; SSE41-NEXT: packusdw %xmm5, %xmm4
+; SSE41-NEXT: pmulhw %xmm4, %xmm0
; SSE41-NEXT: movdqa %xmm6, %xmm1
; SSE41-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/popcnt.ll b/llvm/test/CodeGen/X86/popcnt.ll
index 5ed14ab..ff3bdd2 100644
--- a/llvm/test/CodeGen/X86/popcnt.ll
+++ b/llvm/test/CodeGen/X86/popcnt.ll
@@ -62,7 +62,7 @@
define i16 @cnt16(i16 %x) nounwind readnone {
; X86-LABEL: cnt16:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movl %eax, %ecx
; X86-NEXT: shrl %ecx
; X86-NEXT: andl $21845, %ecx # imm = 0x5555
@@ -1523,7 +1523,7 @@
define i32 @popcount_i16_zext(i16 zeroext %x) {
; X86-LABEL: popcount_i16_zext:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movl %eax, %ecx
; X86-NEXT: shrl %ecx
; X86-NEXT: andl $21845, %ecx # imm = 0x5555
diff --git a/llvm/test/CodeGen/X86/promote-vec3.ll b/llvm/test/CodeGen/X86/promote-vec3.ll
index 482db62..984db8b 100644
--- a/llvm/test/CodeGen/X86/promote-vec3.ll
+++ b/llvm/test/CodeGen/X86/promote-vec3.ll
@@ -42,13 +42,13 @@
;
; AVX-64-LABEL: zext_i8:
; AVX-64: # %bb.0:
-; AVX-64-NEXT: movzbl %dl, %ecx
-; AVX-64-NEXT: movzbl %sil, %edx
+; AVX-64-NEXT: movzbl %sil, %esi
; AVX-64-NEXT: vmovd %edi, %xmm0
; AVX-64-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX-64-NEXT: movzbl %dl, %ecx
; AVX-64-NEXT: vmovd %xmm0, %eax
; AVX-64-NEXT: # kill: def $ax killed $ax killed $eax
-; AVX-64-NEXT: # kill: def $dx killed $dx killed $edx
+; AVX-64-NEXT: movl %esi, %edx
; AVX-64-NEXT: # kill: def $cx killed $cx killed $ecx
; AVX-64-NEXT: retq
%2 = zext <3 x i8> %0 to <3 x i16>
diff --git a/llvm/test/CodeGen/X86/psubus.ll b/llvm/test/CodeGen/X86/psubus.ll
index 28d4a88..4fa5ef2 100644
--- a/llvm/test/CodeGen/X86/psubus.ll
+++ b/llvm/test/CodeGen/X86/psubus.ll
@@ -1806,58 +1806,58 @@
; SSE2OR3-LABEL: psubus_16i32_max:
; SSE2OR3: # %bb.0: # %vector.ph
; SSE2OR3-NEXT: movdqa {{.*#+}} xmm7 = [2147483648,2147483648,2147483648,2147483648]
-; SSE2OR3-NEXT: movdqa %xmm5, %xmm8
+; SSE2OR3-NEXT: movdqa %xmm3, %xmm8
; SSE2OR3-NEXT: pxor %xmm7, %xmm8
; SSE2OR3-NEXT: movdqa {{.*#+}} xmm6 = [2147549183,2147549183,2147549183,2147549183]
; SSE2OR3-NEXT: movdqa %xmm6, %xmm9
; SSE2OR3-NEXT: pcmpgtd %xmm8, %xmm9
; SSE2OR3-NEXT: pcmpeqd %xmm8, %xmm8
-; SSE2OR3-NEXT: pand %xmm9, %xmm5
-; SSE2OR3-NEXT: pxor %xmm8, %xmm9
-; SSE2OR3-NEXT: por %xmm5, %xmm9
-; SSE2OR3-NEXT: pslld $16, %xmm9
-; SSE2OR3-NEXT: psrad $16, %xmm9
-; SSE2OR3-NEXT: movdqa %xmm4, %xmm10
-; SSE2OR3-NEXT: pxor %xmm7, %xmm10
-; SSE2OR3-NEXT: movdqa %xmm6, %xmm5
-; SSE2OR3-NEXT: pcmpgtd %xmm10, %xmm5
-; SSE2OR3-NEXT: pand %xmm5, %xmm4
-; SSE2OR3-NEXT: pxor %xmm8, %xmm5
-; SSE2OR3-NEXT: por %xmm4, %xmm5
-; SSE2OR3-NEXT: pslld $16, %xmm5
-; SSE2OR3-NEXT: psrad $16, %xmm5
-; SSE2OR3-NEXT: packssdw %xmm9, %xmm5
-; SSE2OR3-NEXT: movdqa %xmm3, %xmm4
-; SSE2OR3-NEXT: pxor %xmm7, %xmm4
-; SSE2OR3-NEXT: movdqa %xmm6, %xmm9
-; SSE2OR3-NEXT: pcmpgtd %xmm4, %xmm9
; SSE2OR3-NEXT: pand %xmm9, %xmm3
; SSE2OR3-NEXT: pxor %xmm8, %xmm9
; SSE2OR3-NEXT: por %xmm3, %xmm9
; SSE2OR3-NEXT: pslld $16, %xmm9
; SSE2OR3-NEXT: psrad $16, %xmm9
-; SSE2OR3-NEXT: pxor %xmm2, %xmm7
+; SSE2OR3-NEXT: movdqa %xmm2, %xmm3
+; SSE2OR3-NEXT: pxor %xmm7, %xmm3
+; SSE2OR3-NEXT: movdqa %xmm6, %xmm10
+; SSE2OR3-NEXT: pcmpgtd %xmm3, %xmm10
+; SSE2OR3-NEXT: pand %xmm10, %xmm2
+; SSE2OR3-NEXT: pxor %xmm8, %xmm10
+; SSE2OR3-NEXT: por %xmm2, %xmm10
+; SSE2OR3-NEXT: pslld $16, %xmm10
+; SSE2OR3-NEXT: psrad $16, %xmm10
+; SSE2OR3-NEXT: packssdw %xmm9, %xmm10
+; SSE2OR3-NEXT: psubusw %xmm10, %xmm0
+; SSE2OR3-NEXT: movdqa %xmm5, %xmm2
+; SSE2OR3-NEXT: pxor %xmm7, %xmm2
+; SSE2OR3-NEXT: movdqa %xmm6, %xmm3
+; SSE2OR3-NEXT: pcmpgtd %xmm2, %xmm3
+; SSE2OR3-NEXT: pand %xmm3, %xmm5
+; SSE2OR3-NEXT: pxor %xmm8, %xmm3
+; SSE2OR3-NEXT: por %xmm5, %xmm3
+; SSE2OR3-NEXT: pslld $16, %xmm3
+; SSE2OR3-NEXT: psrad $16, %xmm3
+; SSE2OR3-NEXT: pxor %xmm4, %xmm7
; SSE2OR3-NEXT: pcmpgtd %xmm7, %xmm6
; SSE2OR3-NEXT: pxor %xmm6, %xmm8
-; SSE2OR3-NEXT: pand %xmm2, %xmm6
+; SSE2OR3-NEXT: pand %xmm4, %xmm6
; SSE2OR3-NEXT: por %xmm8, %xmm6
; SSE2OR3-NEXT: pslld $16, %xmm6
; SSE2OR3-NEXT: psrad $16, %xmm6
-; SSE2OR3-NEXT: packssdw %xmm9, %xmm6
-; SSE2OR3-NEXT: psubusw %xmm6, %xmm0
-; SSE2OR3-NEXT: psubusw %xmm5, %xmm1
+; SSE2OR3-NEXT: packssdw %xmm3, %xmm6
+; SSE2OR3-NEXT: psubusw %xmm6, %xmm1
; SSE2OR3-NEXT: retq
;
; SSE41-LABEL: psubus_16i32_max:
; SSE41: # %bb.0: # %vector.ph
; SSE41-NEXT: movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535]
-; SSE41-NEXT: pminud %xmm6, %xmm5
-; SSE41-NEXT: pminud %xmm6, %xmm4
-; SSE41-NEXT: packusdw %xmm5, %xmm4
; SSE41-NEXT: pminud %xmm6, %xmm3
; SSE41-NEXT: pminud %xmm6, %xmm2
; SSE41-NEXT: packusdw %xmm3, %xmm2
; SSE41-NEXT: psubusw %xmm2, %xmm0
+; SSE41-NEXT: pminud %xmm6, %xmm5
+; SSE41-NEXT: pminud %xmm6, %xmm4
+; SSE41-NEXT: packusdw %xmm5, %xmm4
; SSE41-NEXT: psubusw %xmm4, %xmm1
; SSE41-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/shift-mask.ll b/llvm/test/CodeGen/X86/shift-mask.ll
index 604f5c1..fe52c7b 100644
--- a/llvm/test/CodeGen/X86/shift-mask.ll
+++ b/llvm/test/CodeGen/X86/shift-mask.ll
@@ -142,9 +142,9 @@
define i16 @test_i16_shl_lshr_2(i16 %a0) {
; X86-LABEL: test_i16_shl_lshr_2:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: shrl $2, %eax
-; X86-NEXT: andl $-8, %eax
+; X86-NEXT: andl $16376, %eax # imm = 0x3FF8
; X86-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NEXT: retl
;
@@ -411,7 +411,7 @@
define i16 @test_i16_lshr_lshr_1(i16 %a0) {
; X86-LABEL: test_i16_lshr_lshr_1:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: shrl $2, %eax
; X86-NEXT: andl $2047, %eax # imm = 0x7FF
; X86-NEXT: # kill: def $ax killed $ax killed $eax
diff --git a/llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll b/llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll
index f5b0aa5..cfd9010 100644
--- a/llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll
+++ b/llvm/test/CodeGen/X86/shuffle-strided-with-offset-128.ll
@@ -463,8 +463,8 @@
; SSE2-LABEL: shuffle_v16i8_to_v2i8_2:
; SSE2: # %bb.0:
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = mem[0,2,2,3]
-; SSE2-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
+; SSE2-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: movw %ax, (%rsi)
@@ -541,8 +541,8 @@
; SSE2-LABEL: shuffle_v16i8_to_v2i8_4:
; SSE2: # %bb.0:
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = mem[3,1,2,3]
-; SSE2-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
+; SSE2-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: movw %ax, (%rsi)
@@ -619,8 +619,8 @@
; SSE2-LABEL: shuffle_v16i8_to_v2i8_6:
; SSE2: # %bb.0:
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = mem[3,1,2,3]
-; SSE2-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE2-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: movw %ax, (%rsi)
diff --git a/llvm/test/CodeGen/X86/single_elt_vector_memory_operation.ll b/llvm/test/CodeGen/X86/single_elt_vector_memory_operation.ll
index f65461c..1d8450e 100644
--- a/llvm/test/CodeGen/X86/single_elt_vector_memory_operation.ll
+++ b/llvm/test/CodeGen/X86/single_elt_vector_memory_operation.ll
@@ -56,20 +56,23 @@
;
; AVX-LABEL: store_single_128bit_elt_vector:
; AVX: # %bb.0:
-; AVX-NEXT: vmovaps (%rdi), %xmm0
+; AVX-NEXT: vmovaps (%rdi), %ymm0
; AVX-NEXT: vmovaps %xmm0, (%rdx)
+; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX2-LABEL: store_single_128bit_elt_vector:
; AVX2: # %bb.0:
-; AVX2-NEXT: vmovaps (%rdi), %xmm0
+; AVX2-NEXT: vmovaps (%rdi), %ymm0
; AVX2-NEXT: vmovaps %xmm0, (%rdx)
+; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: store_single_128bit_elt_vector:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovaps (%rdi), %xmm0
+; AVX512F-NEXT: vmovaps (%rdi), %ymm0
; AVX512F-NEXT: vmovaps %xmm0, (%rdx)
+; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
%i0 = load <32 x i8>, ptr %in, align 64
%i1 = bitcast <32 x i8> %i0 to <2 x i128>
@@ -149,7 +152,7 @@
;
; AVX512F-LABEL: store_single_256bit_elt_vector:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovaps (%rdi), %ymm0
+; AVX512F-NEXT: vmovaps (%rdi), %zmm0
; AVX512F-NEXT: vmovaps %ymm0, (%rdx)
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/smax.ll b/llvm/test/CodeGen/X86/smax.ll
index 838ef46..11d21cf 100644
--- a/llvm/test/CodeGen/X86/smax.ll
+++ b/llvm/test/CodeGen/X86/smax.ll
@@ -660,8 +660,8 @@
; X64: # %bb.0:
; X64-NEXT: movswl %si, %eax
; X64-NEXT: movswl %di, %ecx
-; X64-NEXT: shrl $15, %ecx
-; X64-NEXT: shrl $8, %eax
+; X64-NEXT: sarl $15, %ecx
+; X64-NEXT: sarl $8, %eax
; X64-NEXT: cmpw %ax, %cx
; X64-NEXT: cmovgl %ecx, %eax
; X64-NEXT: # kill: def $ax killed $ax killed $eax
@@ -671,7 +671,7 @@
; X86: # %bb.0:
; X86-NEXT: movsbl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movswl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: shrl $15, %eax
+; X86-NEXT: sarl $15, %eax
; X86-NEXT: cmpw %cx, %ax
; X86-NEXT: cmovlel %ecx, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
diff --git a/llvm/test/CodeGen/X86/smin.ll b/llvm/test/CodeGen/X86/smin.ll
index 9436676..f353853 100644
--- a/llvm/test/CodeGen/X86/smin.ll
+++ b/llvm/test/CodeGen/X86/smin.ll
@@ -659,8 +659,8 @@
; X64: # %bb.0:
; X64-NEXT: movswl %si, %eax
; X64-NEXT: movswl %di, %ecx
-; X64-NEXT: shrl $15, %ecx
-; X64-NEXT: shrl $8, %eax
+; X64-NEXT: sarl $15, %ecx
+; X64-NEXT: sarl $8, %eax
; X64-NEXT: cmpw %ax, %cx
; X64-NEXT: cmovll %ecx, %eax
; X64-NEXT: # kill: def $ax killed $ax killed $eax
@@ -670,7 +670,7 @@
; X86: # %bb.0:
; X86-NEXT: movsbl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movswl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: shrl $15, %eax
+; X86-NEXT: sarl $15, %eax
; X86-NEXT: cmpw %cx, %ax
; X86-NEXT: cmovgel %ecx, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
diff --git a/llvm/test/CodeGen/X86/smulo-128-legalisation-lowering.ll b/llvm/test/CodeGen/X86/smulo-128-legalisation-lowering.ll
index 3c4bb04..4fe4dcc 100644
--- a/llvm/test/CodeGen/X86/smulo-128-legalisation-lowering.ll
+++ b/llvm/test/CodeGen/X86/smulo-128-legalisation-lowering.ll
@@ -89,8 +89,8 @@
; X86-NEXT: .cfi_offset %edi, -16
; X86-NEXT: .cfi_offset %ebx, -12
; X86-NEXT: .cfi_offset %ebp, -8
-; X86-NEXT: movl {{[0-9]+}}(%esp), %esi
; X86-NEXT: movl {{[0-9]+}}(%esp), %edi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %esi
; X86-NEXT: movl {{[0-9]+}}(%esp), %ebx
; X86-NEXT: movl %edi, %eax
; X86-NEXT: mull %ebx
@@ -251,10 +251,10 @@
; X86-NEXT: addl %eax, %esi
; X86-NEXT: adcl %edx, %ecx
; X86-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) ## 4-byte Spill
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: sarl $31, %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: sarl $31, %eax
-; X86-NEXT: movl %eax, %ecx
-; X86-NEXT: mull {{[0-9]+}}(%esp)
+; X86-NEXT: mull %ecx
; X86-NEXT: movl %edx, %edi
; X86-NEXT: movl %eax, %ebx
; X86-NEXT: movl %eax, %ebp
@@ -585,8 +585,8 @@
; X86-NEXT: .cfi_offset %edi, -16
; X86-NEXT: .cfi_offset %ebx, -12
; X86-NEXT: .cfi_offset %ebp, -8
-; X86-NEXT: movl {{[0-9]+}}(%esp), %edi
; X86-NEXT: movl {{[0-9]+}}(%esp), %ebp
+; X86-NEXT: movl {{[0-9]+}}(%esp), %edi
; X86-NEXT: movl {{[0-9]+}}(%esp), %ebx
; X86-NEXT: movl %ebp, %eax
; X86-NEXT: mull %ebx
@@ -1295,8 +1295,8 @@
; X86-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) ## 4-byte Spill
; X86-NEXT: addl %edx, %ecx
; X86-NEXT: adcl $0, %esi
-; X86-NEXT: movl %edi, %eax
-; X86-NEXT: mull {{[0-9]+}}(%esp)
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: mull %edi
; X86-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) ## 4-byte Spill
; X86-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) ## 4-byte Spill
; X86-NEXT: addl %eax, %ecx
@@ -1315,9 +1315,9 @@
; X86-NEXT: adcl $0, %esi
; X86-NEXT: adcl $0, %ebx
; X86-NEXT: movl %ebx, {{[-0-9]+}}(%e{{[sb]}}p) ## 4-byte Spill
-; X86-NEXT: movl %edi, %eax
; X86-NEXT: movl %edi, %ecx
; X86-NEXT: movl %edi, {{[-0-9]+}}(%e{{[sb]}}p) ## 4-byte Spill
+; X86-NEXT: movl %edi, %eax
; X86-NEXT: mull {{[0-9]+}}(%esp)
; X86-NEXT: movl %edx, %ebp
; X86-NEXT: movl %eax, %ebx
@@ -1379,9 +1379,9 @@
; X86-NEXT: movl %ebx, {{[-0-9]+}}(%e{{[sb]}}p) ## 4-byte Spill
; X86-NEXT: adcl %ebp, %eax
; X86-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) ## 4-byte Spill
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ebp ## 4-byte Reload
-; X86-NEXT: mull %ebp
+; X86-NEXT: movl %ebp, %eax
+; X86-NEXT: mull {{[0-9]+}}(%esp)
; X86-NEXT: movl %edx, %esi
; X86-NEXT: movl %eax, %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
diff --git a/llvm/test/CodeGen/X86/umax.ll b/llvm/test/CodeGen/X86/umax.ll
index f86f8ea..2c93f6a 100644
--- a/llvm/test/CodeGen/X86/umax.ll
+++ b/llvm/test/CodeGen/X86/umax.ll
@@ -1259,8 +1259,8 @@
; X64: # %bb.0:
; X64-NEXT: movswl %si, %eax
; X64-NEXT: movswl %di, %ecx
-; X64-NEXT: shrl $15, %ecx
-; X64-NEXT: shrl $8, %eax
+; X64-NEXT: sarl $15, %ecx
+; X64-NEXT: sarl $8, %eax
; X64-NEXT: cmpw %ax, %cx
; X64-NEXT: cmoval %ecx, %eax
; X64-NEXT: # kill: def $ax killed $ax killed $eax
@@ -1270,7 +1270,7 @@
; X86: # %bb.0:
; X86-NEXT: movsbl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movswl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: shrl $15, %eax
+; X86-NEXT: sarl $15, %eax
; X86-NEXT: cmpw %cx, %ax
; X86-NEXT: cmovbel %ecx, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
diff --git a/llvm/test/CodeGen/X86/umin.ll b/llvm/test/CodeGen/X86/umin.ll
index 29d3e0d..0a747b8 100644
--- a/llvm/test/CodeGen/X86/umin.ll
+++ b/llvm/test/CodeGen/X86/umin.ll
@@ -668,8 +668,8 @@
; X64: # %bb.0:
; X64-NEXT: movswl %si, %eax
; X64-NEXT: movswl %di, %ecx
-; X64-NEXT: shrl $15, %ecx
-; X64-NEXT: shrl $8, %eax
+; X64-NEXT: sarl $15, %ecx
+; X64-NEXT: sarl $8, %eax
; X64-NEXT: cmpw %ax, %cx
; X64-NEXT: cmovbl %ecx, %eax
; X64-NEXT: # kill: def $ax killed $ax killed $eax
@@ -679,7 +679,7 @@
; X86: # %bb.0:
; X86-NEXT: movsbl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movswl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: shrl $15, %eax
+; X86-NEXT: sarl $15, %eax
; X86-NEXT: cmpw %cx, %ax
; X86-NEXT: cmovael %ecx, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
diff --git a/llvm/test/CodeGen/X86/vector-fshl-256.ll b/llvm/test/CodeGen/X86/vector-fshl-256.ll
index 1720193..bdbc7bf 100644
--- a/llvm/test/CodeGen/X86/vector-fshl-256.ll
+++ b/llvm/test/CodeGen/X86/vector-fshl-256.ll
@@ -1001,7 +1001,7 @@
;
; AVX2-LABEL: splatvar_funnnel_v16i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX2-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX2-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
@@ -1012,7 +1012,7 @@
;
; AVX512F-LABEL: splatvar_funnnel_v16i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512F-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512F-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
@@ -1023,7 +1023,7 @@
;
; AVX512VL-LABEL: splatvar_funnnel_v16i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512VL-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VL-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
@@ -1034,7 +1034,7 @@
;
; AVX512BW-LABEL: splatvar_funnnel_v16i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512BW-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512BW-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
@@ -1054,7 +1054,7 @@
;
; AVX512VLBW-LABEL: splatvar_funnnel_v16i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512VLBW-NEXT: vpsrlw $1, %ymm1, %ymm1
; AVX512VLBW-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
@@ -1089,7 +1089,7 @@
;
; XOPAVX2-LABEL: splatvar_funnnel_v16i16:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm4
; XOPAVX2-NEXT: vpsrlw $1, %ymm1, %ymm1
; XOPAVX2-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
diff --git a/llvm/test/CodeGen/X86/vector-fshl-512.ll b/llvm/test/CodeGen/X86/vector-fshl-512.ll
index 114d706..8cb042c 100644
--- a/llvm/test/CodeGen/X86/vector-fshl-512.ll
+++ b/llvm/test/CodeGen/X86/vector-fshl-512.ll
@@ -580,7 +580,7 @@
;
; AVX512BW-LABEL: splatvar_funnnel_v32i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512BW-NEXT: vpsrlw $1, %zmm1, %zmm1
; AVX512BW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
@@ -597,7 +597,7 @@
;
; AVX512VLBW-LABEL: splatvar_funnnel_v32i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm4
; AVX512VLBW-NEXT: vpsrlw $1, %zmm1, %zmm1
; AVX512VLBW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
diff --git a/llvm/test/CodeGen/X86/vector-fshl-rot-256.ll b/llvm/test/CodeGen/X86/vector-fshl-rot-256.ll
index b292297..d03b0c2 100644
--- a/llvm/test/CodeGen/X86/vector-fshl-rot-256.ll
+++ b/llvm/test/CodeGen/X86/vector-fshl-rot-256.ll
@@ -775,7 +775,7 @@
;
; AVX2-LABEL: splatvar_funnnel_v16i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX2-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX2-NEXT: vpsrlw $1, %ymm0, %ymm4
; AVX2-NEXT: vpsrlw %xmm3, %ymm4, %ymm3
@@ -786,7 +786,7 @@
;
; AVX512F-LABEL: splatvar_funnnel_v16i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512F-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512F-NEXT: vpsrlw $1, %ymm0, %ymm4
; AVX512F-NEXT: vpsrlw %xmm3, %ymm4, %ymm3
@@ -797,7 +797,7 @@
;
; AVX512VL-LABEL: splatvar_funnnel_v16i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512VL-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512VL-NEXT: vpsrlw $1, %ymm0, %ymm4
; AVX512VL-NEXT: vpsrlw %xmm3, %ymm4, %ymm3
@@ -808,7 +808,7 @@
;
; AVX512BW-LABEL: splatvar_funnnel_v16i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512BW-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512BW-NEXT: vpsrlw $1, %ymm0, %ymm4
; AVX512BW-NEXT: vpsrlw %xmm3, %ymm4, %ymm3
@@ -819,7 +819,7 @@
;
; AVX512VLBW-LABEL: splatvar_funnnel_v16i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512VLBW-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512VLBW-NEXT: vpsrlw $1, %ymm0, %ymm4
; AVX512VLBW-NEXT: vpsrlw %xmm3, %ymm4, %ymm3
diff --git a/llvm/test/CodeGen/X86/vector-fshl-rot-512.ll b/llvm/test/CodeGen/X86/vector-fshl-rot-512.ll
index 0a473dd..81874d4 100644
--- a/llvm/test/CodeGen/X86/vector-fshl-rot-512.ll
+++ b/llvm/test/CodeGen/X86/vector-fshl-rot-512.ll
@@ -334,7 +334,7 @@
;
; AVX512BW-LABEL: splatvar_funnnel_v32i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512BW-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512BW-NEXT: vpsrlw $1, %zmm0, %zmm4
; AVX512BW-NEXT: vpsrlw %xmm3, %zmm4, %zmm3
@@ -345,7 +345,7 @@
;
; AVX512VLBW-LABEL: splatvar_funnnel_v32i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512VLBW-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512VLBW-NEXT: vpsrlw $1, %zmm0, %zmm4
; AVX512VLBW-NEXT: vpsrlw %xmm3, %zmm4, %zmm3
diff --git a/llvm/test/CodeGen/X86/vector-fshr-256.ll b/llvm/test/CodeGen/X86/vector-fshr-256.ll
index b39b7c1..053b3bb 100644
--- a/llvm/test/CodeGen/X86/vector-fshr-256.ll
+++ b/llvm/test/CodeGen/X86/vector-fshr-256.ll
@@ -1034,7 +1034,7 @@
;
; AVX2-LABEL: splatvar_funnnel_v16i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
; AVX2-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
; AVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
@@ -1045,7 +1045,7 @@
;
; AVX512F-LABEL: splatvar_funnnel_v16i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512F-NEXT: vpand %xmm3, %xmm2, %xmm4
; AVX512F-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
; AVX512F-NEXT: vpandn %xmm3, %xmm2, %xmm2
@@ -1056,7 +1056,7 @@
;
; AVX512VL-LABEL: splatvar_funnnel_v16i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512VL-NEXT: vpand %xmm3, %xmm2, %xmm4
; AVX512VL-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
; AVX512VL-NEXT: vpandn %xmm3, %xmm2, %xmm2
@@ -1067,7 +1067,7 @@
;
; AVX512BW-LABEL: splatvar_funnnel_v16i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
; AVX512BW-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
@@ -1087,7 +1087,7 @@
;
; AVX512VLBW-LABEL: splatvar_funnnel_v16i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4
; AVX512VLBW-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
@@ -1123,7 +1123,7 @@
;
; XOPAVX2-LABEL: splatvar_funnnel_v16i16:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; XOPAVX2-NEXT: vpand %xmm3, %xmm2, %xmm4
; XOPAVX2-NEXT: vpsrlw %xmm4, %ymm1, %ymm1
; XOPAVX2-NEXT: vpandn %xmm3, %xmm2, %xmm2
diff --git a/llvm/test/CodeGen/X86/vector-fshr-512.ll b/llvm/test/CodeGen/X86/vector-fshr-512.ll
index 2bd0350..40f15af 100644
--- a/llvm/test/CodeGen/X86/vector-fshr-512.ll
+++ b/llvm/test/CodeGen/X86/vector-fshr-512.ll
@@ -582,7 +582,7 @@
;
; AVX512BW-LABEL: splatvar_funnnel_v32i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512BW-NEXT: vpand %xmm3, %xmm2, %xmm4
; AVX512BW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
; AVX512BW-NEXT: vpandn %xmm3, %xmm2, %xmm2
@@ -600,7 +600,7 @@
;
; AVX512VLBW-LABEL: splatvar_funnnel_v32i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq {{.*#+}} xmm3 = [15,0,0,0,15,0,0,0]
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm3 = [15,0,0,0]
; AVX512VLBW-NEXT: vpand %xmm3, %xmm2, %xmm4
; AVX512VLBW-NEXT: vpsrlw %xmm4, %zmm1, %zmm1
; AVX512VLBW-NEXT: vpandn %xmm3, %xmm2, %xmm2
diff --git a/llvm/test/CodeGen/X86/vector-fshr-rot-256.ll b/llvm/test/CodeGen/X86/vector-fshr-rot-256.ll
index ecc832f..8c196c5 100644
--- a/llvm/test/CodeGen/X86/vector-fshr-rot-256.ll
+++ b/llvm/test/CodeGen/X86/vector-fshr-rot-256.ll
@@ -814,7 +814,7 @@
;
; AVX2-LABEL: splatvar_funnnel_v16i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX2-NEXT: vpand %xmm2, %xmm1, %xmm3
; AVX2-NEXT: vpsrlw %xmm3, %ymm0, %ymm3
; AVX2-NEXT: vpandn %xmm2, %xmm1, %xmm1
@@ -825,7 +825,7 @@
;
; AVX512F-LABEL: splatvar_funnnel_v16i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512F-NEXT: vpand %xmm2, %xmm1, %xmm3
; AVX512F-NEXT: vpsrlw %xmm3, %ymm0, %ymm3
; AVX512F-NEXT: vpandn %xmm2, %xmm1, %xmm1
@@ -836,7 +836,7 @@
;
; AVX512VL-LABEL: splatvar_funnnel_v16i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512VL-NEXT: vpand %xmm2, %xmm1, %xmm3
; AVX512VL-NEXT: vpsrlw %xmm3, %ymm0, %ymm3
; AVX512VL-NEXT: vpandn %xmm2, %xmm1, %xmm1
@@ -847,7 +847,7 @@
;
; AVX512BW-LABEL: splatvar_funnnel_v16i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512BW-NEXT: vpand %xmm2, %xmm1, %xmm3
; AVX512BW-NEXT: vpsrlw %xmm3, %ymm0, %ymm3
; AVX512BW-NEXT: vpandn %xmm2, %xmm1, %xmm1
@@ -858,7 +858,7 @@
;
; AVX512VLBW-LABEL: splatvar_funnnel_v16i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512VLBW-NEXT: vpand %xmm2, %xmm1, %xmm3
; AVX512VLBW-NEXT: vpsrlw %xmm3, %ymm0, %ymm3
; AVX512VLBW-NEXT: vpandn %xmm2, %xmm1, %xmm1
diff --git a/llvm/test/CodeGen/X86/vector-fshr-rot-512.ll b/llvm/test/CodeGen/X86/vector-fshr-rot-512.ll
index 3c17bf2..1f0cf8b 100644
--- a/llvm/test/CodeGen/X86/vector-fshr-rot-512.ll
+++ b/llvm/test/CodeGen/X86/vector-fshr-rot-512.ll
@@ -332,7 +332,7 @@
;
; AVX512BW-LABEL: splatvar_funnnel_v32i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512BW-NEXT: vpand %xmm2, %xmm1, %xmm3
; AVX512BW-NEXT: vpsrlw %xmm3, %zmm0, %zmm3
; AVX512BW-NEXT: vpandn %xmm2, %xmm1, %xmm1
@@ -343,7 +343,7 @@
;
; AVX512VLBW-LABEL: splatvar_funnnel_v32i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512VLBW-NEXT: vpand %xmm2, %xmm1, %xmm3
; AVX512VLBW-NEXT: vpsrlw %xmm3, %zmm0, %zmm3
; AVX512VLBW-NEXT: vpandn %xmm2, %xmm1, %xmm1
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-2.ll
index 3ce186d..0432c09 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-2.ll
@@ -262,39 +262,39 @@
define void @load_i16_stride2_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nounwind {
; SSE-LABEL: load_i16_stride2_vf32:
; SSE: # %bb.0:
-; SSE-NEXT: movdqa 64(%rdi), %xmm0
-; SSE-NEXT: movdqa 80(%rdi), %xmm4
+; SSE-NEXT: movdqa (%rdi), %xmm0
+; SSE-NEXT: movdqa 16(%rdi), %xmm5
+; SSE-NEXT: movdqa 32(%rdi), %xmm2
+; SSE-NEXT: movdqa 48(%rdi), %xmm4
; SSE-NEXT: movdqa 96(%rdi), %xmm1
-; SSE-NEXT: movdqa 112(%rdi), %xmm6
-; SSE-NEXT: movdqa (%rdi), %xmm2
-; SSE-NEXT: movdqa 16(%rdi), %xmm7
-; SSE-NEXT: movdqa 32(%rdi), %xmm3
-; SSE-NEXT: movdqa 48(%rdi), %xmm9
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm9[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm5[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm3[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm8[0]
+; SSE-NEXT: movdqa 112(%rdi), %xmm7
+; SSE-NEXT: movdqa 64(%rdi), %xmm3
+; SSE-NEXT: movdqa 80(%rdi), %xmm9
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm9[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm6[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm3[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm8[0]
; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm7[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm8[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm2[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm1[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
; SSE-NEXT: punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm10[0]
-; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm6[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm5[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm10[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm1[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
; SSE-NEXT: punpcklqdq {{.*#+}} xmm10 = xmm10[0],xmm11[0]
; SSE-NEXT: pshuflw {{.*#+}} xmm11 = xmm4[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm11[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm0[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm2[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm12[0,2,2,3]
; SSE-NEXT: punpcklqdq {{.*#+}} xmm12 = xmm12[0],xmm11[0]
@@ -302,22 +302,22 @@
; SSE-NEXT: psrad $16, %xmm3
; SSE-NEXT: packssdw %xmm9, %xmm3
; SSE-NEXT: psrad $16, %xmm7
-; SSE-NEXT: psrad $16, %xmm2
-; SSE-NEXT: packssdw %xmm7, %xmm2
-; SSE-NEXT: psrad $16, %xmm6
; SSE-NEXT: psrad $16, %xmm1
-; SSE-NEXT: packssdw %xmm6, %xmm1
-; SSE-NEXT: psrad $16, %xmm4
+; SSE-NEXT: packssdw %xmm7, %xmm1
+; SSE-NEXT: psrad $16, %xmm5
; SSE-NEXT: psrad $16, %xmm0
-; SSE-NEXT: packssdw %xmm4, %xmm0
-; SSE-NEXT: movdqa %xmm12, 32(%rsi)
-; SSE-NEXT: movdqa %xmm10, 48(%rsi)
-; SSE-NEXT: movdqa %xmm8, (%rsi)
-; SSE-NEXT: movdqa %xmm5, 16(%rsi)
-; SSE-NEXT: movdqa %xmm0, 32(%rdx)
+; SSE-NEXT: packssdw %xmm5, %xmm0
+; SSE-NEXT: psrad $16, %xmm4
+; SSE-NEXT: psrad $16, %xmm2
+; SSE-NEXT: packssdw %xmm4, %xmm2
+; SSE-NEXT: movdqa %xmm12, 16(%rsi)
+; SSE-NEXT: movdqa %xmm10, (%rsi)
+; SSE-NEXT: movdqa %xmm8, 48(%rsi)
+; SSE-NEXT: movdqa %xmm6, 32(%rsi)
+; SSE-NEXT: movdqa %xmm2, 16(%rdx)
+; SSE-NEXT: movdqa %xmm0, (%rdx)
; SSE-NEXT: movdqa %xmm1, 48(%rdx)
-; SSE-NEXT: movdqa %xmm2, (%rdx)
-; SSE-NEXT: movdqa %xmm3, 16(%rdx)
+; SSE-NEXT: movdqa %xmm3, 32(%rdx)
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i16_stride2_vf32:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
index 1967248..a2a3f9e 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
@@ -1246,27 +1246,44 @@
define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
; SSE-LABEL: load_i16_stride3_vf64:
; SSE: # %bb.0:
-; SSE-NEXT: subq $408, %rsp # imm = 0x198
-; SSE-NEXT: movdqa 192(%rdi), %xmm11
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 272(%rdi), %xmm4
+; SSE-NEXT: subq $392, %rsp # imm = 0x188
+; SSE-NEXT: movdqa 336(%rdi), %xmm14
+; SSE-NEXT: movdqa 176(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 240(%rdi), %xmm13
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 256(%rdi), %xmm5
-; SSE-NEXT: movdqa 80(%rdi), %xmm9
+; SSE-NEXT: movdqa 144(%rdi), %xmm8
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 160(%rdi), %xmm5
+; SSE-NEXT: movdqa 320(%rdi), %xmm9
; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa (%rdi), %xmm12
-; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 16(%rdi), %xmm15
-; SSE-NEXT: movdqa 32(%rdi), %xmm7
+; SSE-NEXT: movdqa 288(%rdi), %xmm11
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 304(%rdi), %xmm10
+; SSE-NEXT: movdqa 128(%rdi), %xmm6
+; SSE-NEXT: movdqa 96(%rdi), %xmm12
+; SSE-NEXT: movdqa 112(%rdi), %xmm7
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 48(%rdi), %xmm6
-; SSE-NEXT: movdqa 64(%rdi), %xmm10
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,0,65535,65535,0]
; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm7, %xmm1
+; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,2,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm1[0,3,2,1,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm7[0,1,2,3,4,7,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[0,1,2,1]
+; SSE-NEXT: movdqa %xmm6, %xmm15
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,0],xmm1[2,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pandn %xmm10, %xmm1
-; SSE-NEXT: movdqa %xmm6, %xmm2
+; SSE-NEXT: movdqa %xmm11, %xmm2
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,2,1,3]
@@ -1274,7 +1291,6 @@
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm1[0,3,2,1,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm10[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm9[0,1,2,1]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,0],xmm1[2,0]
@@ -1282,32 +1298,16 @@
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: pandn %xmm5, %xmm1
-; SSE-NEXT: movdqa %xmm13, %xmm2
+; SSE-NEXT: movdqa %xmm8, %xmm2
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,2,1,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm1[0,3,2,1,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm5[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa %xmm5, %xmm9
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm4[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,0],xmm1[2,0]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm15, %xmm1
-; SSE-NEXT: movdqa %xmm12, %xmm2
-; SSE-NEXT: pand %xmm0, %xmm2
-; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm15[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[0,1,2,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[0,1,2,1]
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,0],xmm1[2,0]
-; SSE-NEXT: movdqa 208(%rdi), %xmm4
+; SSE-NEXT: movdqa 352(%rdi), %xmm4
+; SSE-NEXT: movdqa %xmm4, (%rsp) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,2,1,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
@@ -1316,7 +1316,48 @@
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: pandn %xmm4, %xmm1
-; SSE-NEXT: movdqa %xmm11, %xmm2
+; SSE-NEXT: movdqa %xmm14, %xmm2
+; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movdqa 368(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,0],xmm3[2,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 16(%rdi), %xmm6
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm6, %xmm1
+; SSE-NEXT: movdqa (%rdi), %xmm13
+; SSE-NEXT: movdqa %xmm13, %xmm2
+; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movdqa 32(%rdi), %xmm11
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,1,2,1]
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,4,7,6,7]
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,0],xmm3[2,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 208(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm2, %xmm1
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 192(%rdi), %xmm10
+; SSE-NEXT: movdqa %xmm10, %xmm2
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: movdqa 224(%rdi), %xmm1
@@ -1324,7 +1365,6 @@
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,0],xmm3[2,0]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
@@ -1332,40 +1372,21 @@
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 160(%rdi), %xmm5
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm5, %xmm1
-; SSE-NEXT: movdqa 144(%rdi), %xmm14
-; SSE-NEXT: movdqa %xmm14, %xmm2
-; SSE-NEXT: pand %xmm0, %xmm2
-; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: movdqa 176(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,0],xmm3[2,0]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 352(%rdi), %xmm2
+; SSE-NEXT: movdqa 64(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: movdqa %xmm2, %xmm7
+; SSE-NEXT: movdqa %xmm2, %xmm4
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 336(%rdi), %xmm12
-; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: movdqa 48(%rdi), %xmm9
+; SSE-NEXT: movdqa %xmm9, %xmm2
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: movdqa 368(%rdi), %xmm11
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,1,2,1]
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 80(%rdi), %xmm8
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[0,1,2,1]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm7[0,1,2,3,4,7,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,7,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,0],xmm3[2,0]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
@@ -1373,35 +1394,14 @@
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: movdqa %xmm2, %xmm3
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 96(%rdi), %xmm8
-; SSE-NEXT: movdqa %xmm8, %xmm2
-; SSE-NEXT: pand %xmm0, %xmm2
-; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: movdqa 128(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,0],xmm3[2,0]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 288(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT: movdqa 240(%rdi), %xmm4
+; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: pand %xmm0, %xmm1
-; SSE-NEXT: movdqa 304(%rdi), %xmm2
+; SSE-NEXT: movdqa 256(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pandn %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa 320(%rdi), %xmm1
+; SSE-NEXT: movdqa 272(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
@@ -1413,319 +1413,319 @@
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{.*#+}} xmm13 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: pandn %xmm6, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm6, %xmm1
-; SSE-NEXT: movdqa %xmm13, %xmm2
-; SSE-NEXT: pandn %xmm10, %xmm2
-; SSE-NEXT: pand %xmm13, %xmm1
-; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm7[0,3,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
-; SSE-NEXT: movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT: movdqa %xmm6, %xmm0
-; SSE-NEXT: pandn %xmm2, %xmm0
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pandn %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm2
-; SSE-NEXT: pandn %xmm9, %xmm2
-; SSE-NEXT: pand %xmm13, %xmm1
-; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[0,3,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
-; SSE-NEXT: movdqa %xmm6, %xmm0
-; SSE-NEXT: pandn %xmm2, %xmm0
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm9
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pandn %xmm1, %xmm9
-; SSE-NEXT: movdqa %xmm13, %xmm2
-; SSE-NEXT: pandn %xmm15, %xmm2
-; SSE-NEXT: pand %xmm13, %xmm1
-; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[0,3,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
-; SSE-NEXT: movdqa %xmm6, %xmm0
-; SSE-NEXT: pandn %xmm2, %xmm0
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pandn %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm2
-; SSE-NEXT: pandn %xmm4, %xmm2
-; SSE-NEXT: pand %xmm13, %xmm1
-; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[0,3,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
-; SSE-NEXT: movdqa %xmm6, %xmm0
-; SSE-NEXT: pandn %xmm2, %xmm0
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm4
-; SSE-NEXT: pandn %xmm14, %xmm4
-; SSE-NEXT: movdqa %xmm14, %xmm1
-; SSE-NEXT: movdqa %xmm13, %xmm2
-; SSE-NEXT: pandn %xmm5, %xmm2
-; SSE-NEXT: pand %xmm13, %xmm1
-; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm15[0,3,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
-; SSE-NEXT: movdqa %xmm6, %xmm14
-; SSE-NEXT: pandn %xmm2, %xmm14
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm14
-; SSE-NEXT: movdqa %xmm13, %xmm0
+; SSE-NEXT: movdqa {{.*#+}} xmm7 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT: movdqa %xmm7, %xmm0
; SSE-NEXT: pandn %xmm12, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm2
-; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: pand %xmm13, %xmm12
-; SSE-NEXT: por %xmm2, %xmm12
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm11[0,3,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
-; SSE-NEXT: movdqa %xmm6, %xmm10
-; SSE-NEXT: pandn %xmm2, %xmm10
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm12[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm10
-; SSE-NEXT: movdqa %xmm13, %xmm5
-; SSE-NEXT: pandn %xmm8, %xmm5
-; SSE-NEXT: movdqa %xmm8, %xmm2
-; SSE-NEXT: movdqa %xmm13, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: pandn %xmm12, %xmm3
-; SSE-NEXT: pand %xmm13, %xmm2
-; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm11[0,3,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,4,4,5,6]
-; SSE-NEXT: movdqa %xmm6, %xmm8
-; SSE-NEXT: pandn %xmm1, %xmm8
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm8
-; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: movdqa %xmm7, %xmm1
; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm0
+; SSE-NEXT: pand %xmm7, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm15[0,3,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,6]
-; SSE-NEXT: movdqa %xmm6, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: movdqa {{.*#+}} xmm12 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: pandn %xmm1, %xmm2
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE-NEXT: pand %xmm6, %xmm0
-; SSE-NEXT: por %xmm0, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm2
-; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,4,7,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,1,0,2]
-; SSE-NEXT: movdqa %xmm6, %xmm7
-; SSE-NEXT: pandn %xmm1, %xmm7
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[3,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm7
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm7, %xmm15
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: pandn %xmm1, %xmm15
+; SSE-NEXT: movdqa %xmm7, %xmm2
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: pand %xmm7, %xmm1
+; SSE-NEXT: por %xmm2, %xmm1
+; SSE-NEXT: pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
+; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: pandn %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm7, %xmm2
+; SSE-NEXT: pandn %xmm5, %xmm2
+; SSE-NEXT: pand %xmm7, %xmm1
+; SSE-NEXT: por %xmm2, %xmm1
+; SSE-NEXT: pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
+; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm7, %xmm5
+; SSE-NEXT: pandn %xmm14, %xmm5
+; SSE-NEXT: movdqa %xmm7, %xmm2
+; SSE-NEXT: pandn (%rsp), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: pand %xmm7, %xmm14
+; SSE-NEXT: por %xmm2, %xmm14
+; SSE-NEXT: pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm14[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
+; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: pandn %xmm13, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm7, %xmm2
+; SSE-NEXT: pandn %xmm6, %xmm2
+; SSE-NEXT: pand %xmm7, %xmm1
+; SSE-NEXT: por %xmm2, %xmm1
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm11[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
+; SSE-NEXT: movdqa %xmm12, %xmm13
+; SSE-NEXT: pandn %xmm2, %xmm13
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
+; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: por %xmm1, %xmm13
+; SSE-NEXT: movdqa %xmm7, %xmm6
+; SSE-NEXT: pandn %xmm10, %xmm6
+; SSE-NEXT: movdqa %xmm7, %xmm3
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: pand %xmm7, %xmm10
+; SSE-NEXT: por %xmm3, %xmm10
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm14[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,6]
+; SSE-NEXT: movdqa %xmm12, %xmm11
+; SSE-NEXT: pandn %xmm3, %xmm11
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm10[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT: pand %xmm12, %xmm2
+; SSE-NEXT: por %xmm2, %xmm11
+; SSE-NEXT: movdqa %xmm7, %xmm2
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: movdqa %xmm9, %xmm3
+; SSE-NEXT: pand %xmm7, %xmm3
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm8[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
+; SSE-NEXT: movdqa %xmm12, %xmm9
+; SSE-NEXT: pandn %xmm2, %xmm9
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm3[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT: pand %xmm12, %xmm2
+; SSE-NEXT: por %xmm2, %xmm9
+; SSE-NEXT: movdqa %xmm7, %xmm1
+; SSE-NEXT: pandn %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm4, %xmm2
+; SSE-NEXT: movdqa %xmm7, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pandn %xmm10, %xmm3
+; SSE-NEXT: pand %xmm7, %xmm2
+; SSE-NEXT: por %xmm3, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm4[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,4,5,6]
+; SSE-NEXT: movdqa %xmm12, %xmm8
+; SSE-NEXT: pandn %xmm0, %xmm8
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm8
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm0
-; SSE-NEXT: por %xmm9, %xmm0
-; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
-; SSE-NEXT: movdqa %xmm6, %xmm9
-; SSE-NEXT: pandn %xmm1, %xmm9
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm9
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm0
-; SSE-NEXT: por %xmm4, %xmm0
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm15[0,1,2,3,4,7,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
-; SSE-NEXT: movdqa %xmm6, %xmm4
-; SSE-NEXT: pandn %xmm1, %xmm4
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: pand %xmm13, %xmm12
-; SSE-NEXT: por %xmm5, %xmm12
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm11[0,1,2,3,4,7,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,1,0,2]
-; SSE-NEXT: movdqa %xmm6, %xmm1
-; SSE-NEXT: pandn %xmm5, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm12[3,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[2,1,0,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[2,1,0,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm5
-; SSE-NEXT: por %xmm5, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm0
-; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,1,0,2]
-; SSE-NEXT: movdqa %xmm6, %xmm15
-; SSE-NEXT: pandn %xmm5, %xmm15
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[3,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[2,1,0,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[2,1,0,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm5
-; SSE-NEXT: por %xmm5, %xmm15
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm0
-; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: pand %xmm7, %xmm0
+; SSE-NEXT: por %xmm15, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm2
-; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,1,0,2]
-; SSE-NEXT: movdqa %xmm6, %xmm0
-; SSE-NEXT: pandn %xmm5, %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm2[3,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[2,1,0,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[2,1,0,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm5
+; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
+; SSE-NEXT: movdqa %xmm12, %xmm15
+; SSE-NEXT: pandn %xmm0, %xmm15
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[3,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm15
+; SSE-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT: pand %xmm7, %xmm0
; SSE-NEXT: por %xmm5, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm2
-; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,1,0,2]
-; SSE-NEXT: movdqa %xmm6, %xmm11
-; SSE-NEXT: pandn %xmm5, %xmm11
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm2[3,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[2,1,0,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[2,1,0,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm5
-; SSE-NEXT: por %xmm5, %xmm11
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm2
-; SSE-NEXT: pandn (%rsp), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT: por %xmm2, %xmm13
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm13[3,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[2,1,0,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[2,1,0,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm5
-; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT: # xmm13 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm13[0,1,0,2]
-; SSE-NEXT: pandn %xmm13, %xmm6
-; SSE-NEXT: por %xmm5, %xmm6
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 96(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 112(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 48(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 64(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movaps %xmm5, (%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movaps %xmm5, 80(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movaps %xmm5, 16(%rsi)
-; SSE-NEXT: movdqa %xmm3, 96(%rdx)
-; SSE-NEXT: movdqa %xmm8, 32(%rdx)
-; SSE-NEXT: movdqa %xmm10, 112(%rdx)
-; SSE-NEXT: movdqa %xmm14, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, (%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 80(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 16(%rdx)
-; SSE-NEXT: movdqa %xmm6, 96(%rcx)
-; SSE-NEXT: movdqa %xmm11, 112(%rcx)
-; SSE-NEXT: movdqa %xmm0, 64(%rcx)
-; SSE-NEXT: movdqa %xmm15, 80(%rcx)
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
+; SSE-NEXT: movdqa %xmm12, %xmm5
+; SSE-NEXT: pandn %xmm0, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[3,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm5
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pand %xmm7, %xmm0
+; SSE-NEXT: por %xmm6, %xmm0
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,4,7,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
+; SSE-NEXT: movdqa %xmm12, %xmm3
+; SSE-NEXT: pandn %xmm0, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[3,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm3
+; SSE-NEXT: pand %xmm7, %xmm10
+; SSE-NEXT: por %xmm1, %xmm10
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
+; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: pandn %xmm0, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[3,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: por %xmm0, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pand %xmm7, %xmm0
+; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: movdqa %xmm0, %xmm4
+; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm0[0,1,0,2]
+; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: pandn %xmm6, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm4[3,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[2,1,0,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[2,1,0,3,4,5,6,7]
+; SSE-NEXT: pand %xmm12, %xmm6
+; SSE-NEXT: por %xmm6, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pand %xmm7, %xmm4
+; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,0,2]
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pandn %xmm6, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm4[3,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[2,1,0,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[2,1,0,3,4,5,6,7]
+; SSE-NEXT: pand %xmm12, %xmm6
+; SSE-NEXT: por %xmm6, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pand %xmm7, %xmm4
+; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,0,2]
+; SSE-NEXT: movdqa %xmm12, %xmm14
+; SSE-NEXT: pandn %xmm6, %xmm14
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm4[3,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[2,1,0,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[2,1,0,3,4,5,6,7]
+; SSE-NEXT: pand %xmm12, %xmm6
+; SSE-NEXT: por %xmm6, %xmm14
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pand %xmm7, %xmm4
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: por %xmm4, %xmm7
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm7[3,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[2,1,0,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[2,1,0,3,4,5,6,7]
+; SSE-NEXT: pand %xmm12, %xmm6
+; SSE-NEXT: pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,0,2]
+; SSE-NEXT: pandn %xmm7, %xmm12
+; SSE-NEXT: por %xmm6, %xmm12
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 16(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 64(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, (%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 112(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 96(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movaps %xmm6, 32(%rsi)
+; SSE-NEXT: movdqa %xmm8, 80(%rdx)
+; SSE-NEXT: movdqa %xmm9, 16(%rdx)
+; SSE-NEXT: movdqa %xmm11, 64(%rdx)
+; SSE-NEXT: movdqa %xmm13, (%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 112(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 48(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 96(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 32(%rdx)
+; SSE-NEXT: movdqa %xmm12, 16(%rcx)
+; SSE-NEXT: movdqa %xmm14, (%rcx)
+; SSE-NEXT: movdqa %xmm0, 48(%rcx)
; SSE-NEXT: movdqa %xmm1, 32(%rcx)
-; SSE-NEXT: movdqa %xmm4, 48(%rcx)
-; SSE-NEXT: movdqa %xmm9, (%rcx)
-; SSE-NEXT: movdqa %xmm7, 16(%rcx)
-; SSE-NEXT: addq $408, %rsp # imm = 0x198
+; SSE-NEXT: movdqa %xmm2, 80(%rcx)
+; SSE-NEXT: movdqa %xmm3, 64(%rcx)
+; SSE-NEXT: movdqa %xmm5, 112(%rcx)
+; SSE-NEXT: movdqa %xmm15, 96(%rcx)
+; SSE-NEXT: addq $392, %rsp # imm = 0x188
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i16_stride3_vf64:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll
index 9b347e0..7220654 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-4.ll
@@ -1165,19 +1165,19 @@
; SSE-LABEL: load_i16_stride4_vf32:
; SSE: # %bb.0:
; SSE-NEXT: subq $248, %rsp
-; SSE-NEXT: movdqa 224(%rdi), %xmm3
+; SSE-NEXT: movdqa 96(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm4
+; SSE-NEXT: movdqa 64(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 208(%rdi), %xmm5
+; SSE-NEXT: movdqa 80(%rdi), %xmm5
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 96(%rdi), %xmm2
+; SSE-NEXT: movdqa 160(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rdi), %xmm6
+; SSE-NEXT: movdqa 176(%rdi), %xmm6
; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 64(%rdi), %xmm1
+; SSE-NEXT: movdqa 128(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 80(%rdi), %xmm0
+; SSE-NEXT: movdqa 144(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1199,10 +1199,10 @@
; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm4[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm9[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 240(%rdi), %xmm0
+; SSE-NEXT: movdqa 112(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm0[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm6[0,1,0,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[0,1,0,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm3[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm7[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
@@ -1212,34 +1212,34 @@
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 16(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm0[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm1[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm11[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm1[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm14[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 32(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 48(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm0[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[0,1,0,2,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm2[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm14[0,1,0,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm11[0,1,0,2,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm2[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm6[0,1,0,2,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 128(%rdi), %xmm0
+; SSE-NEXT: movdqa 192(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 144(%rdi), %xmm1
+; SSE-NEXT: movdqa 208(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm5[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm4[0,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 160(%rdi), %xmm2
+; SSE-NEXT: movdqa 224(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 176(%rdi), %xmm1
+; SSE-NEXT: movdqa 240(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm3[0,1,0,2,4,5,6,7]
@@ -1262,18 +1262,19 @@
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm10[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm9[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm6[0,1,1,3,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm7[0,1,1,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[0,1,1,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
+; SSE-NEXT: movapd %xmm7, (%rsp) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[1,3,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm14[1,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm11[0,1,1,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1]
-; SSE-NEXT: movapd %xmm6, (%rsp) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[1,3,2,3,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm11[1,3,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[0,1,1,3,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm14[0,1,1,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
+; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[1,3,2,3,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm4[1,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
@@ -1295,118 +1296,117 @@
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[3,1,2,3]
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm3[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm1[0],xmm10[1]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[3,1,2,3]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[3,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[2,0,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm6[2,0,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[3,1,2,3]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[3,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm4[0,1,2,0,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm13 = xmm2[0,1,2,0,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm5[0],xmm13[1],xmm5[1]
+; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm3[0],xmm13[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3]
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[2,0,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,1,2,3]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm0[2,0,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm1[2,0,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,1,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; SSE-NEXT: # xmm14 = mem[3,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm13 = xmm14[0,1,2,0,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm1[0],xmm13[1]
-; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,1,2,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3]
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,0,2,3,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[2,0,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm0[0,1,2,0,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm14[0,1,2,0,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm5[0],xmm9[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; SSE-NEXT: # xmm11 = mem[3,1,2,3]
-; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: # xmm7 = mem[3,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm11[0,1,2,0,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm7[0,1,2,0,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
-; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT: # xmm6 = mem[3,1,2,3]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[3,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm11[2,0,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[2,0,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; SSE-NEXT: # xmm5 = mem[3,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm6[2,0,2,3,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[2,0,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[3,1,2,3]
-; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm4[0,1,2,0,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm2[0,1,2,0,4,5,6,7]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[3,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm5[0,1,2,0,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm3[0,1,2,0,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm6[3,1,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,1,3,1,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,1,3,1,4,5,6,7]
+; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,3,1,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,3,1,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm14[0,1,3,1,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm11[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm5[0,1,3,1,4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm14 = xmm14[0,1,3,1,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm1[0],xmm14[1]
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm11[0,1,3,1,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,1,3,1,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm6[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm5[3,1,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%rsi)
-; SSE-NEXT: movapd %xmm12, 32(%rdx)
-; SSE-NEXT: movapd %xmm8, (%rdx)
-; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%rdx)
-; SSE-NEXT: movapd %xmm15, 32(%rcx)
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm4[0],xmm3[1]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, (%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 16(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 32(%rsi)
+; SSE-NEXT: movapd %xmm12, 48(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, (%rdx)
+; SSE-NEXT: movaps (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 16(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 32(%rdx)
+; SSE-NEXT: movapd %xmm15, 48(%rcx)
; SSE-NEXT: movapd %xmm9, (%rcx)
-; SSE-NEXT: movapd %xmm13, 48(%rcx)
-; SSE-NEXT: movapd %xmm10, 16(%rcx)
-; SSE-NEXT: movapd %xmm2, 32(%r8)
-; SSE-NEXT: movapd %xmm7, (%r8)
-; SSE-NEXT: movapd %xmm14, 48(%r8)
-; SSE-NEXT: movapd %xmm3, 16(%r8)
+; SSE-NEXT: movapd %xmm13, 16(%rcx)
+; SSE-NEXT: movapd %xmm10, 32(%rcx)
+; SSE-NEXT: movapd %xmm3, 48(%r8)
+; SSE-NEXT: movapd %xmm6, (%r8)
+; SSE-NEXT: movapd %xmm1, 32(%r8)
+; SSE-NEXT: movapd %xmm0, 16(%r8)
; SSE-NEXT: addq $248, %rsp
; SSE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
index 8bda8ab..888dbf5 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
@@ -1756,24 +1756,23 @@
define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4) nounwind {
; SSE-LABEL: load_i16_stride5_vf32:
; SSE: # %bb.0:
-; SSE-NEXT: subq $408, %rsp # imm = 0x198
-; SSE-NEXT: movdqa 64(%rdi), %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa (%rdi), %xmm12
-; SSE-NEXT: movdqa 16(%rdi), %xmm9
-; SSE-NEXT: movdqa 32(%rdi), %xmm8
-; SSE-NEXT: movdqa 48(%rdi), %xmm6
-; SSE-NEXT: movdqa 224(%rdi), %xmm7
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 160(%rdi), %xmm5
-; SSE-NEXT: movdqa 176(%rdi), %xmm10
-; SSE-NEXT: movdqa 208(%rdi), %xmm3
+; SSE-NEXT: subq $392, %rsp # imm = 0x188
+; SSE-NEXT: movdqa 304(%rdi), %xmm6
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 240(%rdi), %xmm12
+; SSE-NEXT: movdqa 256(%rdi), %xmm9
+; SSE-NEXT: movdqa 288(%rdi), %xmm7
+; SSE-NEXT: movdqa 272(%rdi), %xmm5
+; SSE-NEXT: movdqa 64(%rdi), %xmm8
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%rdi), %xmm11
+; SSE-NEXT: movdqa 16(%rdi), %xmm10
+; SSE-NEXT: movdqa 32(%rdi), %xmm13
+; SSE-NEXT: movdqa 48(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm11
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm11, %xmm1
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pandn %xmm13, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
@@ -1781,65 +1780,65 @@
; SSE-NEXT: movdqa %xmm10, %xmm15
; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[0,2,2,3]
-; SSE-NEXT: movdqa %xmm5, (%rsp) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[0,2,2,3]
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,3]
; SSE-NEXT: movaps {{.*#+}} xmm10 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT: andps %xmm10, %xmm3
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,1,0,1]
+; SSE-NEXT: andps %xmm10, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[0,1,0,1]
; SSE-NEXT: movaps %xmm10, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: por %xmm3, %xmm2
+; SSE-NEXT: por %xmm4, %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm8, %xmm1
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[0,1,0,3]
-; SSE-NEXT: movdqa %xmm6, %xmm14
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movdqa %xmm5, %xmm8
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[0,1,0,3]
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[3,1,2,3]
; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm12[0,2,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm12[0,2,2,3]
; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,1,0,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,1,0,1]
; SSE-NEXT: movaps %xmm10, %xmm2
; SSE-NEXT: andnps %xmm1, %xmm2
-; SSE-NEXT: movdqa 272(%rdi), %xmm7
-; SSE-NEXT: andps %xmm10, %xmm3
-; SSE-NEXT: orps %xmm3, %xmm2
+; SSE-NEXT: movdqa 192(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: andps %xmm10, %xmm4
+; SSE-NEXT: orps %xmm4, %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm7, %xmm1
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 288(%rdi), %xmm2
+; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: movdqa 208(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: movdqa 256(%rdi), %xmm6
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[3,1,2,3]
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: movdqa 240(%rdi), %xmm3
+; SSE-NEXT: movdqa 176(%rdi), %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[3,1,2,3]
+; SSE-NEXT: movdqa %xmm3, %xmm6
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
-; SSE-NEXT: movdqa 304(%rdi), %xmm1
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: movdqa 160(%rdi), %xmm14
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm14[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,3]
+; SSE-NEXT: movdqa 224(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
; SSE-NEXT: movaps %xmm10, %xmm2
; SSE-NEXT: andnps %xmm1, %xmm2
-; SSE-NEXT: andps %xmm10, %xmm3
-; SSE-NEXT: orps %xmm3, %xmm2
+; SSE-NEXT: andps %xmm10, %xmm4
+; SSE-NEXT: orps %xmm4, %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 128(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1850,12 +1849,12 @@
; SSE-NEXT: pandn %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa 96(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: movdqa 80(%rdi), %xmm13
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm13[0,2,2,3]
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 80(%rdi), %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[0,2,2,3]
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
@@ -1869,17 +1868,17 @@
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm15, %xmm0
; SSE-NEXT: psrlq $48, %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [0,0,0,65535,65535,65535,65535,65535]
; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[1,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm13[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,7,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
@@ -1888,11 +1887,11 @@
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT: movdqa %xmm11, %xmm2
; SSE-NEXT: psllq $48, %xmm2
-; SSE-NEXT: movaps %xmm10, %xmm3
-; SSE-NEXT: andnps %xmm2, %xmm3
+; SSE-NEXT: movaps %xmm10, %xmm4
+; SSE-NEXT: andnps %xmm2, %xmm4
; SSE-NEXT: pand %xmm10, %xmm1
-; SSE-NEXT: orps %xmm1, %xmm3
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: orps %xmm1, %xmm4
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm9, %xmm1
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm12[0,3,2,3]
@@ -1900,12 +1899,12 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm14[1,3,2,3]
-; SSE-NEXT: movdqa %xmm14, %xmm15
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm8[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[1,3,2,3]
+; SSE-NEXT: movdqa %xmm7, %xmm15
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm8[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm4[0,1,2,3,7,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,4,7]
@@ -1914,23 +1913,24 @@
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT: movdqa %xmm9, %xmm1
; SSE-NEXT: psllq $48, %xmm1
-; SSE-NEXT: movdqa %xmm10, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: movdqa %xmm10, %xmm4
+; SSE-NEXT: pandn %xmm1, %xmm4
; SSE-NEXT: pand %xmm10, %xmm2
-; SSE-NEXT: por %xmm2, %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrlq $48, %xmm6
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: por %xmm2, %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm6, %xmm1
+; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm14[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: pandn %xmm2, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm8[1,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm4[0,1,2,3,7,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,4,7]
@@ -1939,23 +1939,23 @@
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT: movdqa %xmm7, %xmm1
; SSE-NEXT: psllq $48, %xmm1
-; SSE-NEXT: movdqa %xmm10, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: movdqa %xmm10, %xmm4
+; SSE-NEXT: pandn %xmm1, %xmm4
; SSE-NEXT: pand %xmm10, %xmm2
-; SSE-NEXT: por %xmm2, %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: por %xmm2, %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%rsp), %xmm12 # 16-byte Reload
; SSE-NEXT: movdqa %xmm12, %xmm1
; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm13[0,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[1,3,2,3]
-; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,7,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
@@ -1969,20 +1969,19 @@
; SSE-NEXT: pandn %xmm1, %xmm10
; SSE-NEXT: por %xmm0, %xmm10
; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm4, %xmm0
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm3, %xmm0
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm13[0,0]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm13[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT: movaps {{.*#+}} xmm3 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: movaps {{.*#+}} xmm4 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT: movaps %xmm4, %xmm1
; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: movdqa (%rsp), %xmm4 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm4[0,1,1,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,1,1,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm10[2],xmm2[3],xmm10[3]
-; SSE-NEXT: pand %xmm3, %xmm2
+; SSE-NEXT: pand %xmm4, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,1,2,0]
@@ -1995,14 +1994,14 @@
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm15[0,0]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm15[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: movaps %xmm4, %xmm1
; SSE-NEXT: andnps %xmm0, %xmm1
; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[0,1,1,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT: pand %xmm3, %xmm2
+; SSE-NEXT: pand %xmm4, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[0,1,2,0]
@@ -2015,13 +2014,13 @@
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm11[0,0]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm11[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: movaps %xmm4, %xmm1
; SSE-NEXT: andnps %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm14[0,1,1,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT: pand %xmm3, %xmm2
+; SSE-NEXT: pand %xmm4, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,1,2,0]
@@ -2033,13 +2032,13 @@
; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,0],xmm1[0,0]
; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm1[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,0,1,3]
-; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: movaps %xmm4, %xmm1
; SSE-NEXT: andnps %xmm5, %xmm1
; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[0,1,1,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm12[2],xmm2[3],xmm12[3]
-; SSE-NEXT: pand %xmm3, %xmm2
+; SSE-NEXT: pand %xmm4, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,1,2,0]
@@ -2048,20 +2047,20 @@
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm13[2,2,2,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm10[0,2,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm3, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[2,0],xmm13[3,0]
-; SSE-NEXT: movaps %xmm3, %xmm0
+; SSE-NEXT: movaps %xmm4, %xmm0
; SSE-NEXT: andnps %xmm13, %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[0,1],xmm6[0,2]
@@ -2073,7 +2072,7 @@
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm15[2,2,2,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
@@ -2083,11 +2082,11 @@
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm3, %xmm9
+; SSE-NEXT: pand %xmm4, %xmm9
; SSE-NEXT: por %xmm1, %xmm9
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm15[3,0]
-; SSE-NEXT: movdqa %xmm3, %xmm10
+; SSE-NEXT: movdqa %xmm4, %xmm10
; SSE-NEXT: pandn %xmm15, %xmm10
; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[0,2]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm15[0,1,2,3,7,4,6,7]
@@ -2097,7 +2096,7 @@
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,1],xmm1[2,0]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm11[2,2,2,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
@@ -2106,11 +2105,11 @@
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm6[0,3,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm3, %xmm6
+; SSE-NEXT: pand %xmm4, %xmm6
; SSE-NEXT: por %xmm1, %xmm6
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm11[3,0]
-; SSE-NEXT: movdqa %xmm3, %xmm15
+; SSE-NEXT: movdqa %xmm4, %xmm15
; SSE-NEXT: pandn %xmm11, %xmm15
; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,1],xmm0[0,2]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,7,4,6,7]
@@ -2121,19 +2120,19 @@
; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm1[2,0]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm11[2,2,2,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; SSE-NEXT: movdqa (%rsp), %xmm2 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm2[0,2,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[0,3,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm3, %xmm12
+; SSE-NEXT: pand %xmm4, %xmm12
; SSE-NEXT: por %xmm1, %xmm12
-; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[3,0]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
@@ -2142,29 +2141,29 @@
; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[0,1],xmm0[0,2]
; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[2,0],xmm5[3,0]
; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm14[0,2]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm2[3,0]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm4[0,2]
-; SSE-NEXT: movaps %xmm2, %xmm4
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm2[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm3[0,2]
+; SSE-NEXT: movaps %xmm2, %xmm3
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT: movaps %xmm7, %xmm2
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm11[3,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[2,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm3, %xmm1
+; SSE-NEXT: pand %xmm4, %xmm1
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm3, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm5[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm14[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm14 = xmm14[2,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm3, %xmm14
-; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm4[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pand %xmm4, %xmm14
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm3[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm13[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm13 = xmm13[2,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm3, %xmm13
-; SSE-NEXT: pandn %xmm11, %xmm3
+; SSE-NEXT: pand %xmm4, %xmm13
+; SSE-NEXT: pandn %xmm11, %xmm4
; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,1],xmm2[0,2]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm11[0,1,2,3,7,4,6,7]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
@@ -2184,62 +2183,62 @@
; SSE-NEXT: por %xmm10, %xmm0
; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[0,2,2,3]
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,1,1,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[3,1],xmm2[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm2[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,0]
; SSE-NEXT: por %xmm15, %xmm14
; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[0,2,2,3]
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,1,1,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[3,1],xmm2[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,1],xmm4[2,0]
-; SSE-NEXT: por %xmm13, %xmm3
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm2[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,1],xmm3[2,0]
+; SSE-NEXT: por %xmm13, %xmm4
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[0,2,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm5[0,1,1,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[3,1],xmm2[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm2[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,0]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 32(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 48(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, (%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 32(%rcx)
; SSE-NEXT: movaps %xmm12, 16(%r8)
-; SSE-NEXT: movaps %xmm6, 48(%r8)
-; SSE-NEXT: movaps %xmm9, (%r8)
+; SSE-NEXT: movaps %xmm6, 32(%r8)
+; SSE-NEXT: movaps %xmm9, 48(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 32(%r8)
-; SSE-NEXT: movaps %xmm3, 16(%r9)
-; SSE-NEXT: movaps %xmm14, 48(%r9)
-; SSE-NEXT: movaps %xmm0, (%r9)
-; SSE-NEXT: movaps %xmm1, 32(%r9)
-; SSE-NEXT: addq $408, %rsp # imm = 0x198
+; SSE-NEXT: movaps %xmm2, (%r8)
+; SSE-NEXT: movaps %xmm4, 16(%r9)
+; SSE-NEXT: movaps %xmm14, 32(%r9)
+; SSE-NEXT: movaps %xmm0, 48(%r9)
+; SSE-NEXT: movaps %xmm1, (%r9)
+; SSE-NEXT: addq $392, %rsp # imm = 0x188
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i16_stride5_vf32:
@@ -3658,29 +3657,30 @@
; SSE-LABEL: load_i16_stride5_vf64:
; SSE: # %bb.0:
; SSE-NEXT: subq $1000, %rsp # imm = 0x3E8
-; SSE-NEXT: movdqa 464(%rdi), %xmm4
-; SSE-NEXT: movdqa 400(%rdi), %xmm10
-; SSE-NEXT: movdqa 416(%rdi), %xmm11
-; SSE-NEXT: movdqa 448(%rdi), %xmm5
+; SSE-NEXT: movdqa 544(%rdi), %xmm4
+; SSE-NEXT: movdqa %xmm4, (%rsp) # 16-byte Spill
+; SSE-NEXT: movdqa 480(%rdi), %xmm10
+; SSE-NEXT: movdqa 496(%rdi), %xmm11
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 528(%rdi), %xmm5
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 432(%rdi), %xmm12
-; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 144(%rdi), %xmm7
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 80(%rdi), %xmm6
-; SSE-NEXT: movdqa 96(%rdi), %xmm9
-; SSE-NEXT: movdqa 128(%rdi), %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rdi), %xmm8
+; SSE-NEXT: movdqa 512(%rdi), %xmm8
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 224(%rdi), %xmm7
+; SSE-NEXT: movdqa 160(%rdi), %xmm6
+; SSE-NEXT: movdqa 176(%rdi), %xmm9
+; SSE-NEXT: movdqa 208(%rdi), %xmm12
+; SSE-NEXT: movdqa 192(%rdi), %xmm13
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm8, %xmm1
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
+; SSE-NEXT: pandn %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm12[0,1,0,3]
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[3,1,2,3]
-; SSE-NEXT: movdqa %xmm9, %xmm13
+; SSE-NEXT: movdqa %xmm9, %xmm14
; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[0,2,2,3]
@@ -3692,47 +3692,99 @@
; SSE-NEXT: movaps {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,65535,65535,0]
; SSE-NEXT: andps %xmm6, %xmm3
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,1,0,1]
+; SSE-NEXT: movdqa %xmm7, %xmm9
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm6, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
; SSE-NEXT: por %xmm3, %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm12, %xmm1
+; SSE-NEXT: pandn %xmm8, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[0,1,0,3]
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[3,1,2,3]
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm10[0,2,2,3]
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,1,0,1]
-; SSE-NEXT: movdqa %xmm4, %xmm9
+; SSE-NEXT: movaps %xmm6, %xmm2
+; SSE-NEXT: andnps %xmm1, %xmm2
+; SSE-NEXT: movdqa 272(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm6, %xmm4
-; SSE-NEXT: andnps %xmm1, %xmm4
-; SSE-NEXT: movdqa 32(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: andps %xmm6, %xmm3
-; SSE-NEXT: orps %xmm3, %xmm4
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: orps %xmm3, %xmm2
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: movdqa 48(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, (%rsp) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,3]
-; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa 16(%rdi), %xmm7
+; SSE-NEXT: pandn %xmm4, %xmm1
+; SSE-NEXT: movdqa 288(%rdi), %xmm11
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm11[0,1,0,3]
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movdqa 256(%rdi), %xmm7
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[3,1,2,3]
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: movdqa (%rdi), %xmm5
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm5[0,2,2,3]
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 240(%rdi), %xmm8
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm8[0,2,2,3]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
+; SSE-NEXT: movdqa 304(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; SSE-NEXT: movaps %xmm6, %xmm2
+; SSE-NEXT: andnps %xmm1, %xmm2
+; SSE-NEXT: andps %xmm6, %xmm3
+; SSE-NEXT: orps %xmm3, %xmm2
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 592(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm2, %xmm1
+; SSE-NEXT: movdqa 608(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movdqa 576(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: movdqa 560(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
+; SSE-NEXT: movdqa 624(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; SSE-NEXT: movaps %xmm6, %xmm2
+; SSE-NEXT: andnps %xmm1, %xmm2
+; SSE-NEXT: andps %xmm6, %xmm3
+; SSE-NEXT: orps %xmm3, %xmm2
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 32(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm2, %xmm1
+; SSE-NEXT: movdqa 48(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,3]
+; SSE-NEXT: pand %xmm0, %xmm3
+; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: movdqa 16(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: movdqa (%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,3]
@@ -3753,526 +3805,547 @@
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,3]
; SSE-NEXT: pand %xmm0, %xmm3
; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa 336(%rdi), %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[3,1,2,3]
-; SSE-NEXT: movdqa %xmm2, %xmm14
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 336(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: movdqa 320(%rdi), %xmm12
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm12[0,2,2,3]
-; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 320(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,3]
; SSE-NEXT: movdqa 384(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
-; SSE-NEXT: movaps %xmm6, %xmm3
-; SSE-NEXT: andnps %xmm1, %xmm3
-; SSE-NEXT: andps %xmm6, %xmm4
-; SSE-NEXT: orps %xmm4, %xmm3
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 272(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: movdqa 288(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,3]
-; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa 256(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: movdqa 240(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,3]
-; SSE-NEXT: movdqa 304(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
-; SSE-NEXT: movaps %xmm6, %xmm3
-; SSE-NEXT: andnps %xmm1, %xmm3
-; SSE-NEXT: andps %xmm6, %xmm4
-; SSE-NEXT: orps %xmm4, %xmm3
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 592(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: movdqa 608(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,3]
-; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa 576(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: movdqa 560(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,3]
-; SSE-NEXT: movdqa 624(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
; SSE-NEXT: movaps %xmm6, %xmm2
; SSE-NEXT: andnps %xmm1, %xmm2
; SSE-NEXT: andps %xmm6, %xmm4
; SSE-NEXT: orps %xmm4, %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: movdqa 208(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,3]
+; SSE-NEXT: movdqa 112(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: movdqa 128(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm1[0,1,0,3]
+; SSE-NEXT: pand %xmm0, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: movdqa 96(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[3,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
+; SSE-NEXT: movdqa 80(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[2,3]
+; SSE-NEXT: movdqa 144(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,1,0,1]
+; SSE-NEXT: movaps %xmm6, %xmm1
+; SSE-NEXT: andnps %xmm3, %xmm1
+; SSE-NEXT: andps %xmm6, %xmm5
+; SSE-NEXT: orps %xmm5, %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 448(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,1,0,3]
; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa 176(%rdi), %xmm1
+; SSE-NEXT: movdqa 432(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: movdqa 160(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[0,2,2,3]
+; SSE-NEXT: pandn %xmm1, %xmm0
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: movdqa 416(%rdi), %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[3,1,2,3]
+; SSE-NEXT: movdqa %xmm1, %xmm2
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
+; SSE-NEXT: movdqa 400(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,3]
-; SSE-NEXT: movdqa 224(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
-; SSE-NEXT: movaps %xmm6, %xmm2
-; SSE-NEXT: andnps %xmm1, %xmm2
-; SSE-NEXT: andps %xmm6, %xmm4
-; SSE-NEXT: orps %xmm4, %xmm2
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 528(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT: pand %xmm0, %xmm1
-; SSE-NEXT: movdqa 512(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pandn %xmm2, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa 496(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: movdqa 480(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,3]
-; SSE-NEXT: movdqa 544(%rdi), %xmm0
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[2,3]
+; SSE-NEXT: movdqa 464(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
; SSE-NEXT: movaps %xmm6, %xmm1
; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: andps %xmm6, %xmm3
+; SSE-NEXT: andps %xmm6, %xmm4
+; SSE-NEXT: orps %xmm4, %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: psrlq $48, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm15[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT: movdqa {{.*#+}} xmm0 = [0,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm0, %xmm4
+; SSE-NEXT: pandn %xmm3, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm12[1,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm13[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm3
+; SSE-NEXT: por %xmm4, %xmm3
+; SSE-NEXT: movdqa %xmm9, %xmm4
+; SSE-NEXT: psllq $48, %xmm4
+; SSE-NEXT: movaps %xmm6, %xmm1
+; SSE-NEXT: andnps %xmm4, %xmm1
+; SSE-NEXT: pand %xmm6, %xmm3
; SSE-NEXT: orps %xmm3, %xmm1
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: psrlq $48, %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,2,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm0 = [0,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm3
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm10[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; SSE-NEXT: movdqa %xmm0, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: pandn %xmm4, %xmm3
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[1,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm8[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
-; SSE-NEXT: pand %xmm0, %xmm1
-; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm13, %xmm3
-; SSE-NEXT: psllq $48, %xmm3
-; SSE-NEXT: movaps %xmm6, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm2
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: orps %xmm1, %xmm2
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrlq $48, %xmm11
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm10[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm11[0],xmm3[1],xmm11[1]
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,3,2,3]
-; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,4,7]
-; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: psllq $48, %xmm9
-; SSE-NEXT: movdqa %xmm6, %xmm2
-; SSE-NEXT: pandn %xmm9, %xmm2
-; SSE-NEXT: pand %xmm6, %xmm3
-; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrlq $48, %xmm7
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: pshufd $237, (%rsp), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,3,2,3]
-; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,4,7]
-; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm1
-; SSE-NEXT: psllq $48, %xmm1
-; SSE-NEXT: movdqa %xmm6, %xmm2
-; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: pand %xmm6, %xmm3
-; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm14, %xmm1
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm12[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,3,2,3]
-; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,4,7]
-; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psllq $48, %xmm1
-; SSE-NEXT: movdqa %xmm6, %xmm2
-; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: pand %xmm6, %xmm3
-; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm14, %xmm1
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm12[1,3,2,3]
-; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,4,7]
-; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm11, %xmm1
-; SSE-NEXT: psllq $48, %xmm1
-; SSE-NEXT: movdqa %xmm6, %xmm2
-; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: pand %xmm6, %xmm3
-; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,3,2,3]
-; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,4,7]
-; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psllq $48, %xmm1
-; SSE-NEXT: movdqa %xmm6, %xmm2
-; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: pand %xmm6, %xmm3
-; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm8, %xmm1
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[1,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm9[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,4,7]
-; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm5, %xmm1
-; SSE-NEXT: psllq $48, %xmm1
-; SSE-NEXT: movdqa %xmm6, %xmm4
-; SSE-NEXT: pandn %xmm1, %xmm4
-; SSE-NEXT: pand %xmm6, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm15[1,3,2,3]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT: psllq $48, %xmm3
+; SSE-NEXT: movdqa %xmm6, %xmm1
+; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: por %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $48, %xmm7
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm8[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: pandn %xmm4, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm11[1,3,2,3]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psllq $48, %xmm3
+; SSE-NEXT: movdqa %xmm6, %xmm1
+; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: por %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm3
+; SSE-NEXT: pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: pandn %xmm4, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm13[1,3,2,3]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: psllq $48, %xmm3
+; SSE-NEXT: movdqa %xmm6, %xmm1
+; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: por %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm3
+; SSE-NEXT: pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: pandn %xmm4, %xmm3
+; SSE-NEXT: pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[1,3,2,3]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psllq $48, %xmm3
+; SSE-NEXT: movdqa %xmm6, %xmm1
+; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: por %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm11[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: pandn %xmm4, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm8[1,3,2,3]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm9, %xmm3
+; SSE-NEXT: psllq $48, %xmm3
+; SSE-NEXT: movdqa %xmm6, %xmm1
+; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: por %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm3
+; SSE-NEXT: pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: pandn %xmm4, %xmm3
+; SSE-NEXT: pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[1,3,2,3]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psllq $48, %xmm3
+; SSE-NEXT: movdqa %xmm6, %xmm1
+; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: por %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: movdqa %xmm2, %xmm12
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $48, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm7[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[1,3,2,3]
-; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
-; SSE-NEXT: pand %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[1,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm10[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,4,7]
+; SSE-NEXT: pand %xmm0, %xmm3
+; SSE-NEXT: pandn %xmm4, %xmm0
+; SSE-NEXT: por %xmm3, %xmm0
; SSE-NEXT: pand %xmm6, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psllq $48, %xmm1
-; SSE-NEXT: pandn %xmm1, %xmm6
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: psllq $48, %xmm3
+; SSE-NEXT: pandn %xmm3, %xmm6
; SSE-NEXT: por %xmm0, %xmm6
; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm3[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
; SSE-NEXT: movaps {{.*#+}} xmm15 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,1,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; SSE-NEXT: pand %xmm15, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: movaps %xmm15, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm3
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT: pand %xmm15, %xmm5
+; SSE-NEXT: por %xmm3, %xmm5
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm13[0,1,1,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm6[2],xmm3[3],xmm6[3]
-; SSE-NEXT: pand %xmm15, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm12, %xmm0
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pshufd $36, (%rsp), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm3[2,0]
+; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[1,0],xmm3[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[2,0],xmm3[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[0,0,1,3]
+; SSE-NEXT: movaps %xmm15, %xmm3
+; SSE-NEXT: andnps %xmm13, %xmm3
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT: pand %xmm15, %xmm5
+; SSE-NEXT: por %xmm3, %xmm5
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm14[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm3[2,0]
+; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[1,0],xmm3[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[2,0],xmm3[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[0,0,1,3]
+; SSE-NEXT: movaps %xmm15, %xmm3
+; SSE-NEXT: andnps %xmm8, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm11[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT: pand %xmm15, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,0]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm10[0,0]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm10[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,1,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm14[2],xmm3[3],xmm14[3]
-; SSE-NEXT: pand %xmm15, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: movaps %xmm15, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm7[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm12[2],xmm1[3],xmm12[3]
+; SSE-NEXT: pand %xmm15, %xmm1
+; SSE-NEXT: por %xmm3, %xmm1
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm2, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm9[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm9[2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm12[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm12[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[0,1,1,3]
-; SSE-NEXT: movdqa %xmm7, %xmm14
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,7,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm8[2],xmm2[3],xmm8[3]
-; SSE-NEXT: pand %xmm15, %xmm2
-; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps %xmm8, %xmm0
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm7[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,1,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,7,6,7]
+; SSE-NEXT: movaps %xmm15, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm3
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm4[0,1,2,3,4,7,6,7]
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
; SSE-NEXT: pand %xmm15, %xmm2
-; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: por %xmm3, %xmm2
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT: pshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm9[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm9[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
+; SSE-NEXT: movaps %xmm15, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm3
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT: pand %xmm15, %xmm1
+; SSE-NEXT: por %xmm3, %xmm1
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movaps %xmm6, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm5[0,0]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm5[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm12[0,1,1,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm11[2],xmm2[3],xmm11[3]
-; SSE-NEXT: pand %xmm15, %xmm2
-; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movaps %xmm15, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm7[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT: pand %xmm15, %xmm1
+; SSE-NEXT: por %xmm3, %xmm1
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm8[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movaps %xmm10, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,0]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,1,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,7,6,7]
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT: pand %xmm15, %xmm2
-; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movaps %xmm15, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm14[0,1,1,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm13[2],xmm4[3],xmm13[3]
+; SSE-NEXT: pand %xmm15, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm4[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,1,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,7,6,7]
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT: pand %xmm15, %xmm2
-; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[2,2,2,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[0,2,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,0]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[2,2,2,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm15, %xmm3
+; SSE-NEXT: pandn %xmm0, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm4[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm15, %xmm0
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: movaps %xmm6, %xmm7
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm5[3,0]
+; SSE-NEXT: movaps %xmm15, %xmm0
+; SSE-NEXT: andnps %xmm5, %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm7[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,7,4,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm8[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,6]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[2,2,2,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm15, %xmm2
+; SSE-NEXT: pandn %xmm0, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm13[0,2,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[0,3,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm15, %xmm0
+; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: movaps %xmm10, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[3,0]
+; SSE-NEXT: movaps %xmm15, %xmm3
+; SSE-NEXT: andnps %xmm1, %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,7,4,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[2,2,2,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm15, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm3
-; SSE-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; SSE-NEXT: movaps %xmm15, %xmm0
-; SSE-NEXT: andnps %xmm2, %xmm0
-; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,7,4,6,7]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm12[3,0]
+; SSE-NEXT: movaps %xmm15, %xmm1
+; SSE-NEXT: andnps %xmm12, %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[0,1],xmm0[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm12[0,1,2,3,7,4,6,7]
; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,0]
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm9[2,2,2,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,2,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm15, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm9[3,0]
+; SSE-NEXT: movaps %xmm15, %xmm1
+; SSE-NEXT: andnps %xmm9, %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,1],xmm0[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,7,4,6,7]
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[2,2,2,2,4,5,6,7]
; SSE-NEXT: movdqa %xmm15, %xmm1
@@ -4302,117 +4375,41 @@
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm9[2,2,2,2,4,5,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[2,2,2,2,4,5,6,7]
; SSE-NEXT: movdqa %xmm15, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm14, %xmm13
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
-; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm15, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, %xmm2
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm9[3,0]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm9, %xmm1
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,1],xmm0[0,2]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,7,4,6,7]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm10[2,2,2,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm15, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, %xmm2
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm10[3,0]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm10, %xmm1
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[0,2]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,7,4,6,7]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[2,2,2,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
-; SSE-NEXT: movdqa %xmm11, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm11[0,2,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm15, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm0[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm15, %xmm10
+; SSE-NEXT: por %xmm1, %xmm10
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm5[3,0]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm5, %xmm1
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[0,2]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,7,4,6,7]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[3,0]
+; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: movaps %xmm15, %xmm2
+; SSE-NEXT: andnps %xmm3, %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[2,2,2,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[0,2,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm11 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm15, %xmm11
-; SSE-NEXT: por %xmm1, %xmm11
-; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[2,0],xmm7[3,0]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm7, %xmm1
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm8[0,2]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,7,4,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,7,4,6,7]
; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,1],xmm1[2,0]
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,1],xmm1[2,0]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm4[2,2,2,2,4,5,6,7]
; SSE-NEXT: movdqa %xmm15, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[0,2,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[0,2,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
@@ -4421,69 +4418,72 @@
; SSE-NEXT: por %xmm1, %xmm9
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[3,0]
-; SSE-NEXT: movaps %xmm15, %xmm1
-; SSE-NEXT: andnps %xmm4, %xmm1
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[0,2]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,7,4,6,7]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: movaps %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm15, %xmm12
+; SSE-NEXT: pandn %xmm4, %xmm12
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,7,4,6,7]
+; SSE-NEXT: pshufd $196, (%rsp), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,1],xmm1[2,0]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm14[2,2,2,2,4,5,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm13[2,2,2,2,4,5,6,7]
; SSE-NEXT: movdqa %xmm15, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm4[0,2,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[0,2,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm0[1,0,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm15, %xmm8
; SSE-NEXT: por %xmm1, %xmm8
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm7[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm1[0,2]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[3,0]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
-; SSE-NEXT: movaps %xmm0, %xmm2
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[3,0]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
; SSE-NEXT: movaps %xmm0, %xmm1
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[2,0],xmm0[3,0]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm13[0,2]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0,2]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm6, %xmm13
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[2,0],xmm13[3,0]
-; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[0,1],xmm6[0,2]
-; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[2,0],xmm3[3,0]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm12[0,2]
-; SSE-NEXT: movaps %xmm3, %xmm12
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[3,0]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[0,2]
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[2,0],xmm7[3,0]
-; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm10[0,2]
-; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm4[3,0]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm5[0,2]
-; SSE-NEXT: movaps %xmm4, %xmm10
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[2,0],xmm11[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,1],xmm14[0,2]
+; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm6, %xmm14
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm6[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,1],xmm0[0,2]
+; SSE-NEXT: movdqa %xmm5, %xmm11
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm5[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,1],xmm4[0,2]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm14[3,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm13[3,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm0[2,1,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm15, %xmm7
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufhw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm0[2,1,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm15, %xmm6
@@ -4492,11 +4492,12 @@
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm0[2,1,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm15, %xmm5
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm0[2,1,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm15, %xmm3
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm12[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufhw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[2,1,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm15, %xmm2
@@ -4505,145 +4506,144 @@
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm0[2,1,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm15, %xmm1
-; SSE-NEXT: pshufhw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm15, %xmm0
-; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm10[0,1,2,3,4,6,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm13[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm13 = xmm13[2,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm15, %xmm13
-; SSE-NEXT: pandn %xmm14, %xmm15
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,1],xmm4[0,2]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm14[0,1,2,3,7,4,6,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm10[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,5,5,6]
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[3,1],xmm4[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[0,1],xmm14[2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm11[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm14[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm14 = xmm14[2,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm15, %xmm14
+; SSE-NEXT: pandn %xmm13, %xmm15
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[0,1],xmm4[0,2]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm13[0,1,2,3,7,4,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm11[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,5,6]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[3,1],xmm4[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[0,1],xmm13[2,0]
; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: pshufd $232, (%rsp), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT: # xmm14 = mem[0,1,1,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = mem[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,5,4,7]
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[3,1],xmm4[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm14[2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[3,1],xmm4[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm13[2,0]
; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT: # xmm14 = mem[0,1,1,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = mem[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,5,4,7]
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[3,1],xmm4[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm14[2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[3,1],xmm4[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm13[2,0]
; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT: # xmm14 = mem[0,1,1,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = mem[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,5,4,7]
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[3,1],xmm4[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm14[2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[3,1],xmm4[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm13[2,0]
; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT: # xmm14 = mem[0,1,1,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = mem[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,5,4,7]
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[3,1],xmm4[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm14[2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[3,1],xmm4[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm13[2,0]
; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT: # xmm14 = mem[0,1,1,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = mem[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,5,4,7]
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[3,1],xmm4[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm14[2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[3,1],xmm4[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm13[2,0]
; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT: # xmm14 = mem[0,1,1,3]
+; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = mem[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,5,4,7]
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[3,1],xmm4[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,0]
-; SSE-NEXT: por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,4,7]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[3,1],xmm4[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,0]
+; SSE-NEXT: por %xmm12, %xmm0
; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT: pshufd $212, (%rsp), %xmm12 # 16-byte Folded Reload
; SSE-NEXT: # xmm12 = mem[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,5,4,7]
; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[3,1],xmm4[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm12[2,0]
-; SSE-NEXT: por %xmm13, %xmm15
+; SSE-NEXT: por %xmm14, %xmm15
; SSE-NEXT: pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[0,2,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm10[0,1,1,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm11[0,1,1,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,5,4,7]
; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[3,1],xmm4[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,1],xmm12[2,0]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 96(%rsi)
+; SSE-NEXT: movaps %xmm4, 80(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 112(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 48(%rsi)
+; SSE-NEXT: movaps %xmm4, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: movaps %xmm4, 64(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: movaps %xmm4, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 80(%rsi)
+; SSE-NEXT: movaps %xmm4, 112(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 16(%rsi)
+; SSE-NEXT: movaps %xmm4, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 96(%rdx)
+; SSE-NEXT: movaps %xmm4, 96(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 32(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 112(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, (%rdx)
+; SSE-NEXT: movaps %xmm4, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: movaps %xmm4, 80(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: movaps %xmm4, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 96(%rcx)
+; SSE-NEXT: movaps %xmm4, 64(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 112(%rcx)
+; SSE-NEXT: movaps %xmm4, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 64(%rcx)
+; SSE-NEXT: movaps %xmm4, 112(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 80(%rcx)
+; SSE-NEXT: movaps %xmm4, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 32(%rcx)
+; SSE-NEXT: movaps %xmm4, 96(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 48(%rcx)
+; SSE-NEXT: movaps %xmm4, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: movaps %xmm4, (%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 16(%rcx)
+; SSE-NEXT: movaps %xmm4, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 80(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 64(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 112(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 96(%rcx)
; SSE-NEXT: movaps %xmm8, 112(%r8)
; SSE-NEXT: movaps %xmm9, 96(%r8)
-; SSE-NEXT: movaps %xmm11, 80(%r8)
+; SSE-NEXT: movaps %xmm10, 80(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: movaps %xmm4, 64(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
@@ -4668,24 +4668,25 @@
; AVX1-ONLY-LABEL: load_i16_stride5_vf64:
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: subq $1032, %rsp # imm = 0x408
-; AVX1-ONLY-NEXT: vmovdqa 304(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 304(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm12
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm15
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm1 = [6,7,2,3,4,5,6,7,6,7,2,3,12,13,6,7]
; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa 256(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,1,3]
+; AVX1-ONLY-NEXT: vmovdqa 256(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm8[0,1,1,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vmovdqa 272(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm3[1]
-; AVX1-ONLY-NEXT: vmovdqa 240(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm13
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 240(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm7[0,2,2,3]
+; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm0[5,6,7]
@@ -4695,9 +4696,9 @@
; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4],xmm0[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm11[3,1,2,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm4
; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -4719,9 +4720,8 @@
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,1,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vmovdqa 592(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, %xmm9
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovdqa 560(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[0,2,2,3]
@@ -4730,7 +4730,7 @@
; AVX1-ONLY-NEXT: vmovdqa 624(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 608(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0,1],xmm3[2,3],xmm0[4,5,6,7]
; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm3, %xmm3
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm3[5,6,7]
@@ -4738,17 +4738,17 @@
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[3,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 480(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm8[0,2,2,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 480(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-ONLY-NEXT: vmovdqa 528(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,1,0,3]
-; AVX1-ONLY-NEXT: vmovdqa 512(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm7[4],xmm4[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 512(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm9[4],xmm4[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4,5,6,7]
; AVX1-ONLY-NEXT: vandps %ymm6, %ymm3, %ymm3
; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm0
@@ -4758,16 +4758,16 @@
; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm3
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[0,1,1,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,1,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vmovdqa 112(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm10[0,2,2,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[0,2,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa 144(%rdi), %xmm3
@@ -4781,9 +4781,9 @@
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[3,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm14
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm14[0,2,2,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-ONLY-NEXT: vmovdqa 32(%rdi), %xmm5
@@ -4811,9 +4811,9 @@
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,1,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vmovdqa 432(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovdqa 432(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm13[1]
+; AVX1-ONLY-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 400(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[0,2,2,3]
@@ -4832,132 +4832,133 @@
; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm0[0,1,0,3]
-; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3],xmm0[4],xmm5[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3],xmm14[4],xmm5[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm5[4,5,6,7]
; AVX1-ONLY-NEXT: vandps %ymm6, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT: vandnps %ymm15, %ymm6, %ymm15
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm15, %ymm2
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT: vandnps %ymm11, %ymm6, %ymm11
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm11, %ymm2
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm10[0,1],xmm8[2,3],xmm10[4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,3,0,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm13[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm7[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm0[0,1,2,3],mem[4,5],xmm0[6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm15[0,1,2,3],xmm12[4,5],xmm15[6,7]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm15, %xmm15
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm15 = xmm2[0,1,2,3,4],xmm15[5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm11[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm11 = xmm12[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[1,2,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm15
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm15 = xmm11[0],xmm15[0],xmm11[1],xmm15[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm7[0,1],xmm10[2,3],xmm7[4,5],xmm10[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm11 = [0,1,0,1,0,1,0,1,10,11,4,5,14,15,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsllq $48, %xmm3, %xmm15
+; AVX1-ONLY-NEXT: vandnps %ymm15, %ymm6, %ymm15
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm15, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,0,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = mem[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm11, %xmm13
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm2[0],xmm13[0],xmm2[1],xmm13[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm11[2,3],mem[4,5],xmm11[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,0,1,0,1,0,1,10,11,4,5,14,15,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm13[0,1,2],xmm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandps %ymm6, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsllq $48, %xmm3, %xmm13
-; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm6, %ymm13
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm13, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm9[0,1],mem[2,3],xmm9[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,0,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[1,2,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm13[0,1],xmm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $207, (%rsp), %xmm3, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,1,2,3],xmm3[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm13[5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm8[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[1,2,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3],xmm2[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm2[5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrlq $48, %xmm3, %xmm15
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm15 = xmm7[0,1],xmm8[2,3],xmm7[4,5],xmm8[6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm15, %xmm15
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2],xmm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandps %ymm6, %ymm13, %ymm13
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsllq $48, %xmm9, %xmm15
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1]
+; AVX1-ONLY-NEXT: vpblendw $204, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm9[0,1],mem[2,3],xmm9[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm15, %xmm15
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm15[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsllq $48, %xmm3, %xmm15
; AVX1-ONLY-NEXT: vandnps %ymm15, %ymm6, %ymm15
-; AVX1-ONLY-NEXT: vorps %ymm15, %ymm13, %ymm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm13, %ymm0
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm15, %ymm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm7[0,1],xmm12[2,3],xmm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm9[2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,3,0,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm10[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[1,2,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm13[0,1],xmm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm10[0,1,2,3],xmm12[4,5],xmm10[6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm13[5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm14[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[1,2,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm14, %xmm15
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm15 = xmm3[0,1],xmm4[2,3],xmm3[4,5],xmm4[6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm15, %xmm15
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2],xmm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandps %ymm6, %ymm13, %ymm13
+; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsllq $48, %xmm5, %xmm15
+; AVX1-ONLY-NEXT: vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3],xmm5[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm2[5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $236, (%rsp), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm15
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = mem[0,1],xmm3[2,3],mem[4,5],xmm3[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm15, %xmm15
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm15[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsllq $48, %xmm15, %xmm15
; AVX1-ONLY-NEXT: vandnps %ymm15, %ymm6, %ymm15
-; AVX1-ONLY-NEXT: vorps %ymm15, %ymm13, %ymm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm13, %ymm0
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm15, %ymm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = xmm0[0,1,2,3],mem[4,5],xmm0[6,7]
; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[0,1],mem[2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm13[0,1],mem[2,3],xmm13[4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,3,0,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[1,2,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4],xmm0[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $204, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[0,1],mem[2,3],xmm1[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm1, %xmm1
; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = mem[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4],xmm0[5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm13[0],xmm2[1],xmm13[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm14[0,1],xmm13[2,3],xmm14[4,5],xmm13[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vandps %ymm6, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
@@ -4966,217 +4967,218 @@
; AVX1-ONLY-NEXT: vorps %ymm2, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm0[0,1,2,3],mem[4,5],xmm0[6,7]
+; AVX1-ONLY-NEXT: vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3],xmm8[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,u,u,8,9,2,3,12,13,12,13,12,13,12,13>
; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm1, %xmm1
; AVX1-ONLY-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = mem[3,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm2[0,1],mem[2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm8[0,1],mem[2,3],xmm8[4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm6, %xmm6
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm1[0,1,2,3,4],xmm6[5,6,7]
-; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,1],xmm11[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm7[0,1],xmm10[2,3],xmm7[4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm1 = [2,3,2,3,2,3,2,3,12,13,6,7,12,13,14,15]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm12[0,1,1,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,7,6,7]
+; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm15[0,1,2],xmm11[3,4,5],xmm15[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm14[0,1,2,0]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,6,5]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5],xmm15[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm11, %ymm6
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm10[0,1,2,3],xmm12[4,5],xmm10[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm6
+; AVX1-ONLY-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = mem[3,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[2,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm11[0,1],xmm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = xmm7[0,1],mem[2,3],xmm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm11[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = xmm11[0,1],mem[2,3],xmm11[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm11, %xmm11
; AVX1-ONLY-NEXT: vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,1,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,7,6,7]
; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm15[0,1,2],xmm13[3,4,5],xmm15[6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm15[0,1,2],xmm11[3,4,5],xmm15[6,7]
; AVX1-ONLY-NEXT: vpshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,2,0]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,6,5]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5],xmm15[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm13, %ymm6
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5],xmm15[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm11, %ymm6
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3],xmm6[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3],xmm9[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[3,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[2,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm13[0,1],xmm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, (%rsp), %xmm11, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm11[0,1],mem[2,3],xmm11[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm13[5,6,7]
-; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,1],xmm8[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = mem[3,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[2,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm11[0,1],xmm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = xmm5[0,1],mem[2,3],xmm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm11[5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,1],xmm3[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vpshufd $212, (%rsp), %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,1,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,7,6,7]
-; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm15[0,1,2],xmm13[3,4,5],xmm15[6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm9[0,1,2,0]
+; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm15 = xmm15[2],xmm4[2],xmm15[3],xmm4[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm15[0,1,2],xmm11[3,4,5],xmm15[6,7]
+; AVX1-ONLY-NEXT: vpshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,2,0]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,6,5]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5],xmm15[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm13, %ymm6
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm7[0,1,2,3],mem[4,5],xmm7[6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[3,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[2,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm13[0,1],xmm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm12[0,1],xmm10[2,3],xmm12[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm13[5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,1,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,7,6,7]
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm15 = xmm15[2],xmm14[2],xmm15[3],xmm14[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm15[0,1,2],xmm13[3,4,5],xmm15[6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm5[0,1,2,0]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,6,5]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5],xmm15[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm13, %ymm3
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5],xmm15[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm11, %ymm3
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm12[0,1],xmm14[2,3],xmm12[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm6, %xmm2
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0,1,2,3],xmm9[4,5],xmm7[6,7]
+; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm9[0,1,2,3],mem[4,5],xmm9[6,7]
; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm8[3,1,2,3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm5[3,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[2,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm6[0,1],xmm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm10[0,1],xmm11[2,3],xmm10[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1],xmm13[2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm4[0,1,1,3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm13[0,1,1,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm5[0,1,2,0]
+; AVX1-ONLY-NEXT: vpshufd $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,0]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm2[6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $204, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm0[0,1],mem[2,3],xmm0[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vpblendw $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm0[2,3],mem[4,5],xmm0[6,7]
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,0,1,10,11,4,5,14,15,14,15,14,15,14,15]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm2
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm0[0,1,2,3],mem[4,5],xmm0[6,7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
+; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm8[0,1,2,3],mem[4,5],xmm8[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm6, %xmm6
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm1[0,1,2,3,4],xmm6[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm2 = [6,7,0,1,10,11,10,11,8,9,10,11,12,13,14,15]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm0[0,1,2,3],mem[4,5],xmm0[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = xmm1[0,1],mem[2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm1 = [6,7,0,1,10,11,10,11,8,9,10,11,12,13,14,15]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm8[0,1,2,3],mem[4,5],xmm8[6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[2,2,2,2,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2],xmm15[3,4,5],xmm13[6,7]
-; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,0,3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2],xmm15[3,4,5],xmm11[6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm14[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,5,6]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5],xmm15[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm13, %ymm6
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5],xmm15[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm11, %ymm6
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $204, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm0[0,1],mem[2,3],xmm0[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[u,u,0,1,10,11,4,5,14,15,14,15,14,15,14,15]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm13[0],xmm6[1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, (%rsp), %xmm13, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm13[0,1,2,3],mem[4,5],xmm13[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm13[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,1],xmm0[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm0[0,1,2,3],mem[4,5],xmm0[6,7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[2,2,2,2,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2],xmm15[3,4,5],xmm13[6,7]
-; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,5,6]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5],xmm15[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm13, %ymm6
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm9[0,1],xmm7[2,3],xmm9[4,5],xmm7[6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm12[0,1],xmm10[2,3],xmm12[4,5],xmm10[6,7]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = <u,u,0,1,10,11,4,5,14,15,14,15,14,15,14,15>
; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm8, %xmm13
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm13[0],xmm6[1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm12[0,1,2,3],xmm14[4,5],xmm12[6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm13[5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm9
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm15 = xmm10[0,1,2,3],xmm11[4,5],xmm10[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm8, %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm11[0],xmm6[1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm7[0,1,2,3],xmm8[4,5],xmm7[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm11[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm12[0,1],xmm7[2,3],xmm12[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm15 = xmm10[0,1,2,3],xmm14[4,5],xmm10[6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[2,2,2,2,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,7,4,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2],xmm15[3,4,5],xmm13[6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm5[0,1,0,3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2],xmm15[3,4,5],xmm11[6,7]
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,5,6]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5],xmm15[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm13, %ymm3
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5],xmm15[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm11, %ymm6
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpblendw $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1],xmm9[2,3],mem[4,5],xmm9[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm6
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm5, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm9
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm11[0],xmm6[1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm3[0,1,2,3],xmm4[4,5],xmm3[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm11[5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm4[0,1],xmm13[2,3],xmm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm3[0,1,2,3],xmm4[4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm6, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1],xmm12[2,3],mem[4,5],xmm12[6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm15 = xmm3[0,1,2,3],xmm5[4,5],xmm3[6,7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[2,2,2,2,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,7,4,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2],xmm15[3,4,5],xmm11[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm13[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,5,6]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5],xmm15[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm11, %ymm6
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm6[0,1,2,3],mem[4,5],xmm6[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm6, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1],xmm6[2,3],mem[4,5],xmm6[6,7]
; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm14, %xmm6
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm6, %xmm6
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm6[0],xmm0[1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm7[0,1],xmm8[2,3],xmm7[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm5[0,1,2,3],xmm9[4,5],xmm5[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $12, (%rsp), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1],mem[2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1,2,3],mem[4,5],xmm2[6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,2,2,2,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,4,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3,4,5],xmm1[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm11[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,6]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm2[6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
@@ -5195,144 +5197,143 @@
; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[0,1,2,3],mem[4,5],xmm1[6,7]
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm0[0,1,2,3],mem[4,5],xmm0[6,7]
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[1,1,1,1]
; AVX1-ONLY-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0,2,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm2 = [8,9,2,3,12,13,12,13,8,9,12,13,12,13,14,15]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm6[3,4,5],xmm1[6,7]
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = [8,9,2,3,12,13,12,13,8,9,12,13,12,13,14,15]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm6[3,4,5],xmm2[6,7]
; AVX1-ONLY-NEXT: vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,1,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm6[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm6[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[3,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $232, (%rsp), %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,2,2,3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm8[0,2,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0,3,2,3]
; AVX1-ONLY-NEXT: vpblendw $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = xmm6[0,1,2],mem[3],xmm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm13[0],xmm6[1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm15[0],xmm6[1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm6[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3],xmm6[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,2,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm15[0],xmm13[0],xmm15[1],xmm13[1]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2],xmm13[3,4,5],xmm6[6,7]
-; AVX1-ONLY-NEXT: vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,1,1,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5],xmm13[6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm12[0,1,2,3],xmm7[4,5],xmm12[6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm10[1,1,1,1]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm11 = xmm14[0,2,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm11[0],xmm15[0],xmm11[1],xmm15[1]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm6
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2],xmm11[3,4,5],xmm6[6,7]
+; AVX1-ONLY-NEXT: vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,1,1,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5],xmm11[6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm6, %ymm1
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm3[3,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,2,1,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm10 = xmm4[0,2,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm10 = xmm10[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm10[0],xmm6[0],xmm10[1],xmm6[1]
-; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = mem[0,3,2,3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm12[3],xmm10[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm14[2,3,2,3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm10 = xmm13[0],xmm10[1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm10[0,1,2,3],xmm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm10 = xmm7[0,1,2,3],xmm8[4,5],xmm7[6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm12 = xmm5[1,1,1,1]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm9[0,2,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm10, %xmm10
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm12[3,4,5],xmm10[6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm12 = xmm11[0,1,1,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,5],xmm12[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm10, %ymm10
; AVX1-ONLY-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[3,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,2,1,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm12 = mem[0,2,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm12[0],xmm6[0],xmm12[1],xmm6[1]
-; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm12 = mem[0,3,2,3]
-; AVX1-ONLY-NEXT: vpblendw $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm12[0,1,2],mem[3],xmm12[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0],xmm11[1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,2,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm11[0],xmm6[0],xmm11[1],xmm6[1]
+; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,3,2,3]
+; AVX1-ONLY-NEXT: vpblendw $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = xmm11[0,1,2],mem[3],xmm11[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0],xmm10[1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm9[0,1,2,3],xmm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm3[0,1,2,3],mem[4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm7, %xmm2
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,2,2,3]
+; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm4[0,1,2,3],mem[4,5],xmm4[6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[1,1,1,1]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm5[0,2,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,1,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm10, (%r9)
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm7, %xmm4
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4,5],xmm4[6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm13[0,1,1,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm4[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vpshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[3,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,2,1,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,2,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,3,2,3]
+; AVX1-ONLY-NEXT: vpblendw $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[0,1,2],mem[3],xmm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0],xmm5[1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2,3],xmm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $48, (%rsp), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[0,1,2,3],mem[4,5],xmm5[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm5, %xmm0
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,2,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm5[3,4,5],xmm0[6,7]
+; AVX1-ONLY-NEXT: vpshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,1,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm5[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%r9)
; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%r9)
; AVX1-ONLY-NEXT: addq $1032, %rsp # imm = 0x408
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
@@ -5815,8 +5816,8 @@
; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%r8)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%r8)
-; AVX2-SLOW-NEXT: vmovdqa %ymm2, 64(%r9)
; AVX2-SLOW-NEXT: vmovdqa %ymm3, (%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, 64(%r9)
; AVX2-SLOW-NEXT: vmovdqa %ymm1, 96(%r9)
; AVX2-SLOW-NEXT: vmovdqa %ymm0, 32(%r9)
; AVX2-SLOW-NEXT: addq $1064, %rsp # imm = 0x428
@@ -5825,178 +5826,171 @@
;
; AVX2-FAST-LABEL: load_i16_stride5_vf64:
; AVX2-FAST: # %bb.0:
-; AVX2-FAST-NEXT: subq $1032, %rsp # imm = 0x408
+; AVX2-FAST-NEXT: subq $1000, %rsp # imm = 0x3E8
; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm6
-; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %ymm7
-; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %ymm8
+; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %ymm5
; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm9
; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %ymm10
-; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm14
-; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm4
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm15
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm15[0],ymm1[1],ymm15[2,3],ymm1[4],ymm15[5],ymm1[6],ymm15[7,8],ymm1[9],ymm15[10,11],ymm1[12],ymm15[13],ymm1[14],ymm15[15]
-; AVX2-FAST-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm12
-; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm12
+; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm14
+; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %ymm13
+; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm13[1],ymm0[2,3],ymm13[4],ymm0[5],ymm13[6],ymm0[7,8],ymm13[9],ymm0[10,11],ymm13[12],ymm0[13],ymm13[14],ymm0[15]
+; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3],xmm0[4,5],xmm1[6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm14[0],ymm4[1,2],ymm14[3],ymm4[4],ymm14[5],ymm4[6,7],ymm14[8],ymm4[9,10],ymm14[11],ymm4[12],ymm14[13],ymm4[14,15]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm12[0],ymm14[1,2],ymm12[3],ymm14[4],ymm12[5],ymm14[6,7],ymm12[8],ymm14[9,10],ymm12[11],ymm14[12],ymm12[13],ymm14[14,15]
; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [1,3,0,2,4,6,1,3]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm2, %ymm3
-; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm5
+; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm11
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,1,6,7,8,9,14,15,4,5,14,15,4,5,2,3,16,17,22,23,24,25,30,31,20,21,30,31,20,21,18,19]
; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm3
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
; AVX2-FAST-NEXT: vpblendvb %ymm4, %ymm0, %ymm3, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm10[0],ymm9[1],ymm10[2,3],ymm9[4],ymm10[5],ymm9[6],ymm10[7,8],ymm9[9],ymm10[10,11],ymm9[12],ymm10[13],ymm9[14],ymm10[15]
-; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm13
+; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4,5],xmm3[6,7]
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm7[0],ymm8[1,2],ymm7[3],ymm8[4],ymm7[5],ymm8[6,7],ymm7[8],ymm8[9,10],ymm7[11],ymm8[12],ymm7[13],ymm8[14,15]
-; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm7[0],ymm5[1,2],ymm7[3],ymm5[4],ymm7[5],ymm5[6,7],ymm7[8],ymm5[9,10],ymm7[11],ymm5[12],ymm7[13],ymm5[14,15]
+; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm7, %ymm15
; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd %ymm3, %ymm5, %ymm3
-; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm11
+; AVX2-FAST-NEXT: vpermd %ymm3, %ymm11, %ymm3
+; AVX2-FAST-NEXT: vmovdqa %ymm11, %ymm7
; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm3
; AVX2-FAST-NEXT: vpblendvb %ymm4, %ymm0, %ymm3, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm5
-; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm0[1],ymm6[2,3],ymm0[4],ymm6[5],ymm0[6],ymm6[7,8],ymm0[9],ymm6[10,11],ymm0[12],ymm6[13],ymm0[14],ymm6[15]
+; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm3[1],ymm6[2,3],ymm3[4],ymm6[5],ymm3[6],ymm6[7,8],ymm3[9],ymm6[10,11],ymm3[12],ymm6[13],ymm3[14],ymm6[15]
+; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm8
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm11
+; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4,5],xmm3[6,7]
-; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0],ymm3[1,2],ymm4[3],ymm3[4],ymm4[5],ymm3[6,7],ymm4[8],ymm3[9,10],ymm4[11],ymm3[12],ymm4[13],ymm3[14,15]
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd %ymm3, %ymm11, %ymm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm6[0],ymm3[1,2],ymm6[3],ymm3[4],ymm6[5],ymm3[6,7],ymm6[8],ymm3[9,10],ymm6[11],ymm3[12],ymm6[13],ymm3[14,15]
+; AVX2-FAST-NEXT: vmovdqa %ymm7, %ymm6
+; AVX2-FAST-NEXT: vpermd %ymm3, %ymm7, %ymm3
; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm3
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm0, %ymm3, %ymm0
+; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm7
+; AVX2-FAST-NEXT: vpblendvb %ymm4, %ymm0, %ymm3, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm5
+; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm4
; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3],ymm5[4],ymm0[5],ymm5[6],ymm0[7,8],ymm5[9],ymm0[10,11],ymm5[12],ymm0[13],ymm5[14],ymm0[15]
-; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3],ymm4[4],ymm0[5],ymm4[6],ymm0[7,8],ymm4[9],ymm0[10,11],ymm4[12],ymm0[13],ymm4[14],ymm0[15]
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3],xmm0[4,5],xmm3[6,7]
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %ymm6
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3],ymm6[4],ymm1[5],ymm6[6,7],ymm1[8],ymm6[9,10],ymm1[11],ymm6[12],ymm1[13],ymm6[14,15]
-; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm9
-; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd %ymm1, %ymm11, %ymm1
+; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0],ymm1[1,2],ymm3[3],ymm1[4],ymm3[5],ymm1[6,7],ymm3[8],ymm1[9,10],ymm3[11],ymm1[12],ymm3[13],ymm1[14,15]
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermd %ymm1, %ymm6, %ymm1
; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm11 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
-; AVX2-FAST-NEXT: vpblendvb %ymm11, %ymm0, %ymm1, %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm12[0],ymm15[1],ymm12[2],ymm15[3],ymm12[4,5],ymm15[6],ymm12[7,8],ymm15[9],ymm12[10],ymm15[11],ymm12[12,13],ymm15[14],ymm12[15]
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm0, %ymm1, %ymm6
+; AVX2-FAST-NEXT: vpblendw $74, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm13[0],mem[1],ymm13[2],mem[3],ymm13[4,5],mem[6],ymm13[7,8],mem[9],ymm13[10],mem[11],ymm13[12,13],mem[14],ymm13[15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6],xmm1[7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = [2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
-; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm0, %xmm1
-; AVX2-FAST-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[0],ymm14[1],mem[2,3],ymm14[4],mem[5],ymm14[6],mem[7,8],ymm14[9],mem[10,11],ymm14[12],mem[13],ymm14[14],mem[15]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <2,u,u,u,4,7,1,6>
-; AVX2-FAST-NEXT: vpermd %ymm0, %ymm3, %ymm2
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm13 = [2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
+; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm0, %xmm1
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm14[0],ymm12[1],ymm14[2,3],ymm12[4],ymm14[5],ymm12[6],ymm14[7,8],ymm12[9],ymm14[10,11],ymm12[12],ymm14[13],ymm12[14],ymm14[15]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = <2,u,u,u,4,7,1,6>
+; AVX2-FAST-NEXT: vpermd %ymm0, %ymm12, %ymm2
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [2,3,4,5,10,11,0,1,14,15,2,3,12,13,0,1,18,19,20,21,26,27,16,17,30,31,18,19,28,29,16,17]
; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vpblendvb %ymm11, %ymm1, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm13[0],ymm10[1],ymm13[2],ymm10[3],ymm13[4,5],ymm10[6],ymm13[7,8],ymm10[9],ymm13[10],ymm10[11],ymm13[12,13],ymm10[14],ymm13[15]
-; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm13
-; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm1, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm9[0],ymm10[1],ymm9[2],ymm10[3],ymm9[4,5],ymm10[6],ymm9[7,8],ymm10[9],ymm9[10],ymm10[11],ymm9[12,13],ymm10[14],ymm9[15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm10
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm10[2,3],xmm1[4,5,6],xmm10[7]
-; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm1, %xmm1
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5],ymm7[6],ymm8[7,8],ymm7[9],ymm8[10,11],ymm7[12],ymm8[13],ymm7[14],ymm8[15]
-; AVX2-FAST-NEXT: vpermd %ymm10, %ymm3, %ymm10
+; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm5[0],ymm15[1],ymm5[2,3],ymm15[4],ymm5[5],ymm15[6],ymm5[7,8],ymm15[9],ymm5[10,11],ymm15[12],ymm5[13],ymm15[14],ymm5[15]
+; AVX2-FAST-NEXT: vpermd %ymm10, %ymm12, %ymm10
; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm10, %ymm10
-; AVX2-FAST-NEXT: vmovdqa %ymm11, %ymm7
-; AVX2-FAST-NEXT: vpblendvb %ymm11, %ymm1, %ymm10, %ymm12
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm14[0],ymm15[1],ymm14[2],ymm15[3],ymm14[4,5],ymm15[6],ymm14[7,8],ymm15[9],ymm14[10],ymm15[11],ymm14[12,13],ymm15[14],ymm14[15]
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm1, %ymm10, %ymm1
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm8[0],ymm11[1],ymm8[2],ymm11[3],ymm8[4,5],ymm11[6],ymm8[7,8],ymm11[9],ymm8[10],ymm11[11],ymm8[12,13],ymm11[14],ymm8[15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1],xmm11[2,3],xmm10[4,5,6],xmm11[7]
-; AVX2-FAST-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm11 = mem[0],ymm4[1],mem[2,3],ymm4[4],mem[5],ymm4[6],mem[7,8],ymm4[9],mem[10,11],ymm4[12],mem[13],ymm4[14],mem[15]
-; AVX2-FAST-NEXT: vpermd %ymm11, %ymm3, %ymm11
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm14[0],ymm15[1],ymm14[2,3],ymm15[4],ymm14[5],ymm15[6],ymm14[7,8],ymm15[9],ymm14[10,11],ymm15[12],ymm14[13],ymm15[14],ymm14[15]
+; AVX2-FAST-NEXT: vpermd %ymm11, %ymm12, %ymm11
; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm11, %ymm11
-; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm10, %xmm10
+; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm10, %xmm10
; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm10, %ymm11, %ymm10
-; AVX2-FAST-NEXT: vpblendw $74, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm11 = ymm5[0],mem[1],ymm5[2],mem[3],ymm5[4,5],mem[6],ymm5[7,8],mem[9],ymm5[10],mem[11],ymm5[12,13],mem[14],ymm5[15]
+; AVX2-FAST-NEXT: vpblendw $74, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm11 = ymm4[0],mem[1],ymm4[2],mem[3],ymm4[4,5],mem[6],ymm4[7,8],mem[9],ymm4[10],mem[11],ymm4[12,13],mem[14],ymm4[15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm11, %xmm8
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm8 = xmm11[0,1],xmm8[2,3],xmm11[4,5,6],xmm8[7]
-; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm8, %xmm4
-; AVX2-FAST-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm8 = ymm9[0],mem[1],ymm9[2,3],mem[4],ymm9[5],mem[6],ymm9[7,8],mem[9],ymm9[10,11],mem[12],ymm9[13],mem[14],ymm9[15]
-; AVX2-FAST-NEXT: vpermd %ymm8, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm8, %xmm4
+; AVX2-FAST-NEXT: vpblendw $173, (%rsp), %ymm3, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0],ymm3[1],mem[2,3],ymm3[4],mem[5],ymm3[6],mem[7,8],ymm3[9],mem[10,11],ymm3[12],mem[13],ymm3[14],mem[15]
+; AVX2-FAST-NEXT: vpermd %ymm8, %ymm12, %ymm3
; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm3, %ymm0
; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm4, %ymm0, %ymm11
-; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [0,3,1,3,0,3,5,7]
-; AVX2-FAST-NEXT: vpermd %ymm1, %ymm4, %ymm3
+; AVX2-FAST-NEXT: vpermd %ymm0, %ymm4, %ymm3
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm9 = [0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27]
; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm3, %ymm3
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm0[0,1,2,3,4],ymm3[5,6,7],ymm0[8,9,10,11,12],ymm3[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm8[0,1,2,3,4],ymm3[5,6,7],ymm8[8,9,10,11,12],ymm3[13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %ymm5
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm4, %ymm8
+; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %ymm13
+; AVX2-FAST-NEXT: vpermd %ymm13, %ymm4, %ymm8
; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm8, %ymm8
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm0[0,1,2,3,4],ymm8[5,6,7],ymm0[8,9,10,11,12],ymm8[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm0
-; AVX2-FAST-NEXT: vpermd %ymm0, %ymm4, %ymm7
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm7, %ymm7
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm3[0,1,2,3,4],ymm7[5,6,7],ymm3[8,9,10,11,12],ymm7[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm3[0,1,2,3,4],ymm8[5,6,7],ymm3[8,9,10,11,12],ymm8[13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm12
+; AVX2-FAST-NEXT: vpermd %ymm12, %ymm4, %ymm8
+; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm8, %ymm8
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm3[0,1,2,3,4],ymm8[5,6,7],ymm3[8,9,10,11,12],ymm8[13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 448(%rdi), %ymm8
; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm8, %ymm4, %ymm4
; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm3[0,1,2,3,4],ymm4[5,6,7],ymm3[8,9,10,11,12],ymm4[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm6[0,1,2,3,4],ymm4[5,6,7],ymm6[8,9,10,11,12],ymm4[13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [1,3,2,3,1,3,6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm4
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd %ymm1, %ymm6, %ymm7
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermd %ymm0, %ymm6, %ymm7
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm3 = [0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25]
; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm7, %ymm7
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm2[0,1,2,3,4],ymm7[5,6,7],ymm2[8,9,10,11,12],ymm7[13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm6, %ymm2
-; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermd %ymm13, %ymm6, %ymm2
; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm12[0,1,2,3,4],ymm2[5,6,7],ymm12[8,9,10,11,12],ymm2[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1,2,3,4],ymm2[5,6,7],ymm1[8,9,10,11,12],ymm2[13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd %ymm0, %ymm6, %ymm1
-; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm7
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermd %ymm12, %ymm6, %ymm1
+; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm4
+; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm10[0,1,2,3,4],ymm1[5,6,7],ymm10[8,9,10,11,12],ymm1[13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm1[4,5,6,7]
@@ -6006,71 +6000,70 @@
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm11[0,1,2,3,4],ymm1[5,6,7],ymm11[8,9,10,11,12],ymm1[13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw $107, (%rsp), %ymm3, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1],ymm3[2],mem[3],ymm3[4],mem[5,6],ymm3[7],mem[8,9],ymm3[10],mem[11],ymm3[12],mem[13,14],ymm3[15]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm5[0,1],ymm12[2],ymm5[3],ymm12[4],ymm5[5,6],ymm12[7],ymm5[8,9],ymm12[10],ymm5[11],ymm12[12],ymm5[13,14],ymm12[15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3,4],xmm1[5,6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = <4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm1, %xmm1
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = ymm12[0],mem[1],ymm12[2,3],mem[4],ymm12[5],mem[6],ymm12[7,8],mem[9],ymm12[10,11],mem[12],ymm12[13],mem[14],ymm12[15]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <0,2,u,u,5,7,2,4>
-; AVX2-FAST-NEXT: vpermd %ymm6, %ymm0, %ymm10
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = <4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0],ymm2[1],mem[2,3],ymm2[4],mem[5],ymm2[6],mem[7,8],ymm2[9],mem[10,11],ymm2[12],mem[13],ymm2[14],mem[15]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <0,2,u,u,5,7,2,4>
+; AVX2-FAST-NEXT: vpermd %ymm6, %ymm2, %ymm10
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,u,u,u,u,0,1,6,7,16,17,22,23,24,25,30,31,20,21,22,23,16,17,22,23>
; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm10, %ymm10
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm1[0,1,2],ymm10[3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [1,4,6,0,1,4,6,0]
; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm11
+; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm11, %ymm11
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm10[0,1,2,3,4],ymm11[5,6,7],ymm10[8,9,10,11,12],ymm11[13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,1],ymm0[2],mem[3],ymm0[4],mem[5,6],ymm0[7],mem[8,9],ymm0[10],mem[11],ymm0[12],mem[13,14],ymm0[15]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm11[3,4],xmm10[5,6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm10, %xmm10
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm11 = mem[0],ymm3[1],mem[2,3],ymm3[4],mem[5],ymm3[6],mem[7,8],ymm3[9],mem[10,11],ymm3[12],mem[13],ymm3[14],mem[15]
+; AVX2-FAST-NEXT: vpermd %ymm11, %ymm2, %ymm11
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm11, %ymm11
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermd %ymm13, %ymm1, %ymm11
+; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm11, %ymm11
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm10[0,1,2,3,4],ymm11[5,6,7],ymm10[8,9,10,11,12],ymm11[13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm10[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = ymm8[0,1],mem[2],ymm8[3],mem[4],ymm8[5,6],mem[7],ymm8[8,9],mem[10],ymm8[11],mem[12],ymm8[13,14],mem[15]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm11[3,4],xmm10[5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm15[0],ymm14[1],ymm15[2,3],ymm14[4],ymm15[5],ymm14[6],ymm15[7,8],ymm14[9],ymm15[10,11],ymm14[12],ymm15[13],ymm14[14],ymm15[15]
+; AVX2-FAST-NEXT: vpermd %ymm11, %ymm2, %ymm11
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm11, %ymm11
+; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm10, %xmm10
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm11
; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm11, %ymm11
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm10[0,1,2,3,4],ymm11[5,6,7],ymm10[8,9,10,11,12],ymm11[13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = mem[0,1],ymm13[2],mem[3],ymm13[4],mem[5,6],ymm13[7],mem[8,9],ymm13[10],mem[11],ymm13[12],mem[13,14],ymm13[15]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm11[3,4],xmm10[5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm10, %xmm10
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm11 = ymm13[0],mem[1],ymm13[2,3],mem[4],ymm13[5],mem[6],ymm13[7,8],mem[9],ymm13[10,11],mem[12],ymm13[13],mem[14],ymm13[15]
-; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm4
-; AVX2-FAST-NEXT: vpermd %ymm11, %ymm0, %ymm11
-; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm11, %ymm11
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm11
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm11, %ymm11
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm10[0,1,2,3,4],ymm11[5,6,7],ymm10[8,9,10,11,12],ymm11[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm14[0,1],ymm15[2],ymm14[3],ymm15[4],ymm14[5,6],ymm15[7],ymm14[8,9],ymm15[10],ymm14[11],ymm15[12],ymm14[13,14],ymm15[15]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm11[3,4],xmm10[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm14[0],ymm15[1],ymm14[2,3],ymm15[4],ymm14[5],ymm15[6],ymm14[7,8],ymm15[9],ymm14[10,11],ymm15[12],ymm14[13],ymm15[14],ymm14[15]
-; AVX2-FAST-NEXT: vpermd %ymm11, %ymm4, %ymm11
-; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm5
-; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm11, %ymm11
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm10, %xmm10
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm7, %ymm1, %ymm11
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm11, %ymm11
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm10[0,1,2,3,4],ymm11[5,6,7],ymm10[8,9,10,11,12],ymm11[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm4[0,1],ymm8[2],ymm4[3],ymm8[4],ymm4[5,6],ymm8[7],ymm4[8,9],ymm8[10],ymm4[11],ymm8[12],ymm4[13,14],ymm8[15]
+; AVX2-FAST-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = ymm4[0,1],mem[2],ymm4[3],mem[4],ymm4[5,6],mem[7],ymm4[8,9],mem[10],ymm4[11],mem[12],ymm4[13,14],mem[15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2],xmm11[3,4],xmm10[5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm10, %xmm7
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = ymm11[0],mem[1],ymm11[2,3],mem[4],ymm11[5],mem[6],ymm11[7,8],mem[9],ymm11[10,11],mem[12],ymm11[13],mem[14],ymm11[15]
-; AVX2-FAST-NEXT: vpermd %ymm10, %ymm5, %ymm2
+; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm10, %xmm7
+; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5],ymm0[6],mem[7,8],ymm0[9],mem[10,11],ymm0[12],mem[13],ymm0[14],mem[15]
+; AVX2-FAST-NEXT: vpermd %ymm10, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
@@ -6078,14 +6071,14 @@
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7],ymm2[8,9,10,11,12],ymm0[13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendw $214, (%rsp), %ymm3, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm3[0],mem[1,2],ymm3[3],mem[4],ymm3[5],mem[6,7],ymm3[8],mem[9,10],ymm3[11],mem[12],ymm3[13],mem[14,15]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm12[0],ymm5[1,2],ymm12[3],ymm5[4],ymm12[5],ymm5[6,7],ymm12[8],ymm5[9,10],ymm12[11],ymm5[12],ymm12[13],ymm5[14,15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm0[1],xmm1[2],xmm0[3]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = <6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm1, %xmm1
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm5[0],ymm12[1],ymm5[2],ymm12[3],ymm5[4,5],ymm12[6],ymm5[7,8],ymm12[9],ymm5[10],ymm12[11],ymm5[12,13],ymm12[14],ymm5[15]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm12[0],ymm11[1],ymm12[2],ymm11[3],ymm12[4,5],ymm11[6],ymm12[7,8],ymm11[9],ymm12[10],ymm11[11],ymm12[12,13],ymm11[14],ymm12[15]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <0,3,u,u,5,0,2,7>
; AVX2-FAST-NEXT: vpermd %ymm6, %ymm2, %ymm6
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,2,3,4,5,18,19,20,21,26,27,16,17,30,31,30,31,18,19,20,21>
@@ -6094,294 +6087,290 @@
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [2,4,7,0,2,4,7,0]
; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm9 = [0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25]
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm10, %ymm10
+; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm5 = [0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25]
+; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm10, %ymm10
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm6[0,1,2,3,4],ymm10[5,6,7],ymm6[8,9,10,11,12],ymm10[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw $41, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0],ymm3[1,2],mem[3],ymm3[4],mem[5],ymm3[6,7],mem[8],ymm3[9,10],mem[11],ymm3[12],mem[13],ymm3[14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm9[0],ymm14[1,2],ymm9[3],ymm14[4],ymm9[5],ymm14[6,7],ymm9[8],ymm14[9,10],ymm9[11],ymm14[12],ymm9[13],ymm14[14,15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm10
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm10[0],xmm6[1],xmm10[2],xmm6[3]
; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm6, %xmm6
-; AVX2-FAST-NEXT: vpblendw $181, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = mem[0],ymm13[1],mem[2],ymm13[3],mem[4,5],ymm13[6],mem[7,8],ymm13[9],mem[10],ymm13[11],mem[12,13],ymm13[14],mem[15]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm3[0],ymm15[1],ymm3[2],ymm15[3],ymm3[4,5],ymm15[6],ymm3[7,8],ymm15[9],ymm3[10],ymm15[11],ymm3[12,13],ymm15[14],ymm3[15]
; AVX2-FAST-NEXT: vpermd %ymm10, %ymm2, %ymm10
; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm10, %ymm10
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2],ymm10[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm13, %ymm3
+; AVX2-FAST-NEXT: vpermd %ymm13, %ymm1, %ymm10
+; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm10, %ymm10
+; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm13
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm6[0,1,2,3,4],ymm10[5,6,7],ymm6[8,9,10,11,12],ymm10[13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendw $41, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0],ymm8[1,2],mem[3],ymm8[4],mem[5],ymm8[6,7],mem[8],ymm8[9,10],mem[11],ymm8[12],mem[13],ymm8[14,15]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm10
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm10[0],xmm6[1],xmm10[2],xmm6[3]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw $181, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0],ymm8[1],mem[2],ymm8[3],mem[4,5],ymm8[6],mem[7,8],ymm8[9],mem[10],ymm8[11],mem[12,13],ymm8[14],mem[15]
+; AVX2-FAST-NEXT: vpermd %ymm10, %ymm2, %ymm10
+; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm10, %ymm10
+; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm6, %xmm6
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2],ymm10[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm10, %ymm10
-; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm12
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm10, %ymm10
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm6[0,1,2,3,4],ymm10[5,6,7],ymm6[8,9,10,11,12],ymm10[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw $41, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0],ymm6[1,2],mem[3],ymm6[4],mem[5],ymm6[6,7],mem[8],ymm6[9,10],mem[11],ymm6[12],mem[13],ymm6[14,15]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm10
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm10[0],xmm6[1],xmm10[2],xmm6[3]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm15[0],ymm14[1],ymm15[2],ymm14[3],ymm15[4,5],ymm14[6],ymm15[7,8],ymm14[9],ymm15[10],ymm14[11],ymm15[12,13],ymm14[14],ymm15[15]
-; AVX2-FAST-NEXT: vpermd %ymm10, %ymm2, %ymm10
-; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm10, %ymm10
-; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm6, %xmm6
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermd %ymm9, %ymm1, %ymm10
-; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm10, %ymm10
-; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm14
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm10 = ymm6[0,1,2,3,4],ymm10[5,6,7],ymm6[8,9,10,11,12],ymm10[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm8[0],ymm4[1,2],ymm8[3],ymm4[4],ymm8[5],ymm4[6,7],ymm8[8],ymm4[9,10],ymm8[11],ymm4[12],ymm8[13],ymm4[14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm5[0],ymm4[1,2],ymm5[3],ymm4[4],ymm5[5],ymm4[6,7],ymm5[8],ymm4[9,10],ymm5[11],ymm4[12],ymm5[13],ymm4[14,15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm10
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm10[0],xmm6[1],xmm10[2],xmm6[3]
; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm6, %xmm0
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm12[0],ymm11[1],ymm12[2],ymm11[3],ymm12[4,5],ymm11[6],ymm12[7,8],ymm11[9],ymm12[10],ymm11[11],ymm12[12,13],ymm11[14],ymm12[15]
+; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm4[0],ymm10[1],ymm4[2],ymm10[3],ymm4[4,5],ymm10[6],ymm4[7,8],ymm10[9],ymm4[10],ymm10[11],ymm4[12,13],ymm10[14],ymm4[15]
; AVX2-FAST-NEXT: vpermd %ymm6, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermd %ymm13, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermd %ymm8, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw $173, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5],ymm0[6],mem[7,8],ymm0[9],mem[10,11],ymm0[12],mem[13],ymm0[14],mem[15]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm0[3,4],xmm1[5,6,7]
-; AVX2-FAST-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm5[0,1],mem[2],ymm5[3],mem[4],ymm5[5,6],mem[7],ymm5[8,9],mem[10],ymm5[11],mem[12],ymm5[13,14],mem[15]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm1, %xmm1
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <1,3,u,u,6,0,3,5>
-; AVX2-FAST-NEXT: vpermd %ymm2, %ymm4, %ymm2
-; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm6 = [16,17,22,23,24,25,30,31,0,0,0,0,0,1,6,7,16,17,22,23,24,25,30,31,0,0,0,0,0,1,6,7]
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5],ymm0[6],mem[7,8],ymm0[9],mem[10,11],ymm0[12],mem[13],ymm0[14],mem[15]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4],xmm2[5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm12[0,1],ymm11[2],ymm12[3],ymm11[4],ymm12[5,6],ymm11[7],ymm12[8,9],ymm11[10],ymm12[11],ymm11[12],ymm12[13,14],ymm11[15]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm12 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = <1,3,u,u,6,0,3,5>
+; AVX2-FAST-NEXT: vpermd %ymm2, %ymm7, %ymm2
+; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm13 = [16,17,22,23,24,25,30,31,0,0,0,0,0,1,6,7,16,17,22,23,24,25,30,31,0,0,0,0,0,1,6,7]
+; AVX2-FAST-NEXT: # ymm13 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = [0,2,1,3,0,2,5,7]
-; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm5 = [16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31]
-; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [0,2,1,3,0,2,5,7]
+; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm0 = [16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31]
+; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm3[0],mem[1],ymm3[2,3],mem[4],ymm3[5],mem[6],ymm3[7,8],mem[9],ymm3[10,11],mem[12],ymm3[13],mem[14],ymm3[15]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm10
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm10[0,1,2],xmm2[3,4],xmm10[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = mem[0,1],ymm3[2],mem[3],ymm3[4],mem[5,6],ymm3[7],mem[8,9],ymm3[10],mem[11],ymm3[12],mem[13,14],ymm3[15]
-; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm2, %xmm2
-; AVX2-FAST-NEXT: vpermd %ymm10, %ymm4, %ymm10
-; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm10, %ymm10
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm14[0],ymm9[1],ymm14[2,3],ymm9[4],ymm14[5],ymm9[6],ymm14[7,8],ymm9[9],ymm14[10,11],ymm9[12],ymm14[13],ymm9[14],ymm14[15]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm11
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm11[0,1,2],xmm2[3,4],xmm11[5,6,7]
+; AVX2-FAST-NEXT: vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm11 = mem[0,1],ymm15[2],mem[3],ymm15[4],mem[5,6],ymm15[7],mem[8,9],ymm15[10],mem[11],ymm15[12],mem[13,14],ymm15[15]
+; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm2, %xmm2
+; AVX2-FAST-NEXT: vpermd %ymm11, %ymm7, %ymm11
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm11, %ymm11
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm11[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermd %ymm3, %ymm6, %ymm3
+; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm3, %ymm3
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm15[0,1],mem[2],ymm15[3],mem[4],ymm15[5,6],mem[7],ymm15[8,9],mem[10],ymm15[11],mem[12],ymm15[13,14],mem[15]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = ymm8[0],mem[1],ymm8[2,3],mem[4],ymm8[5],mem[6],ymm8[7,8],mem[9],ymm8[10,11],mem[12],ymm8[13],mem[14],ymm8[15]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm11
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm10 = xmm11[0,1,2],xmm10[3,4],xmm11[5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm3, %ymm4, %ymm3
-; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm10, %xmm10
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm9, %ymm7, %ymm8
-; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm8, %ymm8
+; AVX2-FAST-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0],ymm5[1],mem[2,3],ymm5[4],mem[5],ymm5[6],mem[7,8],ymm5[9],mem[10,11],ymm5[12],mem[13],ymm5[14],mem[15]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm11
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm11[0,1,2],xmm3[3,4],xmm11[5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm4[0,1],ymm10[2],ymm4[3],ymm10[4],ymm4[5,6],ymm10[7],ymm4[8,9],ymm10[10],ymm4[11],ymm10[12],ymm4[13,14],ymm10[15]
+; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm3, %xmm3
+; AVX2-FAST-NEXT: vpermd %ymm11, %ymm7, %ymm11
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm11, %ymm11
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm11[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermd %ymm8, %ymm6, %ymm8
+; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm8, %ymm8
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm8 = mem[0],ymm8[1],mem[2,3],ymm8[4],mem[5],ymm8[6],mem[7,8],ymm8[9],mem[10,11],ymm8[12],mem[13],ymm8[14],mem[15]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1],ymm4[2],mem[3],ymm4[4],mem[5,6],ymm4[7],mem[8,9],ymm4[10],mem[11],ymm4[12],mem[13,14],ymm4[15]
+; AVX2-FAST-NEXT: vpermd %ymm8, %ymm7, %ymm7
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm7, %ymm7
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = ymm4[0],mem[1],ymm4[2,3],mem[4],ymm4[5],mem[6],ymm4[7,8],mem[9],ymm4[10,11],mem[12],ymm4[13],mem[14],ymm4[15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm10
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm8 = xmm10[0,1,2],xmm8[3,4],xmm10[5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm8, %xmm0
-; AVX2-FAST-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm8 = ymm12[0,1],mem[2],ymm12[3],mem[4],ymm12[5,6],mem[7],ymm12[8,9],mem[10],ymm12[11],mem[12],ymm12[13,14],mem[15]
-; AVX2-FAST-NEXT: vpermd %ymm8, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm13, %ymm7, %ymm4
-; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 64(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, (%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 96(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 32(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 64(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, (%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 96(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 32(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 64(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, (%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 96(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 32(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 64(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, (%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 96(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 32(%r8)
-; AVX2-FAST-NEXT: vmovdqa %ymm0, 64(%r9)
-; AVX2-FAST-NEXT: vmovdqa %ymm3, (%r9)
+; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm8, %xmm4
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm6, %ymm5
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, (%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, (%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, (%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm0, 64(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm0, (%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm0, 96(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX2-FAST-NEXT: vmovdqa %ymm4, (%r9)
+; AVX2-FAST-NEXT: vmovdqa %ymm3, 64(%r9)
; AVX2-FAST-NEXT: vmovdqa %ymm2, 96(%r9)
; AVX2-FAST-NEXT: vmovdqa %ymm1, 32(%r9)
-; AVX2-FAST-NEXT: addq $1032, %rsp # imm = 0x408
+; AVX2-FAST-NEXT: addq $1000, %rsp # imm = 0x3E8
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
;
; AVX2-FAST-PERLANE-LABEL: load_i16_stride5_vf64:
; AVX2-FAST-PERLANE: # %bb.0:
-; AVX2-FAST-PERLANE-NEXT: subq $1032, %rsp # imm = 0x408
+; AVX2-FAST-PERLANE-NEXT: subq $1080, %rsp # imm = 0x438
; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %ymm14
; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3],ymm0[4],ymm1[5],ymm0[6,7],ymm1[8],ymm0[9,10],ymm1[11],ymm0[12],ymm1[13],ymm0[14,15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3],ymm1[4],ymm0[5],ymm1[6,7],ymm0[8],ymm1[9,10],ymm0[11],ymm1[12],ymm0[13],ymm1[14,15]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm1 = [0,1,10,11,4,5,14,15,8,9,10,11,4,5,6,7,16,17,26,27,20,21,30,31,24,25,26,27,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5],ymm2[6],ymm3[7,8],ymm2[9],ymm3[10,11],ymm2[12],ymm3[13],ymm2[14],ymm3[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm5[0],ymm2[1],ymm5[2,3],ymm2[4],ymm5[5],ymm2[6],ymm5[7,8],ymm2[9],ymm5[10,11],ymm2[12],ymm5[13],ymm2[14],ymm5[15]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm2[0],xmm3[1,2,3],xmm2[4,5],xmm3[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,10,11,4,5,14,15,8,9,2,3,12,13,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm3, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm3, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm15[0],ymm8[1,2],ymm15[3],ymm8[4],ymm15[5],ymm8[6,7],ymm15[8],ymm8[9,10],ymm15[11],ymm8[12],ymm15[13],ymm8[14,15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm7[0],ymm14[1,2],ymm7[3],ymm14[4],ymm7[5],ymm14[6,7],ymm7[8],ymm14[9,10],ymm7[11],ymm14[12],ymm7[13],ymm14[14,15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5],ymm0[6],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm5[0],ymm9[1],ymm5[2,3],ymm9[4],ymm5[5],ymm9[6],ymm5[7,8],ymm9[9],ymm5[10,11],ymm9[12],ymm5[13],ymm9[14],ymm5[15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1,2,3],xmm3[4,5],xmm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm6[0],ymm8[1],ymm6[2,3],ymm8[4],ymm6[5],ymm8[6],ymm6[7,8],ymm8[9],ymm6[10,11],ymm8[12],ymm6[13],ymm8[14],ymm6[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm6[1,2,3],xmm3[4,5],xmm6[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm3, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm3, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm12[0],ymm4[1,2],ymm12[3],ymm4[4],ymm12[5],ymm4[6,7],ymm12[8],ymm4[9,10],ymm12[11],ymm4[12],ymm12[13],ymm4[14,15]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5],ymm0[6],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm11
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm13
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm13[0],ymm11[1],ymm13[2,3],ymm11[4],ymm13[5],ymm11[6],ymm13[7,8],ymm11[9],ymm13[10,11],ymm11[12],ymm13[13],ymm11[14],ymm13[15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm11
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm11[0],ymm10[1],ymm11[2,3],ymm10[4],ymm11[5],ymm10[6],ymm11[7,8],ymm10[9],ymm11[10,11],ymm10[12],ymm11[13],ymm10[14],ymm11[15]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1,2,3],xmm3[4,5],xmm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm6[1,2,3],xmm3[4,5],xmm6[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm3, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm3, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm10
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm10[0],ymm0[1,2],ymm10[3],ymm0[4],ymm10[5],ymm0[6,7],ymm10[8],ymm0[9,10],ymm10[11],ymm0[12],ymm10[13],ymm0[14,15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm9[1,2],ymm5[3],ymm9[4],ymm5[5],ymm9[6,7],ymm5[8],ymm9[9,10],ymm5[11],ymm9[12],ymm5[13],ymm9[14,15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5],ymm0[6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm6[0],ymm8[1],ymm6[2,3],ymm8[4],ymm6[5],ymm8[6],ymm6[7,8],ymm8[9],ymm6[10,11],ymm8[12],ymm6[13],ymm8[14],ymm6[15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm7[0],ymm4[1],ymm7[2,3],ymm4[4],ymm7[5],ymm4[6],ymm7[7,8],ymm4[9],ymm7[10,11],ymm4[12],ymm7[13],ymm4[14],ymm7[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm3[1,2,3],xmm1[4,5],xmm3[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm1, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm1, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5],ymm0[6],mem[7,8],ymm0[9],mem[10,11],ymm0[12],mem[13],ymm0[14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5],mem[6],ymm0[7,8],mem[9],ymm0[10,11],mem[12],ymm0[13],mem[14],ymm0[15]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $181, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4,5],ymm1[6],mem[7,8],ymm1[9],mem[10],ymm1[11],mem[12,13],ymm1[14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $74, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2],mem[3],ymm1[4,5],mem[6],ymm1[7,8],mem[9],ymm1[10],mem[11],ymm1[12,13],mem[14],ymm1[15]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6],xmm2[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = [2,3,12,13,6,7,0,1,10,11,6,7,8,9,8,9,18,19,28,29,22,23,16,17,26,27,22,23,24,25,24,25]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = [2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm1, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm1, %ymm0, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0],ymm15[1],mem[2,3],ymm15[4],mem[5],ymm15[6],mem[7,8],ymm15[9],mem[10,11],ymm15[12],mem[13],ymm15[14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm6 = [2,3,12,13,6,7,0,1,10,11,4,5,14,15,10,11]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm14[0],ymm13[1],ymm14[2,3],ymm13[4],ymm14[5],ymm13[6],ymm14[7,8],ymm13[9],ymm14[10,11],ymm13[12],ymm14[13],ymm13[14],ymm14[15]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw $74, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm9[0],mem[1],ymm9[2],mem[3],ymm9[4,5],mem[6],ymm9[7,8],mem[9],ymm9[10],mem[11],ymm9[12,13],mem[14],ymm9[15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm9
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm9[2,3],xmm1[4,5,6],xmm9[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw $74, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm8[0],mem[1],ymm8[2],mem[3],ymm8[4,5],mem[6],ymm8[7,8],mem[9],ymm8[10],mem[11],ymm8[12,13],mem[14],ymm8[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm8[2,3],xmm1[4,5,6],xmm8[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm1, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm1, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm14[0],ymm12[1],ymm14[2,3],ymm12[4],ymm14[5],ymm12[6],ymm14[7,8],ymm12[9],ymm14[10,11],ymm12[12],ymm14[13],ymm12[14],ymm14[15]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm9 = ymm0[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm9[5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm9 = ymm11[0],ymm13[1],ymm11[2],ymm13[3],ymm11[4,5],ymm13[6],ymm11[7,8],ymm13[9],ymm11[10],ymm13[11],ymm11[12,13],ymm13[14],ymm11[15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm9, %xmm12
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm12[2,3],xmm9[4,5,6],xmm12[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0],ymm12[1],mem[2,3],ymm12[4],mem[5],ymm12[6],mem[7,8],ymm12[9],mem[10,11],ymm12[12],mem[13],ymm12[14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm8 = ymm0[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm8 = ymm10[0],ymm11[1],ymm10[2],ymm11[3],ymm10[4,5],ymm11[6],ymm10[7,8],ymm11[9],ymm10[10],ymm11[11],ymm10[12,13],ymm11[14],ymm10[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm8, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1],xmm12[2,3],xmm8[4,5,6],xmm12[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm9, %xmm9
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm9, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0],ymm10[1],mem[2,3],ymm10[4],mem[5],ymm10[6],mem[7,8],ymm10[9],mem[10,11],ymm10[12],mem[13],ymm10[14],mem[15]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm9[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm12[5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm9, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm9 = ymm8[0],ymm6[1],ymm8[2],ymm6[3],ymm8[4,5],ymm6[6],ymm8[7,8],ymm6[9],ymm8[10],ymm6[11],ymm8[12,13],ymm6[14],ymm8[15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm9, %xmm12
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm12[2,3],xmm9[4,5,6],xmm12[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm9, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm4, %ymm2, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 304(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm6 = xmm0[0],xmm2[1],xmm0[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, %xmm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm0, %xmm11
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm8, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm8, %ymm0, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm8 = ymm9[0],ymm5[1],ymm9[2,3],ymm5[4],ymm9[5],ymm5[6],ymm9[7,8],ymm5[9],ymm9[10,11],ymm5[12],ymm9[13],ymm5[14],ymm9[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm8[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm12[5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm8 = ymm4[0],ymm7[1],ymm4[2],ymm7[3],ymm4[4,5],ymm7[6],ymm4[7,8],ymm7[9],ymm4[10],ymm7[11],ymm4[12,13],ymm7[14],ymm4[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm8, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1],xmm12[2,3],xmm8[4,5,6],xmm12[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm8, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm6, %ymm2, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 304(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %xmm10
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm6 = xmm10[0],xmm1[1],xmm10[2,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = [6,7,2,3,4,5,6,7,6,7,2,3,12,13,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm6, %xmm6
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
@@ -6389,342 +6378,346 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm6 = ymm2[0,1,2,3,4],ymm6[5,6,7],ymm2[8,9,10,11,12],ymm6[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 624(%rdi), %xmm12
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm6 = xmm2[0],xmm12[1],xmm2[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, %xmm9
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm6, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 624(%rdi), %xmm13
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm12 = xmm7[0],xmm13[1],xmm7[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm12, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm6 = ymm2[0,1,2,3,4],ymm6[5,6,7],ymm2[8,9,10,11,12],ymm6[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm12 = ymm2[0,1,2,3,4],ymm12[5,6,7],ymm2[8,9,10,11,12],ymm12[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 464(%rdi), %xmm13
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 464(%rdi), %xmm5
; AVX2-FAST-PERLANE-NEXT: vmovdqa 448(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm6 = xmm2[0],xmm13[1],xmm2[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm6, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm6 = ymm2[0,1,2,3,4],ymm6[5,6,7],ymm2[8,9,10,11,12],ymm6[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 144(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm14 = xmm2[0],xmm6[1],xmm2[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm14 = xmm2[0],xmm5[1],xmm2[2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm5, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, %xmm6
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm14, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm14, %xmm14
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm14 = ymm2[0,1,2,3,4],ymm14[5,6,7],ymm2[8,9,10,11,12],ymm14[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 144(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm11 = xmm2[0],xmm5[1],xmm2[2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm11, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm10[0,1,2,3,4],ymm0[5,6,7],ymm10[8,9,10,11,12],ymm0[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6,7],ymm4[8,9,10,11,12],ymm0[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm11, %xmm14
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm11[0,1],xmm15[2],xmm11[3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm15, %xmm10
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm11 = [0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm10[0,1],xmm1[2],xmm10[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm9 = [0,1,2,3,4,5,6,7,8,9,4,5,14,15,8,9]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm0[5,6,7],ymm7[8,9,10,11,12],ymm0[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm9[0,1],xmm12[2],xmm9[3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0,1,2,3,4],ymm0[5,6,7],ymm5[8,9,10,11,12],ymm0[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm4[0,1],xmm13[2],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm7, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm13, %xmm14
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm4, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm7[0,1],xmm13[2],xmm7[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm6[0,1],xmm8[2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm0[5,6,7],ymm15[8,9,10,11,12],ymm0[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm5[2],xmm2[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7],ymm3[8,9,10,11,12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm6[2],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm4[1],ymm2[2,3],ymm4[4],ymm2[5],ymm4[6],ymm2[7,8],ymm4[9],ymm2[10,11],ymm4[12],ymm2[13],ymm4[14],ymm2[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm2[1],ymm5[2,3],ymm2[4],ymm5[5],ymm2[6],ymm5[7,8],ymm2[9],ymm5[10,11],ymm2[12],ymm5[13],ymm2[14],ymm5[15]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $148, (%rsp), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0,1],mem[2],ymm1[3],mem[4],ymm1[5,6],mem[7],ymm1[8,9],mem[10],ymm1[11],mem[12],ymm1[13,14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm6[0,1],ymm11[2],ymm6[3],ymm11[4],ymm6[5,6],ymm11[7],ymm6[8,9],ymm11[10],ymm6[11],ymm11[12],ymm6[13,14],ymm11[15]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3,4],xmm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,u,u,u,u,u,u,u,u,u,u,0,1,10,11,20,21,30,31,24,25,18,19,28,29,26,27,16,17,26,27>
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm15 = <4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm3, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm7 = <4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm3, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm7, %xmm13
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm7 = xmm10[0],xmm14[1],xmm10[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm7 = xmm4[0],xmm10[1],xmm4[2,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,6,7,4,5,6,7,8,9,6,7,0,1,10,11]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm7, %xmm7
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm7 = ymm3[0,1,2,3,4],ymm7[5,6,7],ymm3[8,9,10,11,12],ymm7[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0],ymm6[1],mem[2,3],ymm6[4],mem[5],ymm6[6],mem[7,8],ymm6[9],mem[10,11],ymm6[12],mem[13],ymm6[14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm8[0],mem[1],ymm8[2,3],mem[4],ymm8[5],mem[6],ymm8[7,8],mem[9],ymm8[10,11],mem[12],ymm8[13],mem[14],ymm8[15]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm7[6],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm8[0,1],mem[2],ymm8[3],mem[4],ymm8[5,6],mem[7],ymm8[8,9],mem[10],ymm8[11],mem[12],ymm8[13,14],mem[15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm7, %xmm11
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2],xmm11[3,4],xmm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm10[0,1],mem[2],ymm10[3],mem[4],ymm10[5,6],mem[7],ymm10[8,9],mem[10],ymm10[11],mem[12],ymm10[13,14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm7, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2],xmm9[3,4],xmm7[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm7, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm7, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm13, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm7 = xmm12[0],xmm9[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm7 = xmm14[0],xmm12[1],xmm14[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm7, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm7 = ymm3[0,1,2,3,4],ymm7[5,6,7],ymm3[8,9,10,11,12],ymm7[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm15[0],mem[1],ymm15[2,3],mem[4],ymm15[5],mem[6],ymm15[7,8],mem[9],ymm15[10,11],mem[12],ymm15[13],mem[14],ymm15[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm7[6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm7 = ymm12[0,1],ymm13[2],ymm12[3],ymm13[4],ymm12[5,6],ymm13[7],ymm12[8,9],ymm13[10],ymm12[11],ymm13[12],ymm12[13,14],ymm13[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm7, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2],xmm9[3,4],xmm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm3, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm7, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm7 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm7 = mem[0],xmm4[1],mem[2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm7, %xmm7
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm7 = ymm3[0,1,2,3,4],ymm7[5,6,7],ymm3[8,9,10,11,12],ymm7[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm14[0],mem[1],ymm14[2,3],mem[4],ymm14[5],mem[6],ymm14[7,8],mem[9],ymm14[10,11],mem[12],ymm14[13],mem[14],ymm14[15]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm7[6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm7 = ymm9[0,1],ymm12[2],ymm9[3],ymm12[4],ymm9[5,6],ymm12[7],ymm9[8,9],ymm12[10],ymm9[11],ymm12[12],ymm9[13,14],ymm12[15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm7, %xmm11
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2],xmm11[3,4],xmm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm7, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm7 = xmm13[0],xmm5[1],xmm13[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm7, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm7 = ymm3[0,1,2,3,4],ymm7[5,6,7],ymm3[8,9,10,11,12],ymm7[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm10[0],mem[1],ymm10[2,3],mem[4],ymm10[5],mem[6],ymm10[7,8],mem[9],ymm10[10,11],mem[12],ymm10[13],mem[14],ymm10[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm9[0],ymm14[1],ymm9[2,3],ymm14[4],ymm9[5],ymm14[6],ymm9[7,8],ymm14[9],ymm9[10,11],ymm14[12],ymm9[13],ymm14[14],ymm9[15]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm3[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm7[6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm13[0,1],mem[2],ymm13[3],mem[4],ymm13[5,6],mem[7],ymm13[8,9],mem[10],ymm13[11],mem[12],ymm13[13,14],mem[15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2],xmm7[3,4],xmm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm5, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm1[0,1],mem[2],ymm1[3],mem[4],ymm1[5,6],mem[7],ymm1[8,9],mem[10],ymm1[11],mem[12],ymm1[13,14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2],xmm7[3,4],xmm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm1 = xmm4[4,5,14,15,8,9,2,3,12,13,6,7,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm3 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm3 = xmm15[0],mem[1],xmm15[2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm3, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0],ymm2[1],ymm4[2],ymm2[3],ymm4[4,5],ymm2[6],ymm4[7,8],ymm2[9],ymm4[10],ymm2[11],ymm4[12,13],ymm2[14],ymm4[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm5[1],ymm2[2],ymm5[3],ymm2[4,5],ymm5[6],ymm2[7,8],ymm5[9],ymm2[10],ymm5[11],ymm2[12,13],ymm5[14],ymm2[15]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5],ymm1[6],ymm0[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $41, (%rsp), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0],ymm1[1,2],mem[3],ymm1[4],mem[5],ymm1[6,7],mem[8],ymm1[9,10],mem[11],ymm1[12],mem[13],ymm1[14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm11[0],ymm6[1,2],ymm11[3],ymm6[4],ymm11[5],ymm6[6,7],ymm11[8],ymm6[9,10],ymm11[11],ymm6[12],ymm11[13],ymm6[14,15]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2],xmm1[3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,2,3,12,13,22,23,16,17,26,27,20,21,30,31,30,31,18,19,28,29>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = <6,7,0,1,10,11,4,5,14,15,8,9,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm2[0,1],xmm4[2],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm1[0,1,2,3,4],ymm5[5,6,7],ymm1[8,9,10,11,12],ymm5[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw $74, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm6[0],mem[1],ymm6[2],mem[3],ymm6[4,5],mem[6],ymm6[7,8],mem[9],ymm6[10],mem[11],ymm6[12,13],mem[14],ymm6[15]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm1[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4],ymm1[5],ymm5[6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw $41, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0],ymm8[1,2],mem[3],ymm8[4],mem[5],ymm8[6,7],mem[8],ymm8[9,10],mem[11],ymm8[12],mem[13],ymm8[14,15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm11
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm11[0],xmm5[1],xmm11[2],xmm5[3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $4, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm5 = xmm5[0,1],mem[2],xmm5[3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm1[0,1,2,3,4],ymm5[5,6,7],ymm1[8,9,10,11,12],ymm5[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm5[0,1],xmm6[2],xmm5[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,2,3,4,5,6,7,8,9,8,9,2,3,12,13]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm1[0,1,2,3,4],ymm4[5,6,7],ymm1[8,9,10,11,12],ymm4[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw $181, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0],ymm14[1],mem[2],ymm14[3],mem[4,5],ymm14[6],mem[7,8],ymm14[9],mem[10],ymm14[11],mem[12,13],ymm14[14],mem[15]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm1[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4],ymm1[5],ymm5[6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm12[0],ymm9[1,2],ymm12[3],ymm9[4],ymm12[5],ymm9[6,7],ymm12[8],ymm9[9,10],ymm12[11],ymm9[12],ymm12[13],ymm9[14,15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm11
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm11[0],xmm5[1],xmm11[2],xmm5[3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $4, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm5 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm5 = xmm12[0,1],mem[2],xmm12[3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm1[0,1,2,3,4],ymm5[5,6,7],ymm1[8,9,10,11,12],ymm5[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw $181, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0],ymm8[1],mem[2],ymm8[3],mem[4,5],ymm8[6],mem[7,8],ymm8[9],mem[10],ymm8[11],mem[12,13],ymm8[14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4],ymm1[5],ymm4[6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw $41, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0],ymm10[1,2],mem[3],ymm10[4],mem[5],ymm10[6,7],mem[8],ymm10[9,10],mem[11],ymm10[12],mem[13],ymm10[14,15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $4, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm1[0,1,2,3,4],ymm4[5,6,7],ymm1[8,9,10,11,12],ymm4[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm8[0],ymm10[1],ymm8[2],ymm10[3],ymm8[4,5],ymm10[6],ymm8[7,8],ymm10[9],ymm8[10],ymm10[11],ymm8[12,13],ymm10[14],ymm8[15]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm1[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4],ymm1[5],ymm5[6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm11[0],ymm13[1,2],ymm11[3],ymm13[4],ymm11[5],ymm13[6,7],ymm11[8],ymm13[9,10],ymm11[11],ymm13[12],ymm11[13],ymm13[14,15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm5 = xmm7[0],xmm5[1],xmm7[2],xmm5[3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm5, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm8[0],ymm15[1],ymm8[2],ymm15[3],ymm8[4,5],ymm15[6],ymm8[7,8],ymm15[9],ymm8[10],ymm15[11],ymm8[12,13],ymm15[14],ymm8[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4],ymm1[5],ymm4[6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm13[0],ymm12[1,2],ymm13[3],ymm12[4],ymm13[5],ymm12[6,7],ymm13[8],ymm12[9,10],ymm13[11],ymm12[12],ymm13[13],ymm12[14,15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm13[0,1],xmm12[2],xmm13[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm1[0,1,2,3,4],ymm4[5,6,7],ymm1[8,9,10,11,12],ymm4[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm14[0],ymm9[1],ymm14[2],ymm9[3],ymm14[4,5],ymm9[6],ymm14[7,8],ymm9[9],ymm14[10],ymm9[11],ymm14[12,13],ymm9[14],ymm14[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4],ymm1[5],ymm4[6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $214, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm14[0],mem[1,2],ymm14[3],mem[4],ymm14[5],mem[6,7],ymm14[8],mem[9,10],ymm14[11],mem[12],ymm14[13],mem[14,15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm4, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm15, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm15[0,1],xmm7[2],xmm15[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm10[0,1],xmm7[2],xmm10[3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm15 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0,1],mem[2],ymm0[3],mem[4],ymm0[5,6],mem[7],ymm0[8,9],mem[10],ymm0[11],mem[12],ymm0[13,14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1],ymm0[2],mem[3],ymm0[4],mem[5,6],ymm0[7],mem[8,9],ymm0[10],mem[11],ymm0[12],mem[13,14],ymm0[15]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5],ymm0[6],mem[7,8],ymm0[9],mem[10,11],ymm0[12],mem[13],ymm0[14],mem[15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm3[0,1,2],xmm0[3,4],xmm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [24,25,18,19,28,29,22,23,0,0,0,0,4,5,14,15,24,25,18,19,28,29,22,23,0,0,0,0,4,5,14,15]
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm13 = [12,13,14,15,4,5,14,15,8,9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm2, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,0,1,10,11,8,9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm4, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0],ymm11[1],mem[2,3],ymm11[4],mem[5],ymm11[6],mem[7,8],ymm11[9],mem[10,11],ymm11[12],mem[13],ymm11[14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3,4],xmm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [24,25,18,19,28,29,22,23,0,0,0,0,4,5,14,15,24,25,18,19,28,29,22,23,0,0,0,0,4,5,14,15]
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm9 = [8,9,2,3,12,13,6,7,0,1,10,11,0,1,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm11 = [12,13,14,15,4,5,14,15,8,9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm5, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,0,1,10,11,8,9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm6, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2],mem[3],ymm1[4],mem[5,6],ymm1[7],mem[8,9],ymm1[10],mem[11],ymm1[12],mem[13,14],ymm1[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4],ymm1[5,6],ymm4[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1],ymm4[2],mem[3],ymm4[4],mem[5,6],ymm4[7],mem[8,9],ymm4[10],mem[11],ymm4[12],mem[13,14],ymm4[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0],ymm4[1],mem[2,3],ymm4[4],mem[5],ymm4[6],mem[7,8],ymm4[9],mem[10,11],ymm4[12],mem[13],ymm4[14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4],xmm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm8[0,1],ymm15[2],ymm8[3],ymm15[4],ymm8[5,6],ymm15[7],ymm8[8,9],ymm15[10],ymm8[11],ymm15[12],ymm8[13,14],ymm15[15]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6],ymm5[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0],ymm5[1],mem[2,3],ymm5[4],mem[5],ymm5[6],mem[7,8],ymm5[9],mem[10,11],ymm5[12],mem[13],ymm5[14],mem[15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm14
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm14[0,1,2],xmm5[3,4],xmm14[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm5, %xmm5
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm9, %xmm9
-; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm9[0],xmm5[0],xmm9[1],xmm5[1]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm13, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm12, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm6, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm7, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm6 = ymm8[0,1],ymm10[2],ymm8[3],ymm10[4],ymm8[5,6],ymm10[7],ymm8[8,9],ymm10[10],ymm8[11],ymm10[12],ymm8[13,14],ymm10[15]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm8 = ymm6[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4],ymm6[5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0],ymm11[1],mem[2,3],ymm11[4],mem[5],ymm11[6],mem[7,8],ymm11[9],mem[10,11],ymm11[12],mem[13],ymm11[14],mem[15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm8, %xmm9
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm8 = xmm9[0,1,2],xmm8[3,4],xmm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm6, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm8, %xmm8
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1],ymm6[2],mem[3],ymm6[4],mem[5,6],ymm6[7],mem[8,9],ymm6[10],mem[11],ymm6[12],mem[13,14],ymm6[15]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm8 = ymm6[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4],ymm6[5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm6, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $82, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3],mem[4],ymm6[5],mem[6],ymm6[7,8],mem[9],ymm6[10,11],mem[12],ymm6[13],mem[14],ymm6[15]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm8
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3,4],xmm8[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm6, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm12, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm6, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, (%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm15, (%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, 64(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, (%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, 96(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, 32(%r9)
-; AVX2-FAST-PERLANE-NEXT: addq $1032, %rsp # imm = 0x408
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm10, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm7, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm5[0,1],mem[2],ymm5[3],mem[4],ymm5[5,6],mem[7],ymm5[8,9],mem[10],ymm5[11],mem[12],ymm5[13,14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm5[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4],ymm5[5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm5, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendw $173, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0],ymm14[1],mem[2,3],ymm14[4],mem[5],ymm14[6],mem[7,8],ymm14[9],mem[10,11],ymm14[12],mem[13],ymm14[14],mem[15]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 64(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 96(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 64(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 96(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 64(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 96(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 64(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 96(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, (%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, 96(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, 32(%r9)
+; AVX2-FAST-PERLANE-NEXT: addq $1080, %rsp # imm = 0x438
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
index 9930293..e7ab8a8 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-6.ll
@@ -2209,28 +2209,29 @@
define void @load_i16_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
; SSE-LABEL: load_i16_stride6_vf32:
; SSE: # %bb.0:
-; SSE-NEXT: subq $488, %rsp # imm = 0x1E8
-; SSE-NEXT: movdqa 304(%rdi), %xmm5
-; SSE-NEXT: movdqa 320(%rdi), %xmm7
-; SSE-NEXT: movdqa 64(%rdi), %xmm2
-; SSE-NEXT: movdqa 80(%rdi), %xmm10
-; SSE-NEXT: movdqa (%rdi), %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 16(%rdi), %xmm4
+; SSE-NEXT: subq $472, %rsp # imm = 0x1D8
+; SSE-NEXT: movdqa 304(%rdi), %xmm8
+; SSE-NEXT: movdqa 320(%rdi), %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 240(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 32(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 48(%rdi), %xmm6
+; SSE-NEXT: movdqa 256(%rdi), %xmm2
+; SSE-NEXT: movdqa 272(%rdi), %xmm10
+; SSE-NEXT: movdqa 208(%rdi), %xmm6
; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 192(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 224(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm7 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT: movdqa %xmm7, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[0,1,0,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT: pand %xmm8, %xmm0
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm6[2],xmm0[3],xmm6[3]
+; SSE-NEXT: pand %xmm7, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm2, %xmm11
@@ -2239,7 +2240,7 @@
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm10, %xmm0
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[3,0]
-; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm10[0,0]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm10[2,3]
@@ -2247,24 +2248,23 @@
; SSE-NEXT: pslld $16, %xmm10
; SSE-NEXT: psrldq {{.*#+}} xmm11 = xmm11[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; SSE-NEXT: punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[0,3,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[3,1],xmm0[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm11[2,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[0,1,1,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm7, %xmm10
+; SSE-NEXT: movdqa %xmm7, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa 288(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: movdqa %xmm5, %xmm6
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT: pand %xmm8, %xmm0
-; SSE-NEXT: movdqa %xmm8, %xmm10
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,4,6,6,7]
+; SSE-NEXT: movdqa %xmm8, %xmm6
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; SSE-NEXT: pand %xmm10, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: movdqa 352(%rdi), %xmm4
@@ -2273,7 +2273,7 @@
; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; SSE-NEXT: movdqa %xmm0, %xmm3
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm4[3,0]
-; SSE-NEXT: movaps %xmm3, (%rsp) # 16-byte Spill
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm4, %xmm3
; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm0[0,0]
@@ -2290,24 +2290,25 @@
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 224(%rdi), %xmm15
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm15[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm8, %xmm3
-; SSE-NEXT: pandn %xmm0, %xmm3
-; SSE-NEXT: movdqa 208(%rdi), %xmm1
-; SSE-NEXT: movdqa 192(%rdi), %xmm0
+; SSE-NEXT: movdqa 32(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm10, %xmm3
+; SSE-NEXT: pandn %xmm0, %xmm3
+; SSE-NEXT: movdqa (%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 16(%rdi), %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT: movdqa %xmm1, %xmm14
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm8, %xmm0
+; SSE-NEXT: pand %xmm10, %xmm0
; SSE-NEXT: por %xmm3, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: movdqa 256(%rdi), %xmm4
-; SSE-NEXT: movdqa 272(%rdi), %xmm0
+; SSE-NEXT: movdqa 64(%rdi), %xmm4
+; SSE-NEXT: movdqa 80(%rdi), %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm4[2,2,3,3]
; SSE-NEXT: punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1],xmm13[2],xmm0[2],xmm13[3],xmm0[3]
; SSE-NEXT: movdqa %xmm0, %xmm3
@@ -2321,7 +2322,7 @@
; SSE-NEXT: pslld $16, %xmm0
; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
-; SSE-NEXT: movdqa 240(%rdi), %xmm0
+; SSE-NEXT: movdqa 48(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm0[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[0,1,0,2,4,5,6,7]
@@ -2331,15 +2332,14 @@
; SSE-NEXT: movdqa 128(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm8, %xmm3
+; SSE-NEXT: movdqa %xmm10, %xmm3
; SSE-NEXT: pandn %xmm0, %xmm3
-; SSE-NEXT: movdqa 112(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 112(%rdi), %xmm15
; SSE-NEXT: movdqa 96(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm0[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,4,6,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
; SSE-NEXT: pand %xmm10, %xmm0
; SSE-NEXT: por %xmm3, %xmm0
; SSE-NEXT: movdqa 160(%rdi), %xmm1
@@ -2383,10 +2383,10 @@
; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm6, %xmm0
; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: pshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = mem[0,1,2,3,5,7,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm9 = xmm7[0,1,2,3,5,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm0[2],xmm9[3],xmm0[3]
; SSE-NEXT: movdqa %xmm10, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT: pandn %xmm7, %xmm0
; SSE-NEXT: pand %xmm10, %xmm9
; SSE-NEXT: por %xmm0, %xmm9
@@ -2401,29 +2401,31 @@
; SSE-NEXT: # xmm14 = mem[0,1,2,3,5,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm0[2],xmm14[3],xmm0[3]
; SSE-NEXT: movdqa %xmm10, %xmm0
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pandn %xmm15, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: pandn %xmm9, %xmm0
; SSE-NEXT: pand %xmm10, %xmm14
; SSE-NEXT: por %xmm0, %xmm14
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm12[0,1,1,3,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[3,1],xmm0[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,1],xmm13[2,0]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm15, %xmm12
+; SSE-NEXT: movdqa %xmm15, %xmm0
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm8[0,1,2,3,5,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; SSE-NEXT: pand %xmm10, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: pandn %xmm12, %xmm10
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: pandn %xmm15, %xmm10
; SSE-NEXT: por %xmm2, %xmm10
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm4[0,1,1,3,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,1],xmm3[2,0]
; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm1, %xmm11
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[1,1,1,1]
; SSE-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,65535,65535,65535]
; SSE-NEXT: movdqa %xmm1, %xmm3
@@ -2437,7 +2439,7 @@
; SSE-NEXT: pand %xmm1, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: shufps $132, (%rsp), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
@@ -2448,13 +2450,12 @@
; SSE-NEXT: pandn %xmm3, %xmm5
; SSE-NEXT: pand %xmm0, %xmm4
; SSE-NEXT: por %xmm4, %xmm5
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm7, %xmm9
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm5, (%rsp) # 16-byte Spill
; SSE-NEXT: movdqa %xmm7, %xmm3
+; SSE-NEXT: movdqa %xmm7, %xmm13
; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm11[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm10[1,1,1,1]
; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm4[0]
; SSE-NEXT: movdqa %xmm1, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
@@ -2468,7 +2469,7 @@
; SSE-NEXT: pand %xmm1, %xmm5
; SSE-NEXT: por %xmm4, %xmm5
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: shufps $132, (%rsp), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = xmm4[0,1],mem[0,2]
; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,6,6,7]
@@ -2478,13 +2479,14 @@
; SSE-NEXT: pandn %xmm4, %xmm6
; SSE-NEXT: pand %xmm0, %xmm5
; SSE-NEXT: por %xmm5, %xmm6
-; SSE-NEXT: movdqa %xmm6, (%rsp) # 16-byte Spill
-; SSE-NEXT: psrldq {{.*#+}} xmm15 = xmm15[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,1,1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm15 = xmm15[0],xmm5[0]
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm9, %xmm4
+; SSE-NEXT: psrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm9[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm5[0]
; SSE-NEXT: movdqa %xmm1, %xmm5
-; SSE-NEXT: pandn %xmm15, %xmm5
+; SSE-NEXT: pandn %xmm4, %xmm5
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm6[0,0]
@@ -2501,23 +2503,22 @@
; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,5,4]
-; SSE-NEXT: movdqa %xmm0, %xmm8
-; SSE-NEXT: pandn %xmm5, %xmm8
+; SSE-NEXT: movdqa %xmm0, %xmm7
+; SSE-NEXT: pandn %xmm5, %xmm7
; SSE-NEXT: pand %xmm0, %xmm6
-; SSE-NEXT: por %xmm6, %xmm8
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm12, %xmm15
-; SSE-NEXT: movdqa %xmm12, %xmm5
+; SSE-NEXT: por %xmm6, %xmm7
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm15, %xmm5
; SSE-NEXT: psrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm10[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm7[1,1,1,1]
; SSE-NEXT: punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm6[0]
; SSE-NEXT: movdqa %xmm1, %xmm6
; SSE-NEXT: pandn %xmm5, %xmm6
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,0],xmm8[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm8[2,3]
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,0],xmm12[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm12[2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm5[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm12[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm12[1,0,2,3,4,5,6,7]
@@ -2535,11 +2536,10 @@
; SSE-NEXT: pand %xmm0, %xmm12
; SSE-NEXT: por %xmm12, %xmm8
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm12
-; SSE-NEXT: movdqa %xmm13, %xmm6
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm12, %xmm6
; SSE-NEXT: psrlq $48, %xmm6
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = mem[2,2,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm11[2,2,3,3]
; SSE-NEXT: punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm6[0]
; SSE-NEXT: movdqa %xmm1, %xmm6
; SSE-NEXT: pandn %xmm8, %xmm6
@@ -2556,10 +2556,9 @@
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm2, %xmm8
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm2
-; SSE-NEXT: psrlq $48, %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm9[2,2,3,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm2[0]
+; SSE-NEXT: psrlq $48, %xmm10
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm13[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm10[0]
; SSE-NEXT: movdqa %xmm1, %xmm2
; SSE-NEXT: pandn %xmm6, %xmm2
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
@@ -2575,10 +2574,10 @@
; SSE-NEXT: pand %xmm0, %xmm3
; SSE-NEXT: por %xmm3, %xmm6
; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrlq $48, %xmm7
+; SSE-NEXT: psrlq $48, %xmm9
; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[2,2,3,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm7[0]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm9[0]
; SSE-NEXT: movdqa %xmm1, %xmm2
; SSE-NEXT: pandn %xmm3, %xmm2
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm4[3,1,2,3,4,5,6,7]
@@ -2589,15 +2588,13 @@
; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[0,1,2,3,7,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
-; SSE-NEXT: movdqa %xmm0, %xmm13
-; SSE-NEXT: pandn %xmm2, %xmm13
+; SSE-NEXT: movdqa %xmm0, %xmm11
+; SSE-NEXT: pandn %xmm2, %xmm11
; SSE-NEXT: pand %xmm0, %xmm3
-; SSE-NEXT: por %xmm3, %xmm13
-; SSE-NEXT: movdqa %xmm10, %xmm9
-; SSE-NEXT: movdqa %xmm10, %xmm2
-; SSE-NEXT: psrlq $48, %xmm2
+; SSE-NEXT: por %xmm3, %xmm11
+; SSE-NEXT: psrlq $48, %xmm7
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm15[2,2,3,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm7[0]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm5[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
@@ -2607,37 +2604,35 @@
; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[0,1,2,3,7,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
-; SSE-NEXT: movdqa %xmm0, %xmm7
-; SSE-NEXT: pandn %xmm2, %xmm7
+; SSE-NEXT: movdqa %xmm0, %xmm9
+; SSE-NEXT: pandn %xmm2, %xmm9
; SSE-NEXT: pand %xmm0, %xmm1
-; SSE-NEXT: por %xmm1, %xmm7
+; SSE-NEXT: por %xmm1, %xmm9
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm4[2,3,2,3]
; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT: # xmm8 = mem[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,4,5,4,6]
; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm12[1]
; SSE-NEXT: movss {{.*#+}} xmm3 = xmm2[0],xmm3[1,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm11[0,2,2,3,4,5,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm12[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm0, %xmm6
-; SSE-NEXT: pandn %xmm2, %xmm6
+; SSE-NEXT: movdqa %xmm0, %xmm7
+; SSE-NEXT: pandn %xmm2, %xmm7
; SSE-NEXT: andps %xmm0, %xmm3
-; SSE-NEXT: por %xmm3, %xmm6
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: por %xmm3, %xmm7
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm15[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm15[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,3,2,3]
; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,0,3]
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,5,4,6]
+; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm13[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm13[0,1,2,3,4,5,4,6]
; SSE-NEXT: punpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = xmm2[1],mem[1]
; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
@@ -2645,10 +2640,10 @@
; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm0, %xmm5
-; SSE-NEXT: pandn %xmm3, %xmm5
+; SSE-NEXT: movdqa %xmm0, %xmm6
+; SSE-NEXT: pandn %xmm3, %xmm6
; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: por %xmm2, %xmm5
+; SSE-NEXT: por %xmm2, %xmm6
; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
@@ -2665,10 +2660,10 @@
; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: pandn %xmm1, %xmm4
+; SSE-NEXT: movdqa %xmm0, %xmm5
+; SSE-NEXT: pandn %xmm1, %xmm5
; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: por %xmm2, %xmm4
+; SSE-NEXT: por %xmm2, %xmm5
; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
@@ -2678,10 +2673,11 @@
; SSE-NEXT: # xmm1 = mem[0,1,0,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,6]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm9[1]
+; SSE-NEXT: punpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[1],mem[1]
; SSE-NEXT: movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm9[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,6]
; SSE-NEXT: movdqa %xmm0, %xmm3
@@ -2689,30 +2685,30 @@
; SSE-NEXT: andps %xmm0, %xmm1
; SSE-NEXT: por %xmm1, %xmm3
; SSE-NEXT: psrlq $48, %xmm10
+; SSE-NEXT: psrldq {{.*#+}} xmm4 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm10[0],xmm4[1],xmm10[1],xmm4[2],xmm10[2],xmm4[3],xmm10[3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3]
-; SSE-NEXT: psrld $16, %xmm12
+; SSE-NEXT: psrld $16, %xmm1
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm8[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm12[1]
-; SSE-NEXT: movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm11[3,1,2,3,4,5,6,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm1[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm4[0],xmm2[1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm12[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm1[0,1,2,3,4,4,5,7]
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: pandn %xmm8, %xmm1
; SSE-NEXT: andps %xmm0, %xmm2
; SSE-NEXT: por %xmm2, %xmm1
+; SSE-NEXT: psrlq $48, %xmm15
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3]
+; SSE-NEXT: movdqa %xmm2, %xmm4
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm8
-; SSE-NEXT: psrldq {{.*#+}} xmm15 = xmm15[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm11
-; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm11[1]
-; SSE-NEXT: movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
+; SSE-NEXT: psrld $16, %xmm8
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm13[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm8[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm4[0],xmm2[1,2,3]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT: # xmm8 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,1,0,3]
@@ -2721,18 +2717,18 @@
; SSE-NEXT: pandn %xmm8, %xmm15
; SSE-NEXT: andps %xmm0, %xmm2
; SSE-NEXT: por %xmm2, %xmm15
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm8
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm4
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
-; SSE-NEXT: movdqa %xmm2, %xmm8
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm11
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm8
; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm11[1]
-; SSE-NEXT: movss {{.*#+}} xmm2 = xmm8[0],xmm2[1,2,3]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm8[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm4[0],xmm2[1,2,3]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT: # xmm8 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,1,0,3]
@@ -2741,20 +2737,21 @@
; SSE-NEXT: pandn %xmm8, %xmm12
; SSE-NEXT: andps %xmm0, %xmm2
; SSE-NEXT: por %xmm2, %xmm12
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm8
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm4
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
-; SSE-NEXT: movdqa %xmm2, %xmm8
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm11
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm8
; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm11[1]
-; SSE-NEXT: movss {{.*#+}} xmm2 = xmm8[0],xmm2[1,2,3]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm8[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm4[0],xmm2[1,2,3]
; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm9[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,5,7]
; SSE-NEXT: pandn %xmm8, %xmm0
@@ -2762,42 +2759,42 @@
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 32(%rsi)
+; SSE-NEXT: movaps %xmm2, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, (%rsi)
+; SSE-NEXT: movaps %xmm2, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 16(%rdx)
-; SSE-NEXT: movaps %xmm14, 32(%rdx)
+; SSE-NEXT: movaps %xmm14, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, (%rdx)
+; SSE-NEXT: movaps %xmm2, 32(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 32(%rcx)
-; SSE-NEXT: movaps (%rsp), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 48(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, (%rcx)
-; SSE-NEXT: movdqa %xmm7, 16(%r8)
-; SSE-NEXT: movdqa %xmm13, 32(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 48(%rcx)
+; SSE-NEXT: movaps (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, 32(%rcx)
+; SSE-NEXT: movdqa %xmm9, 16(%r8)
+; SSE-NEXT: movdqa %xmm11, (%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps %xmm2, 48(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, (%r8)
+; SSE-NEXT: movaps %xmm2, 32(%r8)
; SSE-NEXT: movdqa %xmm3, 16(%r9)
-; SSE-NEXT: movdqa %xmm4, 32(%r9)
-; SSE-NEXT: movdqa %xmm5, 48(%r9)
-; SSE-NEXT: movdqa %xmm6, (%r9)
+; SSE-NEXT: movdqa %xmm5, (%r9)
+; SSE-NEXT: movdqa %xmm6, 48(%r9)
+; SSE-NEXT: movdqa %xmm7, 32(%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT: movdqa %xmm0, 16(%rax)
-; SSE-NEXT: movdqa %xmm12, 32(%rax)
+; SSE-NEXT: movdqa %xmm12, (%rax)
; SSE-NEXT: movdqa %xmm15, 48(%rax)
-; SSE-NEXT: movdqa %xmm1, (%rax)
-; SSE-NEXT: addq $488, %rsp # imm = 0x1E8
+; SSE-NEXT: movdqa %xmm1, 32(%rax)
+; SSE-NEXT: addq $472, %rsp # imm = 0x1D8
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i16_stride6_vf32:
@@ -5313,30 +5310,29 @@
define void @load_i16_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
; SSE-LABEL: load_i16_stride6_vf64:
; SSE: # %bb.0:
-; SSE-NEXT: subq $1160, %rsp # imm = 0x488
-; SSE-NEXT: movdqa 496(%rdi), %xmm5
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 512(%rdi), %xmm10
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 144(%rdi), %xmm8
+; SSE-NEXT: subq $1128, %rsp # imm = 0x468
+; SSE-NEXT: movdqa 592(%rdi), %xmm7
+; SSE-NEXT: movdqa 608(%rdi), %xmm8
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 160(%rdi), %xmm3
-; SSE-NEXT: movdqa 176(%rdi), %xmm0
-; SSE-NEXT: movdqa 112(%rdi), %xmm7
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 96(%rdi), %xmm4
+; SSE-NEXT: movdqa 240(%rdi), %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 256(%rdi), %xmm3
+; SSE-NEXT: movdqa 272(%rdi), %xmm0
+; SSE-NEXT: movdqa 208(%rdi), %xmm6
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 192(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 128(%rdi), %xmm1
+; SSE-NEXT: movdqa 224(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa {{.*#+}} xmm12 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: movdqa {{.*#+}} xmm11 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT: movdqa %xmm11, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,1,0,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
-; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; SSE-NEXT: pand %xmm11, %xmm1
; SSE-NEXT: por %xmm2, %xmm1
; SSE-NEXT: movdqa %xmm1, %xmm2
; SSE-NEXT: movdqa %xmm3, %xmm1
@@ -5353,41 +5349,42 @@
; SSE-NEXT: pslld $16, %xmm0
; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[0,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[0,3,2,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm10[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[0,1,1,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa 480(%rdi), %xmm0
+; SSE-NEXT: movdqa 576(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm7[2],xmm0[3],xmm7[3]
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm2
-; SSE-NEXT: movdqa 544(%rdi), %xmm4
-; SSE-NEXT: movdqa 560(%rdi), %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT: movdqa 640(%rdi), %xmm3
+; SSE-NEXT: movdqa 656(%rdi), %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm1, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[3,0]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm4, %xmm0
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm1[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[2,3]
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm3, %xmm0
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm1[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[2,3]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslld $16, %xmm1
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: movdqa 528(%rdi), %xmm1
+; SSE-NEXT: movdqa 624(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5395,26 +5392,104 @@
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 32(%rdi), %xmm0
+; SSE-NEXT: movdqa 320(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa (%rdi), %xmm0
+; SSE-NEXT: movdqa 304(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 288(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 16(%rdi), %xmm6
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm6[2],xmm0[3],xmm6[3]
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: pand %xmm11, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: movdqa 352(%rdi), %xmm3
+; SSE-NEXT: movdqa 368(%rdi), %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm3[2,2,3,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1],xmm15[2],xmm0[2],xmm15[3],xmm0[3]
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[3,0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm0[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[2,3]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pslld $16, %xmm0
+; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT: movdqa 336(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 704(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movdqa 688(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 672(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: pand %xmm11, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: movdqa 736(%rdi), %xmm3
+; SSE-NEXT: movdqa 752(%rdi), %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm3[2,2,3,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[3,0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm0[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[2,3]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pslld $16, %xmm0
+; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT: movdqa 720(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 32(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm11, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movdqa (%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 16(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: movdqa 64(%rdi), %xmm3
; SSE-NEXT: movdqa 80(%rdi), %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm3[2,2,3,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm3[2,2,3,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[3,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5437,7 +5512,7 @@
; SSE-NEXT: movdqa 416(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa 400(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5447,21 +5522,21 @@
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm2
-; SSE-NEXT: movdqa 448(%rdi), %xmm4
+; SSE-NEXT: movdqa 448(%rdi), %xmm3
; SSE-NEXT: movdqa 464(%rdi), %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm4[2,2,3,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm3[2,2,3,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[3,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[3,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm0[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm0[2,3]
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm0[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[2,3]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslld $16, %xmm0
; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
@@ -5473,204 +5548,109 @@
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 320(%rdi), %xmm0
+; SSE-NEXT: movdqa 128(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa 304(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 288(%rdi), %xmm0
+; SSE-NEXT: movdqa 112(%rdi), %xmm13
+; SSE-NEXT: movdqa 96(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT: pand %xmm12, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, %xmm2
-; SSE-NEXT: movdqa 352(%rdi), %xmm4
-; SSE-NEXT: movdqa 368(%rdi), %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm4[2,2,3,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[3,0]
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm0[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm0[2,3]
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pslld $16, %xmm0
-; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa 336(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 704(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm12, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa 688(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 672(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT: pand %xmm12, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, %xmm2
-; SSE-NEXT: movdqa 736(%rdi), %xmm4
-; SSE-NEXT: movdqa 752(%rdi), %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm4[2,2,3,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[3,0]
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: movaps %xmm4, (%rsp) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm0[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm0[2,3]
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pslld $16, %xmm0
-; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa 720(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 224(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm12, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa 208(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm0[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,4,6,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm13[2],xmm0[3],xmm13[3]
+; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: movdqa 256(%rdi), %xmm3
-; SSE-NEXT: movdqa 272(%rdi), %xmm0
+; SSE-NEXT: movdqa 160(%rdi), %xmm3
+; SSE-NEXT: movdqa 176(%rdi), %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[2,2,3,3]
; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; SSE-NEXT: movdqa %xmm0, %xmm4
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm3[3,0]
; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm3, %xmm4
+; SSE-NEXT: movdqa %xmm3, %xmm5
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm0[0,0]
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[2,3]
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslld $16, %xmm0
-; SSE-NEXT: psrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: movdqa 240(%rdi), %xmm0
+; SSE-NEXT: psrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE-NEXT: movdqa 144(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[3,1],xmm0[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm0[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm14[0,1,0,2,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[3,1],xmm0[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 608(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 512(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,2,4,5,6,7]
-; SSE-NEXT: movdqa %xmm12, %xmm4
-; SSE-NEXT: pandn %xmm0, %xmm4
-; SSE-NEXT: movdqa 592(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm11, %xmm5
+; SSE-NEXT: pandn %xmm0, %xmm5
+; SSE-NEXT: movdqa 496(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 576(%rdi), %xmm0
+; SSE-NEXT: movdqa 480(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm0[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm15[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm0[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,4,6,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: pand %xmm12, %xmm0
-; SSE-NEXT: por %xmm4, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm0
+; SSE-NEXT: por %xmm5, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: movdqa 640(%rdi), %xmm7
-; SSE-NEXT: movdqa 656(%rdi), %xmm4
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[2,2,3,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
-; SSE-NEXT: movdqa %xmm4, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[3,0]
+; SSE-NEXT: movdqa 544(%rdi), %xmm12
+; SSE-NEXT: movdqa 560(%rdi), %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm12[2,2,3,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; SSE-NEXT: movdqa %xmm5, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm12[3,0]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm7, %xmm0
-; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[1,0],xmm4[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm4[2,3]
-; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pslld $16, %xmm4
-; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT: movdqa 624(%rdi), %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm4[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm13 = xmm9[0,1,0,2,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm13[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm0
-; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: pshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,3,5,7,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm12, %xmm13
; SSE-NEXT: movdqa %xmm12, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: pandn %xmm4, %xmm0
-; SSE-NEXT: pand %xmm12, %xmm1
-; SSE-NEXT: por %xmm0, %xmm1
-; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[3,1],xmm0[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,0]
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[1,0],xmm5[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[2,0],xmm5[2,3]
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pslld $16, %xmm5
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
+; SSE-NEXT: movdqa 528(%rdi), %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm5[0,1,0,2,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm12[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: pshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,2,3,5,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm13, %xmm0
+; SSE-NEXT: movdqa %xmm11, %xmm12
+; SSE-NEXT: movdqa %xmm11, %xmm0
; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: pand %xmm13, %xmm1
+; SSE-NEXT: pand %xmm11, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[3,1],xmm0[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,0]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrld $16, %xmm6
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: pshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,2,3,5,7,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
-; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: pandn %xmm6, %xmm0
-; SSE-NEXT: pand %xmm13, %xmm1
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT: movdqa %xmm11, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[3,1],xmm0[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm11[2,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5679,25 +5659,25 @@
; SSE-NEXT: pshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,2,3,5,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: pandn %xmm11, %xmm0
-; SSE-NEXT: pand %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: pand %xmm12, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[3,1],xmm0[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,0]
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[3,1],xmm0[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm15, %xmm0
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: pshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,2,3,5,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: pandn %xmm14, %xmm0
-; SSE-NEXT: pand %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: pandn %xmm11, %xmm0
+; SSE-NEXT: pand %xmm12, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
@@ -5709,57 +5689,72 @@
; SSE-NEXT: pshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,2,3,5,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: pandn %xmm12, %xmm0
-; SSE-NEXT: pand %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: pand %xmm12, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[3,1],xmm0[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,0]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[3,1],xmm0[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm8, %xmm0
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: pshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,2,3,5,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm13, %xmm0
+; SSE-NEXT: movdqa %xmm12, %xmm0
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT: pandn %xmm7, %xmm0
-; SSE-NEXT: pand %xmm13, %xmm1
+; SSE-NEXT: pand %xmm12, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,1,3,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[3,1],xmm0[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrld $16, %xmm13
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm10[0,1,2,3,5,7,6,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm13[2],xmm1[3],xmm13[3]
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pandn %xmm10, %xmm0
+; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm14[0,1,1,3,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[3,1],xmm0[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm14, %xmm0
; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm15[0,1,2,3,5,7,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm9[0,1,2,3,5,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; SSE-NEXT: pand %xmm13, %xmm2
-; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT: por %xmm2, %xmm13
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm9[0,1,1,3,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[0,1],xmm3[2,0]
-; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm4, %xmm0
-; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: pand %xmm12, %xmm2
+; SSE-NEXT: movdqa (%rsp), %xmm6 # 16-byte Reload
+; SSE-NEXT: pandn %xmm6, %xmm12
+; SSE-NEXT: por %xmm2, %xmm12
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm5[0,1,1,3,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[3,1],xmm0[1,3]
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[0,1],xmm4[2,0]
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[1,1,1,1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm10[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm10[2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[0,2,2,3,4,5,6,7]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT: movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm5, %xmm1
+; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[2,3]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[1,0,2,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm2
+; SSE-NEXT: pand %xmm5, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
@@ -5768,55 +5763,28 @@
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,6,5,4]
-; SSE-NEXT: movdqa {{.*#+}} xmm15 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT: movdqa %xmm15, %xmm5
-; SSE-NEXT: pandn %xmm1, %xmm5
-; SSE-NEXT: pand %xmm15, %xmm2
-; SSE-NEXT: por %xmm2, %xmm5
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrldq {{.*#+}} xmm6 = xmm6[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[1,1,1,1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm2[0]
-; SSE-NEXT: movdqa %xmm4, %xmm2
-; SSE-NEXT: pandn %xmm6, %xmm2
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[1,0],xmm0[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[2,0],xmm0[2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm8[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,2,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm1
-; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
-; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,5,4]
-; SSE-NEXT: movdqa %xmm15, %xmm5
-; SSE-NEXT: pandn %xmm2, %xmm5
-; SSE-NEXT: pand %xmm15, %xmm1
-; SSE-NEXT: por %xmm1, %xmm5
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm1, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm2
+; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[1,1,1,1]
; SSE-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
-; SSE-NEXT: movdqa %xmm4, %xmm2
+; SSE-NEXT: movdqa %xmm5, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,0],xmm5[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm5[2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm1[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[1,0,2,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm5
-; SSE-NEXT: por %xmm2, %xmm5
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,0],xmm15[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm15[2,3]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm1[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,0,2,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm3
+; SSE-NEXT: por %xmm2, %xmm3
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
@@ -5824,46 +5792,72 @@
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,5,4]
-; SSE-NEXT: movdqa %xmm15, %xmm6
-; SSE-NEXT: pandn %xmm2, %xmm6
-; SSE-NEXT: pand %xmm15, %xmm5
-; SSE-NEXT: por %xmm5, %xmm6
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrldq {{.*#+}} xmm7 = xmm7[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = mem[1,1,1,1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm7 = xmm7[0],xmm5[0]
-; SSE-NEXT: movdqa %xmm4, %xmm5
-; SSE-NEXT: pandn %xmm7, %xmm5
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[1,0],xmm0[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[2,0],xmm0[2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm9[0,2,2,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm3
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm7, %xmm2
+; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE-NEXT: movdqa %xmm5, %xmm3
+; SSE-NEXT: pandn %xmm2, %xmm3
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[1,0],xmm8[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[2,0],xmm8[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm13[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,0,2,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm2
-; SSE-NEXT: por %xmm5, %xmm2
+; SSE-NEXT: pand %xmm5, %xmm2
+; SSE-NEXT: por %xmm3, %xmm2
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,5,4]
-; SSE-NEXT: movdqa %xmm15, %xmm6
-; SSE-NEXT: pandn %xmm5, %xmm6
-; SSE-NEXT: pand %xmm15, %xmm2
-; SSE-NEXT: por %xmm2, %xmm6
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm14, %xmm2
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,4]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm2
+; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm6, %xmm2
; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = mem[1,1,1,1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm5[0]
-; SSE-NEXT: movdqa %xmm4, %xmm5
-; SSE-NEXT: pandn %xmm2, %xmm5
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE-NEXT: movdqa %xmm5, %xmm3
+; SSE-NEXT: pandn %xmm2, %xmm3
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[1,0],xmm14[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[2,0],xmm14[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm11[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,0,2,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm2
+; SSE-NEXT: por %xmm3, %xmm2
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,4]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm2
+; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm14[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE-NEXT: movdqa %xmm5, %xmm3
+; SSE-NEXT: pandn %xmm2, %xmm3
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
@@ -5871,36 +5865,36 @@
; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm2[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[1,0,2,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm6
-; SSE-NEXT: por %xmm5, %xmm6
+; SSE-NEXT: pand %xmm5, %xmm6
+; SSE-NEXT: por %xmm3, %xmm6
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,5,4]
-; SSE-NEXT: movdqa %xmm15, %xmm7
-; SSE-NEXT: pandn %xmm5, %xmm7
-; SSE-NEXT: pand %xmm15, %xmm6
-; SSE-NEXT: por %xmm6, %xmm7
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm5
-; SSE-NEXT: psrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm10[1,1,1,1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm6[0]
-; SSE-NEXT: movdqa %xmm4, %xmm6
-; SSE-NEXT: pandn %xmm5, %xmm6
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,4]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm6
+; SSE-NEXT: por %xmm6, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm6[0]
+; SSE-NEXT: movdqa %xmm5, %xmm6
+; SSE-NEXT: pandn %xmm3, %xmm6
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[1,0],xmm0[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[2,0],xmm0[2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm11[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[1,0,2,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm5
-; SSE-NEXT: por %xmm6, %xmm5
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm0[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm3[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[1,0,2,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm7
+; SSE-NEXT: por %xmm6, %xmm7
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
@@ -5908,56 +5902,28 @@
; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm0[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,5,4]
-; SSE-NEXT: movdqa %xmm15, %xmm7
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm6, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm7
+; SSE-NEXT: por %xmm7, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm8, %xmm6
+; SSE-NEXT: psrldq {{.*#+}} xmm6 = xmm6[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm15[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm7[0]
+; SSE-NEXT: movdqa %xmm5, %xmm7
; SSE-NEXT: pandn %xmm6, %xmm7
-; SSE-NEXT: pand %xmm15, %xmm5
-; SSE-NEXT: por %xmm5, %xmm7
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm12, %xmm5
-; SSE-NEXT: psrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT: # xmm6 = mem[1,1,1,1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm6[0]
-; SSE-NEXT: movdqa %xmm4, %xmm7
-; SSE-NEXT: pandn %xmm5, %xmm7
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[1,0],xmm5[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[2,0],xmm5[2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm6[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[1,0,2,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm5
-; SSE-NEXT: por %xmm7, %xmm5
-; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
-; SSE-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,6,5,4]
-; SSE-NEXT: movdqa %xmm15, %xmm12
-; SSE-NEXT: pandn %xmm7, %xmm12
-; SSE-NEXT: pand %xmm15, %xmm5
-; SSE-NEXT: por %xmm5, %xmm12
-; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm13, %xmm5
-; SSE-NEXT: psrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: # xmm7 = mem[1,1,1,1]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm7[0]
-; SSE-NEXT: movdqa %xmm4, %xmm7
-; SSE-NEXT: pandn %xmm5, %xmm7
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[1,0],xmm0[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[2,0],xmm0[2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm12[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[1,0,2,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm5
-; SSE-NEXT: por %xmm7, %xmm5
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[1,0],xmm0[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[2,0],xmm0[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm6[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm9[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm9[1,0,2,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm9
+; SSE-NEXT: por %xmm7, %xmm9
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,2]
@@ -5965,565 +5931,592 @@
; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm0[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,6,5,4]
-; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: movdqa %xmm4, %xmm0
; SSE-NEXT: pandn %xmm7, %xmm0
-; SSE-NEXT: pand %xmm15, %xmm5
-; SSE-NEXT: por %xmm5, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm9
+; SSE-NEXT: por %xmm9, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm5
+; SSE-NEXT: movdqa %xmm10, %xmm1
+; SSE-NEXT: movdqa %xmm10, %xmm7
+; SSE-NEXT: psrldq {{.*#+}} xmm7 = xmm7[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[2,2,3,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm7 = xmm7[0],xmm5[0]
-; SSE-NEXT: movdqa %xmm4, %xmm5
-; SSE-NEXT: pandn %xmm7, %xmm5
-; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm0[1,1,1,1]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm7 = xmm7[0],xmm9[0]
+; SSE-NEXT: movdqa %xmm5, %xmm9
+; SSE-NEXT: pandn %xmm7, %xmm9
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[1,0],xmm10[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm10[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm7[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm12[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm12[1,0,2,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm12
+; SSE-NEXT: por %xmm9, %xmm12
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: shufps $132, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT: # xmm9 = xmm9[0,1],mem[0,2]
+; SSE-NEXT: movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm9[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,6,5,4]
+; SSE-NEXT: movdqa %xmm4, %xmm10
+; SSE-NEXT: pandn %xmm9, %xmm10
+; SSE-NEXT: pand %xmm4, %xmm12
+; SSE-NEXT: por %xmm12, %xmm10
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm15, %xmm9
+; SSE-NEXT: psrlq $48, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm8[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm12 = xmm12[0],xmm9[0]
+; SSE-NEXT: movdqa %xmm5, %xmm9
+; SSE-NEXT: pandn %xmm12, %xmm9
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm6
+; SSE-NEXT: por %xmm9, %xmm6
+; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT: # xmm9 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm9[0,1,0,2]
+; SSE-NEXT: movdqa %xmm4, %xmm8
+; SSE-NEXT: pandn %xmm9, %xmm8
+; SSE-NEXT: pand %xmm4, %xmm6
+; SSE-NEXT: por %xmm6, %xmm8
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $48, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm1[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm9 = xmm9[0],xmm0[0]
+; SSE-NEXT: movdqa %xmm5, %xmm6
+; SSE-NEXT: pandn %xmm9, %xmm6
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm7
+; SSE-NEXT: por %xmm6, %xmm7
+; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,0,2]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm6, %xmm0
; SSE-NEXT: pand %xmm4, %xmm7
-; SSE-NEXT: por %xmm5, %xmm7
-; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = mem[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,1,0,2]
-; SSE-NEXT: movdqa %xmm15, %xmm8
-; SSE-NEXT: pandn %xmm5, %xmm8
-; SSE-NEXT: pand %xmm15, %xmm7
-; SSE-NEXT: por %xmm7, %xmm8
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm8, %xmm5
-; SSE-NEXT: psrlq $48, %xmm5
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: # xmm7 = mem[2,2,3,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm7 = xmm7[0],xmm5[0]
-; SSE-NEXT: movdqa %xmm4, %xmm5
-; SSE-NEXT: pandn %xmm7, %xmm5
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm3
-; SSE-NEXT: por %xmm5, %xmm3
-; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = mem[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,1,0,2]
-; SSE-NEXT: movdqa %xmm15, %xmm7
-; SSE-NEXT: pandn %xmm5, %xmm7
-; SSE-NEXT: pand %xmm15, %xmm3
-; SSE-NEXT: por %xmm3, %xmm7
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm7, %xmm3
-; SSE-NEXT: psrlq $48, %xmm3
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = mem[2,2,3,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm3[0]
-; SSE-NEXT: movdqa %xmm4, %xmm3
-; SSE-NEXT: pandn %xmm5, %xmm3
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm9[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[0,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm5
-; SSE-NEXT: por %xmm3, %xmm5
-; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
-; SSE-NEXT: movdqa %xmm15, %xmm9
-; SSE-NEXT: pandn %xmm3, %xmm9
-; SSE-NEXT: pand %xmm15, %xmm5
-; SSE-NEXT: por %xmm5, %xmm9
-; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm3
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = mem[2,2,3,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm3[0]
-; SSE-NEXT: movdqa %xmm4, %xmm3
-; SSE-NEXT: pandn %xmm5, %xmm3
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm1
-; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
-; SSE-NEXT: movdqa %xmm15, %xmm5
-; SSE-NEXT: pandn %xmm3, %xmm5
-; SSE-NEXT: pand %xmm15, %xmm1
-; SSE-NEXT: por %xmm1, %xmm5
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrlq $48, %xmm10
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[2,2,3,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm10[0]
-; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm11[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
-; SSE-NEXT: movdqa %xmm15, %xmm5
-; SSE-NEXT: pandn %xmm1, %xmm5
-; SSE-NEXT: pand %xmm15, %xmm3
-; SSE-NEXT: por %xmm3, %xmm5
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm9, %xmm1
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm14[2,2,3,3]
-; SSE-NEXT: movdqa %xmm14, %xmm11
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm1[0]
-; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: por %xmm7, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm14, %xmm6
+; SSE-NEXT: psrlq $48, %xmm6
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm14[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm7 = xmm7[0],xmm6[0]
+; SSE-NEXT: movdqa %xmm5, %xmm6
+; SSE-NEXT: pandn %xmm7, %xmm6
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm2
+; SSE-NEXT: por %xmm6, %xmm2
+; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,0,2]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm6, %xmm0
; SSE-NEXT: pand %xmm4, %xmm2
-; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
-; SSE-NEXT: movdqa %xmm15, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: pand %xmm15, %xmm2
+; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm7, %xmm2
+; SSE-NEXT: psrlq $48, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm12[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm2[0]
+; SSE-NEXT: movdqa %xmm5, %xmm2
+; SSE-NEXT: pandn %xmm6, %xmm2
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm3
; SSE-NEXT: por %xmm2, %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm5, %xmm1
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm13[2,2,3,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm3
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm6, %xmm2
+; SSE-NEXT: psrlq $48, %xmm2
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT: movdqa %xmm5, %xmm2
+; SSE-NEXT: pandn %xmm3, %xmm2
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm13[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm3
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm3
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm9, %xmm2
+; SSE-NEXT: psrlq $48, %xmm2
+; SSE-NEXT: pshufd $250, (%rsp), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT: movdqa %xmm5, %xmm2
+; SSE-NEXT: pandn %xmm3, %xmm2
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm11[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm3
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm3
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm11, %xmm2
+; SSE-NEXT: psrlq $48, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT: movdqa %xmm5, %xmm2
+; SSE-NEXT: pandn %xmm3, %xmm2
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm3
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm12[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm2
-; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: pand %xmm4, %xmm3
+; SSE-NEXT: por %xmm3, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm10, %xmm2
+; SSE-NEXT: psrlq $48, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm8[2,2,3,3]
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm5, %xmm1
+; SSE-NEXT: pandn %xmm3, %xmm5
+; SSE-NEXT: por %xmm1, %xmm5
; SSE-NEXT: pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,2,3,7,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
-; SSE-NEXT: movdqa %xmm15, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: pand %xmm15, %xmm2
-; SSE-NEXT: por %xmm2, %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm1
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[2,2,3,3]
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm1[0]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm6[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm4, %xmm1
-; SSE-NEXT: pandn %xmm2, %xmm4
-; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: pshufhw $231, (%rsp), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
-; SSE-NEXT: movdqa %xmm15, %xmm2
+; SSE-NEXT: movdqa %xmm4, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: pand %xmm15, %xmm4
-; SSE-NEXT: por %xmm4, %xmm2
+; SSE-NEXT: pand %xmm4, %xmm5
+; SSE-NEXT: por %xmm5, %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[2,3,2,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5,4,6]
-; SSE-NEXT: punpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[1],mem[1]
-; SSE-NEXT: movss {{.*#+}} xmm1 = xmm3[0],xmm1[1,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm14[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: andps %xmm15, %xmm1
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[2,3,2,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,5,4,6]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm8[1]
-; SSE-NEXT: movss {{.*#+}} xmm4 = xmm3[0],xmm4[1,2,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm15[1]
+; SSE-NEXT: movss {{.*#+}} xmm3 = xmm2[0],xmm3[1,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm15[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm4, %xmm5
+; SSE-NEXT: pandn %xmm2, %xmm5
+; SSE-NEXT: andps %xmm4, %xmm3
+; SSE-NEXT: por %xmm3, %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm13[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = xmm2[1],mem[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: andps %xmm15, %xmm4
-; SSE-NEXT: por %xmm4, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[0,1,0,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,5,4,6]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm7[1]
-; SSE-NEXT: movss {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3]
-; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm4, %xmm0
-; SSE-NEXT: andps %xmm15, %xmm3
-; SSE-NEXT: por %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[0,1,0,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,5,4,6]
-; SSE-NEXT: punpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[1],mem[1]
-; SSE-NEXT: movss {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3]
-; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm4, %xmm0
-; SSE-NEXT: andps %xmm15, %xmm3
-; SSE-NEXT: por %xmm3, %xmm0
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,0,3]
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
-; SSE-NEXT: punpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[1],mem[1]
-; SSE-NEXT: movss {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3]
-; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm15, %xmm12
-; SSE-NEXT: pandn %xmm4, %xmm12
-; SSE-NEXT: andps %xmm15, %xmm3
-; SSE-NEXT: por %xmm3, %xmm12
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[0,1,0,3]
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm9[1]
-; SSE-NEXT: movss {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3]
-; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm15, %xmm11
-; SSE-NEXT: pandn %xmm4, %xmm11
-; SSE-NEXT: andps %xmm15, %xmm3
-; SSE-NEXT: por %xmm3, %xmm11
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,0,3]
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm5[1]
-; SSE-NEXT: movss {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3]
-; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm15, %xmm9
-; SSE-NEXT: pandn %xmm4, %xmm9
-; SSE-NEXT: andps %xmm15, %xmm3
-; SSE-NEXT: por %xmm3, %xmm9
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,0,3]
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm10[1]
-; SSE-NEXT: movss {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3]
-; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,2,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,6]
-; SSE-NEXT: movdqa %xmm15, %xmm8
-; SSE-NEXT: pandn %xmm4, %xmm8
-; SSE-NEXT: andps %xmm15, %xmm3
-; SSE-NEXT: por %xmm3, %xmm8
-; SSE-NEXT: psrlq $48, %xmm13
-; SSE-NEXT: movdqa %xmm6, %xmm3
-; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm4
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm4[1]
-; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm14[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,7]
-; SSE-NEXT: movdqa %xmm15, %xmm7
-; SSE-NEXT: pandn %xmm3, %xmm7
-; SSE-NEXT: andps %xmm15, %xmm2
-; SSE-NEXT: por %xmm2, %xmm7
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; SSE-NEXT: movdqa %xmm2, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm2
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; SSE-NEXT: movss {{.*#+}} xmm1 = xmm3[0],xmm1[1,2,3]
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,7]
-; SSE-NEXT: movdqa %xmm15, %xmm6
-; SSE-NEXT: pandn %xmm2, %xmm6
-; SSE-NEXT: andps %xmm15, %xmm1
-; SSE-NEXT: por %xmm1, %xmm6
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT: movdqa %xmm1, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm3
-; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSE-NEXT: movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,7]
-; SSE-NEXT: movdqa %xmm15, %xmm10
-; SSE-NEXT: pandn %xmm2, %xmm10
-; SSE-NEXT: andps %xmm15, %xmm1
-; SSE-NEXT: por %xmm1, %xmm10
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT: movdqa %xmm1, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm3
-; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSE-NEXT: movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,7]
-; SSE-NEXT: movdqa %xmm15, %xmm14
-; SSE-NEXT: pandn %xmm2, %xmm14
-; SSE-NEXT: andps %xmm15, %xmm1
-; SSE-NEXT: por %xmm1, %xmm14
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT: movdqa %xmm1, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm3
-; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSE-NEXT: movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,7]
-; SSE-NEXT: movdqa %xmm15, %xmm13
-; SSE-NEXT: pandn %xmm2, %xmm13
-; SSE-NEXT: andps %xmm15, %xmm1
-; SSE-NEXT: por %xmm1, %xmm13
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: psrlq $48, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT: movdqa %xmm1, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm3
-; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSE-NEXT: movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm2[0,1,2,3,4,4,5,7]
-; SSE-NEXT: movdqa %xmm15, %xmm5
+; SSE-NEXT: movdqa %xmm4, %xmm5
; SSE-NEXT: pandn %xmm3, %xmm5
-; SSE-NEXT: andps %xmm15, %xmm1
-; SSE-NEXT: por %xmm1, %xmm5
+; SSE-NEXT: andps %xmm4, %xmm2
+; SSE-NEXT: por %xmm2, %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm14[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm14[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = xmm2[1],mem[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm4, %xmm5
+; SSE-NEXT: pandn %xmm3, %xmm5
+; SSE-NEXT: andps %xmm4, %xmm2
+; SSE-NEXT: por %xmm2, %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm12[0,1,0,3]
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm7[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm4, %xmm12
+; SSE-NEXT: pandn %xmm3, %xmm12
+; SSE-NEXT: andps %xmm4, %xmm2
+; SSE-NEXT: por %xmm2, %xmm12
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,0,3]
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm6[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm4, %xmm7
+; SSE-NEXT: pandn %xmm3, %xmm7
+; SSE-NEXT: andps %xmm4, %xmm2
+; SSE-NEXT: por %xmm2, %xmm7
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT: pshufd $196, (%rsp), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,0,3]
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm9[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm4, %xmm6
+; SSE-NEXT: pandn %xmm3, %xmm6
+; SSE-NEXT: andps %xmm4, %xmm2
+; SSE-NEXT: por %xmm2, %xmm6
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
+; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm11[1]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm3[0],xmm2[1,2,3]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm9 = xmm3[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm4, %xmm5
+; SSE-NEXT: pandn %xmm9, %xmm5
+; SSE-NEXT: andps %xmm4, %xmm2
+; SSE-NEXT: por %xmm2, %xmm5
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT: # xmm9 = mem[2,3,2,3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm2[0],xmm9[1],xmm2[1],xmm9[2],xmm2[2],xmm9[3],xmm2[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[0,1,0,3]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,6]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm10[1]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm9[0],xmm0[1,2,3]
+; SSE-NEXT: pshuflw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,2,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm9 = xmm2[0,1,2,3,4,5,4,6]
+; SSE-NEXT: movdqa %xmm4, %xmm10
+; SSE-NEXT: pandn %xmm9, %xmm10
+; SSE-NEXT: andps %xmm4, %xmm0
+; SSE-NEXT: por %xmm0, %xmm10
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: psrlq $48, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm3
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm15[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,7]
+; SSE-NEXT: movdqa %xmm4, %xmm9
+; SSE-NEXT: pandn %xmm1, %xmm9
+; SSE-NEXT: andps %xmm4, %xmm0
+; SSE-NEXT: por %xmm0, %xmm9
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: psrld $16, %xmm2
-; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm2[1]
-; SSE-NEXT: movss {{.*#+}} xmm3 = xmm1[0],xmm3[1,2,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,4,5,7]
-; SSE-NEXT: movdqa %xmm15, %xmm2
-; SSE-NEXT: pandn %xmm4, %xmm2
-; SSE-NEXT: andps %xmm15, %xmm3
-; SSE-NEXT: por %xmm3, %xmm2
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,7]
+; SSE-NEXT: movdqa %xmm4, %xmm11
+; SSE-NEXT: pandn %xmm1, %xmm11
+; SSE-NEXT: andps %xmm4, %xmm0
+; SSE-NEXT: por %xmm0, %xmm11
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT: movdqa %xmm3, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: psrld $16, %xmm1
-; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,2,3,4,5,5,7]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm1[1]
-; SSE-NEXT: movss {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3]
-; SSE-NEXT: andps %xmm15, %xmm3
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,5,7]
-; SSE-NEXT: pandn %xmm4, %xmm15
-; SSE-NEXT: por %xmm3, %xmm15
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 96(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 112(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 48(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 64(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, (%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 80(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 16(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 96(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 112(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 80(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 16(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 96(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 112(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 64(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 80(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 48(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 16(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 112(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 96(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 80(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 64(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 48(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 16(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%r8)
-; SSE-NEXT: movdqa %xmm8, 112(%r9)
-; SSE-NEXT: movdqa %xmm9, 96(%r9)
-; SSE-NEXT: movdqa %xmm11, 80(%r9)
-; SSE-NEXT: movdqa %xmm12, 64(%r9)
-; SSE-NEXT: movdqa %xmm0, 48(%r9)
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm1[0,1,2,3,4,4,5,7]
+; SSE-NEXT: movdqa %xmm4, %xmm3
+; SSE-NEXT: pandn %xmm8, %xmm3
+; SSE-NEXT: andps %xmm4, %xmm0
+; SSE-NEXT: por %xmm0, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm2
+; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,5,7]
+; SSE-NEXT: movdqa %xmm4, %xmm15
+; SSE-NEXT: pandn %xmm8, %xmm15
+; SSE-NEXT: andps %xmm4, %xmm0
+; SSE-NEXT: por %xmm0, %xmm15
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm2
+; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm8[0,1,2,3,4,4,5,7]
+; SSE-NEXT: movdqa %xmm4, %xmm8
+; SSE-NEXT: pandn %xmm13, %xmm8
+; SSE-NEXT: andps %xmm4, %xmm0
+; SSE-NEXT: por %xmm0, %xmm8
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm2
+; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm13[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,4,5,7]
+; SSE-NEXT: movdqa %xmm4, %xmm14
+; SSE-NEXT: pandn %xmm13, %xmm14
+; SSE-NEXT: andps %xmm4, %xmm0
+; SSE-NEXT: por %xmm0, %xmm14
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm13
+; SSE-NEXT: pshufhw $212, (%rsp), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm13[1]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT: # xmm13 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm13[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm13[0,1,2,3,4,4,5,7]
+; SSE-NEXT: movdqa %xmm4, %xmm13
+; SSE-NEXT: pandn %xmm1, %xmm13
+; SSE-NEXT: andps %xmm4, %xmm0
+; SSE-NEXT: por %xmm0, %xmm13
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm1
+; SSE-NEXT: pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,2,3,4,5,5,7]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
+; SSE-NEXT: andps %xmm4, %xmm0
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,7]
+; SSE-NEXT: pandn %xmm1, %xmm4
+; SSE-NEXT: por %xmm0, %xmm4
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 64(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, (%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 112(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 96(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 80(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 64(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 112(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 48(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 96(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, (%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 80(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 64(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 112(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 96(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 112(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 96(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 80(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 64(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 48(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, (%r8)
+; SSE-NEXT: movdqa %xmm10, 112(%r9)
+; SSE-NEXT: movdqa %xmm5, 96(%r9)
+; SSE-NEXT: movdqa %xmm6, 80(%r9)
+; SSE-NEXT: movdqa %xmm7, 64(%r9)
+; SSE-NEXT: movdqa %xmm12, 48(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 32(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%r9)
-; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movdqa %xmm15, 112(%rax)
-; SSE-NEXT: movdqa %xmm2, 96(%rax)
-; SSE-NEXT: movdqa %xmm5, 80(%rax)
-; SSE-NEXT: movdqa %xmm13, 64(%rax)
-; SSE-NEXT: movdqa %xmm14, 48(%rax)
-; SSE-NEXT: movdqa %xmm10, 32(%rax)
-; SSE-NEXT: movdqa %xmm6, 16(%rax)
-; SSE-NEXT: movdqa %xmm7, (%rax)
-; SSE-NEXT: addq $1160, %rsp # imm = 0x488
+; SSE-NEXT: movdqa %xmm4, 112(%rax)
+; SSE-NEXT: movdqa %xmm13, 96(%rax)
+; SSE-NEXT: movdqa %xmm14, 80(%rax)
+; SSE-NEXT: movdqa %xmm8, 64(%rax)
+; SSE-NEXT: movdqa %xmm15, 48(%rax)
+; SSE-NEXT: movdqa %xmm3, 32(%rax)
+; SSE-NEXT: movdqa %xmm11, 16(%rax)
+; SSE-NEXT: movdqa %xmm9, (%rax)
+; SSE-NEXT: addq $1128, %rsp # imm = 0x468
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i16_stride6_vf64:
@@ -6549,13 +6542,14 @@
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm4
; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 32(%rdi), %xmm5
; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm2[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm5[0,1,0,2,4,5,6,7]
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm5, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,0,2,4,5,6,7]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -6564,7 +6558,7 @@
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4,5],xmm2[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT: vpslld $16, %xmm1, %xmm1
; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -6626,9 +6620,9 @@
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; AVX1-ONLY-NEXT: vmovdqa 512(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm1, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa 512(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm6, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 528(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm11 = xmm1[0,3,2,3]
@@ -6644,7 +6638,7 @@
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpslld $16, %xmm0, %xmm0
; AVX1-ONLY-NEXT: vmovdqa 256(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm1
@@ -6676,9 +6670,9 @@
; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpslld $16, %xmm1, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm6[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; AVX1-ONLY-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -6758,14 +6752,14 @@
; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm5[0,1,1,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,1,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm14 = xmm5[1,1,1,1]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
; AVX1-ONLY-NEXT: vpshufhw $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm14 = mem[0,1,2,3,5,7,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm15, %xmm15
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm5, %xmm15
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm14 = xmm14[2],xmm15[2],xmm14[3],xmm15[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm14[0,1,2],xmm1[3,4,5],xmm14[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -6776,7 +6770,7 @@
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm14 = mem[2,2,3,3]
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vpunpcklwd (%rsp), %xmm14, %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm14 = xmm14[0],mem[0],xmm14[1],mem[1],xmm14[2],mem[2],xmm14[3],mem[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm14[6,7]
; AVX1-ONLY-NEXT: vandps %ymm0, %ymm10, %ymm0
@@ -6805,11 +6799,10 @@
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm12[0,1,2],xmm1[3,4,5],xmm12[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm11[0,1,1,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm11 = xmm6[1,1,1,1]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
-; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = mem[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm11 = xmm13[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm11 = xmm11[0],mem[0],xmm11[1],mem[1],xmm11[2],mem[2],xmm11[3],mem[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm11[6,7]
@@ -6822,7 +6815,7 @@
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrld $16, %xmm1, %xmm1
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX1-ONLY-NEXT: vpshufd $250, (%rsp), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
@@ -6833,8 +6826,8 @@
; AVX1-ONLY-NEXT: # xmm8 = mem[1,1,1,1]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm8 = xmm9[0,1,2,3,5,7,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm9, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm6, %xmm9
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm8 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm8[0,1,2],xmm1[3,4,5],xmm8[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -6842,7 +6835,8 @@
; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = mem[1,1,1,1]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm6[2,2,3,3]
+; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = xmm7[0],mem[0],xmm7[1],mem[1],xmm7[2],mem[2],xmm7[3],mem[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm7[6,7]
@@ -6852,13 +6846,13 @@
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,5,7,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm12, %xmm1
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm13[2,2,3,3]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1],xmm1[2],xmm11[2],xmm1[3],xmm11[3]
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm11, %xmm1
+; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm12[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vpshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,1,3,4,5,6,7]
@@ -6872,13 +6866,13 @@
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3,4,5],xmm3[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm2[0,1,1,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm4[1,1,1,1]
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[1,1,1,1]
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm8[2,2,3,3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm2[6,7]
; AVX1-ONLY-NEXT: vandps %ymm0, %ymm10, %ymm0
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
@@ -6887,12 +6881,12 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm1 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm15 = [4,5,0,1,12,13,14,15,8,9,10,11,12,13,14,15]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[0,1],mem[2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm5[0,1],mem[2,3],xmm5[4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm1, %xmm1
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm0[3,4],xmm1[5,6,7]
@@ -6918,7 +6912,7 @@
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vpblendw $48, (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = xmm0[0,1,2,3],mem[4,5],xmm0[6,7]
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm0, %xmm3
@@ -6958,9 +6952,8 @@
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm0[0,1,2,3],mem[4,5],xmm0[6,7]
+; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = xmm13[0,1,2,3],mem[4,5],xmm13[6,7]
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm0, %xmm3
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3,4],xmm3[5,6,7]
@@ -6978,11 +6971,11 @@
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm0, %xmm2
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm12[0,1],mem[2,3],xmm12[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = xmm11[0,1],mem[2,3],xmm11[4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm0, %xmm2
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm13[0,1,2,3],xmm11[4,5],xmm13[6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm12[0,1,2,3],xmm9[4,5],xmm12[6,7]
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm0, %xmm3
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2
@@ -6991,9 +6984,10 @@
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm1[0]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm8[0,1,2,3],xmm9[4,5],xmm8[6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm8[0,1,2,3],xmm4[4,5],xmm8[6,7]
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm0, %xmm4
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
@@ -7002,8 +6996,8 @@
; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm10, %ymm3
; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm9[1,1,1,1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[1,1,1,1]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
@@ -7014,25 +7008,25 @@
; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm0, %xmm3
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3,4],xmm3[5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm12 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm12, %xmm0
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm3, %xmm0
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $207, (%rsp), %xmm1, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2,3],xmm1[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm3, %xmm1
+; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm1[0,1,2,3],mem[4,5],xmm1[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm15, %xmm1
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm5, %ymm1
; AVX1-ONLY-NEXT: vandps %ymm5, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm11[1,1,1,1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm15[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm11[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $48, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1,2,3],mem[4,5],xmm2[6,7]
+; AVX1-ONLY-NEXT: vpblendw $207, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3],xmm2[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm2, %xmm14
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm14[0,1,2],xmm1[3,4],xmm14[5,6,7]
; AVX1-ONLY-NEXT: vandps %ymm0, %ymm10, %ymm0
@@ -7058,8 +7052,8 @@
; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm5, %ymm0
; AVX1-ONLY-NEXT: vandps %ymm5, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm9, %xmm1
; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm8 = mem[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm8[0],xmm1[0]
@@ -7071,8 +7065,8 @@
; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm10, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm8, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm0
; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
@@ -7089,8 +7083,8 @@
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm13[2,2,3,3]
+; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm7[0],xmm1[0]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm7, %xmm7
@@ -7100,20 +7094,23 @@
; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm10, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm9, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm12, %xmm7
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm12, %xmm0
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm1, %xmm1
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3,4],xmm1[5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm12, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm3, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm15, %xmm3
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm3, %ymm1
; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm5, %ymm0
; AVX1-ONLY-NEXT: vandps %ymm5, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm11, %xmm1
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm15[2,2,3,3]
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm13, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm13, %xmm15
+; AVX1-ONLY-NEXT: vmovdqa %xmm11, %xmm13
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm11[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm3[0],xmm1[0]
; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm2, %xmm2
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4],xmm2[5,6,7]
@@ -7122,8 +7119,8 @@
; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm10, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm0
; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
@@ -7138,8 +7135,8 @@
; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm5, %ymm0
; AVX1-ONLY-NEXT: vandps %ymm5, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm11, %xmm1
; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = mem[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
@@ -7153,12 +7150,12 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm10[2,3,2,3]
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,1],xmm1[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm1[0,1],mem[2,3],xmm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,4,5,4,5,6,7,0,1,4,5,0,1,12,13]
; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm2
@@ -7172,257 +7169,257 @@
; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm8[1]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[1],mem[1]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm5 = [65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm5, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm8 = [65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm8, %ymm2
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1],xmm2[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm13[0,1,0,3]
+; AVX1-ONLY-NEXT: vpblendw $243, (%rsp), %xmm2, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1],xmm2[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,6]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[1],mem[1]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm4, %xmm3
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm9[1]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm5, %xmm3
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm3[5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm2[0,1],mem[2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm13, %xmm2
+; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = mem[0,1],xmm2[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm12, %xmm2
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[1,1,1,1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm11[2,3,2,3]
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm3[0,1,2,3,4,5,4,6]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm6[1],mem[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm5, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,6]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm8, %ymm2
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = mem[0,1],xmm2[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm2[0,1],mem[2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,0,3]
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,6]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[1],mem[1]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm5, %xmm3
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm3[5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1],xmm2[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm7 = xmm6[0,1,2,3,4,5,4,6]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,4,5,4,6]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm7[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm8, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm7 = xmm7[1],xmm2[1]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm8, %xmm10
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm10[5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm2[0,1],mem[2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm13[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm5[0,1,2,3,4,5,4,6]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm15[1]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm3[5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, (%rsp), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm7[0,1],mem[2,3],xmm7[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm7, %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm7, %ymm0
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm10[0],xmm7[0],xmm10[1],xmm7[1],xmm10[2],xmm7[2],xmm10[3],xmm7[3]
-; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,0,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm10 = xmm8[0,1,2,3,4,5,4,6]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm10 = xmm10[1],xmm9[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1],xmm10[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm5, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm7, %ymm7
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm7, %ymm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = mem[0,1],xmm7[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm15[0,1,0,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,4,6]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm7[1],mem[1]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm8, %xmm14
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4],xmm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[2,3,2,3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm2[0,1],mem[2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm15, %xmm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm7 = xmm3[0,1,2,3,4,5,4,6]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm7 = xmm7[1],xmm4[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm8, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm2, %ymm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1],mem[2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm2, %xmm7
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm9 = xmm1[0,1,2,3,4,5,4,6]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm9 = xmm9[1],xmm11[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm9[0,1,2,3,4],xmm7[5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = xmm7[0,1],mem[2,3],xmm7[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm10, %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm7, %ymm0
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm15[0],xmm7[0],xmm15[1],xmm7[1],xmm15[2],xmm7[2],xmm15[3],xmm7[3]
-; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm9 = xmm15[0,1,2,3,4,5,4,6]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = xmm9[1],mem[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm5, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm7, %ymm7
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm7, %ymm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm7[0,1],mem[2,3],xmm7[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm7, %xmm9
-; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = mem[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm8 = xmm14[0,1,2,3,4,5,4,6]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm8 = xmm8[1],xmm1[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1,2,3,4],xmm9[5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm8 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm7 = xmm10[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = [6,7,2,3,4,5,6,7,6,7,6,7,2,3,14,15]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm13, %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm12, %xmm8
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm9 = xmm11[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm9, %xmm9
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,5,7]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm9[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm5, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm2, %xmm4
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,4,5,5,7]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm4[1]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm4[5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm4, %xmm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm9, %ymm7
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm10 = xmm10[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm2
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,1,2,3,4,5,5,7]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm10 = xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1],xmm10[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm8, %ymm7
+; AVX1-ONLY-NEXT: vandps %ymm8, %ymm9, %ymm9
+; AVX1-ONLY-NEXT: vorps %ymm7, %ymm9, %ymm7
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm6 = xmm6[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,2,3,4,5,5,7]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm8[1],xmm6[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm5, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm4, %ymm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3,4,5,5,7]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm4 = xmm6[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm6[5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm6 = xmm6[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3]
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm6, %ymm4
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm8 = xmm8[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm8, %xmm8
-; AVX1-ONLY-NEXT: vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = mem[0,1,2,3,4,5,5,7]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm8 = xmm9[1],xmm8[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm5, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vorps %ymm4, %ymm6, %ymm4
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,2,3,4,5,5,7]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm8[1],xmm6[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm8, %xmm8
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm8[5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm6[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm8 = xmm8[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm10, %xmm8
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm8, %ymm6
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm8, %xmm8
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm9 = xmm9[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm9, %xmm9
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm10 = xmm15[0,1,2,3,4,5,5,7]
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm4, %xmm9
+; AVX1-ONLY-NEXT: vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = mem[0,1,2,3,4,5,5,7]
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm9 = xmm10[1],xmm9[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1],xmm9[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm5, %ymm6
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm8, %ymm5
-; AVX1-ONLY-NEXT: vorps %ymm6, %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm7, %xmm0
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm1, %xmm6
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm14[0,1,2,3,4,5,5,7]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm6[1]
+; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1,2,3,4],xmm10[5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm7[0,1,2,3,4],ymm9[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm7
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm10 = xmm10[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm10[0],xmm7[0],xmm10[1],xmm7[1],xmm10[2],xmm7[2],xmm10[3],xmm7[3]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm12, %xmm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm10, %ymm7
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm11 = xmm11[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm4, %xmm11
+; AVX1-ONLY-NEXT: vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = mem[0,1,2,3,4,5,5,7]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm11 = xmm12[1],xmm11[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1],xmm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm8, %ymm7
+; AVX1-ONLY-NEXT: vandps %ymm8, %ymm10, %ymm10
+; AVX1-ONLY-NEXT: vorps %ymm7, %ymm10, %ymm7
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vpshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,1,2,3,4,5,5,7]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm10 = xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm4, %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm10 = xmm10[0,1,2,3,4],xmm11[5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm10[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm11 = xmm11[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm4, %xmm11
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm11, %ymm10
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm12 = xmm12[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm12[0],xmm11[0],xmm12[1],xmm11[1],xmm12[2],xmm11[2],xmm12[3],xmm11[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm4, %xmm12
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,5,7]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm12[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm11[0,1],xmm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandnps %ymm10, %ymm8, %ymm10
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm8, %ymm6
+; AVX1-ONLY-NEXT: vorps %ymm6, %ymm10, %ymm6
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,5,7]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm5 = xmm5[1],xmm10[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm10[5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm6
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm10 = xmm14[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm10[0],xmm6[0],xmm10[1],xmm6[1],xmm10[2],xmm6[2],xmm10[3],xmm6[3]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm15, %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm6, %xmm6
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm10 = xmm13[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm10[0],xmm6[0],xmm10[1],xmm6[1],xmm10[2],xmm6[2],xmm10[3],xmm6[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm10, %xmm10
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,5,7]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm10[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm6[0,1],xmm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm8, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm8, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm2, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,7]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4],xmm0[5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -7460,136 +7457,134 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r9)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 64(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm9, (%rax)
; AVX1-ONLY-NEXT: addq $1368, %rsp # imm = 0x558
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-SLOW-LABEL: load_i16_stride6_vf64:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $1256, %rsp # imm = 0x4E8
+; AVX2-SLOW-NEXT: subq $1272, %rsp # imm = 0x4F8
+; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm8
+; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm9
+; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovaps 672(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovaps 640(%rdi), %ymm3
; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm4
; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %ymm8
-; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm10
-; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %ymm6
; AVX2-SLOW-NEXT: vmovdqa 448(%rdi), %ymm7
; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm13 = ymm7[2,3],ymm6[2,3]
; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm6 = ymm7[0,1],ymm6[0,1]
-; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm7 = ymm5[2,3],ymm4[2,3]
-; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm11 = ymm5[0,1],ymm4[0,1]
+; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm15 = ymm7[0,1],ymm6[0,1]
+; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm12 = ymm5[2,3],ymm4[2,3]
+; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm14 = ymm5[0,1],ymm4[0,1]
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm3[2,3],ymm2[2,3]
; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm3[0,1],ymm2[0,1]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm5 = ymm0[2,3],ymm1[2,3]
-; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm3 = ymm0[0,1],ymm1[0,1]
+; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm3 = ymm0[2,3],ymm1[2,3]
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm9 = [8,9,4,5,4,5,6,7,0,1,4,5,0,1,12,13,24,25,20,21,20,21,22,23,16,17,20,21,16,17,28,29]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1],ymm10[2],ymm8[3,4],ymm10[5],ymm8[6,7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = <8,9,u,u,0,1,12,13,u,u,12,13,12,13,14,15>
-; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm4, %xmm0
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm1
-; AVX2-SLOW-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
+; AVX2-SLOW-NEXT: vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm1[0,1]
+; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm11 = [8,9,4,5,4,5,6,7,0,1,4,5,0,1,12,13,24,25,20,21,20,21,22,23,16,17,20,21,16,17,28,29]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1],ymm8[2],ymm9[3,4],ymm8[5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm7 = <8,9,u,u,0,1,12,13,u,u,12,13,12,13,14,15>
+; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm2, %xmm0
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm9
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm9[2,2,2,2,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0],ymm6[1],ymm13[2,3,4,5],ymm6[6],ymm13[7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpshufb %ymm9, %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm14 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
-; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm10, %ymm1
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm0, %ymm1, %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm15 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm15, %xmm0
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm15, %xmm2
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm2[2,2,2,2,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm4, %xmm0
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm6
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm6[2,2,2,2,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0],ymm3[1],ymm5[2,3,4,5],ymm3[6],ymm5[7]
-; AVX2-SLOW-NEXT: vpshufb %ymm9, %ymm3, %ymm1
-; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0],ymm15[1],ymm13[2,3,4,5],ymm15[6],ymm13[7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm0, %ymm1, %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm5, %xmm0
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm7
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm7[2,2,2,2,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm13, %xmm0
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm13, %xmm5
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm5[2,2,2,2,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0],ymm11[1],mem[2,3,4,5],ymm11[6],mem[7]
-; AVX2-SLOW-NEXT: vpshufb %ymm9, %ymm6, %ymm1
-; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm15 = ymm12[0],ymm14[1],ymm12[2,3,4,5],ymm14[6],ymm12[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm15, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm0, %ymm1, %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm0, %xmm1
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm12
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm13 = xmm12[2,2,2,2,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm13 = xmm1[0],xmm13[1],xmm1[2,3],xmm13[4],xmm1[5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm0, %xmm1
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm7
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm12 = xmm7[2,2,2,2,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm12 = xmm1[0],xmm12[1],xmm1[2,3],xmm12[4],xmm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0],ymm10[1],ymm8[2,3,4,5],ymm10[6],ymm8[7]
-; AVX2-SLOW-NEXT: vpshufb %ymm9, %ymm1, %ymm9
-; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm13, %ymm9, %ymm9
-; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm9 = <10,11,u,u,2,3,14,15,u,u,10,11,12,13,14,15>
-; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm15, %xmm13
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,2,3]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0],xmm2[1],xmm13[2,3],xmm2[4],xmm13[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm8[0],mem[1],ymm8[2,3,4,5],mem[6],ymm8[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm1, %ymm11
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm12, %ymm11, %ymm11
+; AVX2-SLOW-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm11 = <10,11,u,u,2,3,14,15,u,u,10,11,12,13,14,15>
+; AVX2-SLOW-NEXT: vpshufb %xmm11, %xmm2, %xmm2
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[1,1,2,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,5,5,5]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm9 = xmm2[0],xmm9[1],xmm2[2,3],xmm9[4],xmm2[5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = [10,11,6,7,4,5,6,7,6,7,6,7,2,3,14,15,26,27,22,23,20,21,22,23,22,23,22,23,18,19,30,31]
-; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm3, %ymm3
-; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm13, %ymm3, %ymm3
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm4, %xmm3
-; AVX2-SLOW-NEXT: vpshufd $229, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm4 = mem[1,1,2,3]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3],xmm4[4],xmm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm3, %ymm4, %ymm3
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm5, %xmm3
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm7[1,1,2,3]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3],xmm4[4],xmm3[5,6,7]
-; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm6, %ymm4
-; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm3, %ymm4, %ymm3
+; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm10, %ymm10
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm9, %ymm10, %ymm9
+; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufb %xmm11, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[1,1,2,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,5,5,5]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm6[1],xmm4[2,3],xmm6[4],xmm4[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm6, %ymm6
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm9
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm4, %ymm6, %ymm4
+; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpshufb %xmm11, %xmm13, %xmm4
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm5[1,1,2,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,5,5,5]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm15, %ymm4
+; AVX2-SLOW-NEXT: vpblendvb %ymm9, %ymm3, %ymm4, %ymm3
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[1,1,2,3]
+; AVX2-SLOW-NEXT: vpshufb %xmm11, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm7[1,1,2,3]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3],xmm2[4],xmm0[5,6,7]
-; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm0, %ymm1, %ymm0
+; AVX2-SLOW-NEXT: vpblendvb %ymm9, %ymm0, %ymm1, %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vpblendd $109, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
@@ -7603,8 +7598,8 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4,5],xmm1[6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21>
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3,4,5],ymm1[6],mem[7]
+; AVX2-SLOW-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4,5],mem[6],ymm1[7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm1, %ymm1
; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm5
@@ -7634,11 +7629,11 @@
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm4, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4,5],xmm1[6,7]
-; AVX2-SLOW-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = ymm11[0],mem[1],ymm11[2,3,4,5],mem[6],ymm11[7]
+; AVX2-SLOW-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm14[0],mem[1],ymm14[2,3,4,5],mem[6],ymm14[7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufb %ymm5, %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm11
+; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm10
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7646,11 +7641,11 @@
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm2[2,2,2,2,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm15 = xmm5[0,1,2,2]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm5[0,1,2,2]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm1
; AVX2-SLOW-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm1, %xmm14
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm14 = xmm14[0,1,2],xmm15[3],xmm14[4,5],xmm15[6],xmm14[7]
+; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm1, %xmm15
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm14 = xmm15[0,1,2],xmm14[3],xmm15[4,5],xmm14[6],xmm15[7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7],ymm0[8,9,10],ymm14[11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm14[4,5,6,7]
@@ -7660,14 +7655,15 @@
; AVX2-SLOW-NEXT: # ymm1 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5,6],ymm0[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm0
; AVX2-SLOW-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm15 = xmm0[0,2,0,3]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,6,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm0[0,2,0,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,6,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm1, %xmm13
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1],xmm15[2],xmm13[3],xmm15[4,5],xmm13[6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm8[1],ymm10[2,3,4,5],ymm8[6],ymm10[7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1],xmm14[2],xmm13[3],xmm14[4,5],xmm13[6,7]
+; AVX2-SLOW-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = mem[0],ymm8[1],mem[2,3,4,5],ymm8[6],mem[7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm0, %ymm12
-; AVX2-SLOW-NEXT: vmovdqa %ymm11, %ymm14
+; AVX2-SLOW-NEXT: vpshufb %ymm10, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vmovdqa %ymm10, %ymm14
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 736(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7704,24 +7700,24 @@
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm9 = mem[0],ymm9[1],mem[2,3,4,5],ymm9[6],mem[7]
-; AVX2-SLOW-NEXT: vpshufb %ymm14, %ymm9, %ymm14
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %ymm14, %ymm9, %ymm15
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm0[0,1,2],ymm7[3,4,5,6,7],ymm0[8,9,10],ymm7[11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-SLOW-NEXT: vpshufb {{.*#+}} xmm7 = xmm0[u,u,u,u,10,11,u,u,2,3,14,15,u,u,u,u]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm14 = <2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15>
-; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm6, %xmm6
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm15 = <2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15>
+; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm6, %xmm6
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1],xmm7[2],xmm6[3],xmm7[4,5],xmm6[6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm0, %ymm15
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm15 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
+; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm0, %ymm14
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm14 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,5,5,5]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3],xmm0[4,5],xmm3[6],xmm0[7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
@@ -7731,13 +7727,13 @@
; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm6 = [2,3,14,15,10,11,0,0,2,3,14,15,10,11,0,0]
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm4, %xmm3
+; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm4, %xmm3
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0,1],xmm0[2],xmm3[3],xmm0[4,5],xmm3[6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm3, %ymm3
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm3, %xmm3
+; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm3, %xmm3
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3],xmm3[4,5],xmm2[6],xmm3[7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
@@ -7746,23 +7742,23 @@
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4,5],xmm1[6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm1, %ymm1
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm8, %xmm1
+; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm8, %xmm1
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm13[0,1,2,3,5,5,5,5]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3],xmm1[4,5],xmm2[6],xmm1[7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5,6,7],ymm0[8,9,10],ymm1[11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm11, %xmm0
+; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm11, %xmm0
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm10[0,1,2,3,5,5,5,5]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3],xmm0[4,5],xmm1[6],xmm0[7]
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm5, %xmm1
-; AVX2-SLOW-NEXT: vpshufb %xmm14, %xmm12, %xmm2
+; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm12, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2],xmm2[3],xmm1[4,5],xmm2[6,7]
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm9, %ymm2
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
@@ -7780,8 +7776,9 @@
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; AVX2-SLOW-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm0[2,1,0,3]
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm14[0,0,0,0,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; AVX2-SLOW-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,4]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5,6],xmm1[7]
@@ -7840,8 +7837,8 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,1],ymm0[2],mem[3],ymm0[4],mem[5,6],ymm0[7]
+; AVX2-SLOW-NEXT: vpblendd $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm0[0,1],mem[2],ymm0[3],mem[4],ymm0[5,6],mem[7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm10 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
@@ -7850,15 +7847,15 @@
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm10 = xmm10[2,1,0,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm13 = xmm10[0,0,0,0,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm15 = xmm12[0,1,2,3,6,5,6,4]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3],xmm15[4],xmm13[5,6],xmm15[7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm14 = xmm12[0,1,2,3,6,5,6,4]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3],xmm14[4],xmm13[5,6],xmm14[7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm15 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm15[2,1,2,3]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm15, %xmm15
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm15 = xmm15[0,3,2,1]
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm11 = xmm15[0,0,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm14 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm14[2,1,2,3]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm14, %xmm14
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm14 = xmm14[0,3,2,1]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm11 = xmm14[0,0,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[0,1,3,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm6 = xmm1[2,1,2,0,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0],xmm11[1,2],xmm6[3],xmm11[4,5,6,7]
@@ -7869,9 +7866,9 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm13[5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm11[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = ymm15[0,1],mem[2],ymm15[3,4],mem[5],ymm15[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm6, %xmm11
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm13 = xmm11[0,1,2,1]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[2,1,0,3]
@@ -7901,7 +7898,8 @@
; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufhw $103, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm5 = mem[0,1,2,3,7,5,6,5]
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm8 = xmm14[1,1,1,1,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm8 = mem[1,1,1,1,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,7,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm8[0,1,2,3],xmm5[4],xmm8[5,6],xmm5[7]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[3,1,2,1,4,5,6,7]
@@ -7922,7 +7920,7 @@
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,7,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm7[0,1,2,3],xmm5[4],xmm7[5,6],xmm5[7]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,1,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm7 = xmm15[0,1,3,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm7 = xmm14[0,1,3,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,7,7,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm7[1,2],xmm1[3],xmm7[4,5,6,7]
; AVX2-SLOW-NEXT: vpshufb %ymm8, %ymm3, %ymm3
@@ -7964,9 +7962,8 @@
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm5[0,1,3,2]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm3[5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $146, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6],mem[7]
+; AVX2-SLOW-NEXT: vpblendd $109, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = mem[0],ymm15[1],mem[2,3],ymm15[4],mem[5,6],ymm15[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm5
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm0[0,1,0,2,4,5,6,7]
@@ -8014,13 +8011,6 @@
; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,2,3,4],ymm4[5,6,7]
; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm6 = [10,11,0,0,2,3,14,15,10,11,0,0,2,3,14,15]
-; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm9, %xmm9
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,1,1,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[0,1,3,3]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1,2,3],xmm9[4],xmm8[5],xmm9[6,7]
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,2,3,4],ymm8[5,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm5, %xmm5
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,3,3]
@@ -8028,13 +8018,20 @@
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,2,3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm12, %xmm5
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm9 = xmm11[0,1,1,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[0,1,3,3]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm9[0,1,2,3],xmm5[4],xmm9[5],xmm5[6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm9, %xmm5
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,1,1,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[0,1,3,3]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm8[0,1,2,3],xmm5[4],xmm8[5],xmm5[6,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm12, %xmm8
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm9 = xmm11[0,1,1,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[0,1,3,3]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm8 = xmm9[0,1,2,3],xmm8[4],xmm9[5],xmm8[6,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,2,3,4],ymm8[5,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm15, %xmm6
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm9 = xmm14[0,1,1,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[0,1,3,3]
@@ -8073,131 +8070,129 @@
; AVX2-SLOW-NEXT: vmovdqa %ymm2, (%r8)
; AVX2-SLOW-NEXT: vmovdqa %ymm4, 96(%r9)
; AVX2-SLOW-NEXT: vmovdqa %ymm13, 32(%r9)
-; AVX2-SLOW-NEXT: vmovdqa %ymm10, (%r9)
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, 64(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm10, 64(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, (%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT: vmovdqa %ymm6, 96(%rax)
-; AVX2-SLOW-NEXT: vmovdqa %ymm5, 32(%rax)
-; AVX2-SLOW-NEXT: vmovdqa %ymm0, 64(%rax)
-; AVX2-SLOW-NEXT: vmovdqa %ymm8, (%rax)
-; AVX2-SLOW-NEXT: addq $1256, %rsp # imm = 0x4E8
+; AVX2-SLOW-NEXT: vmovdqa %ymm8, 32(%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm5, 64(%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, (%rax)
+; AVX2-SLOW-NEXT: addq $1272, %rsp # imm = 0x4F8
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: load_i16_stride6_vf64:
; AVX2-FAST: # %bb.0:
; AVX2-FAST-NEXT: subq $1272, %rsp # imm = 0x4F8
+; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm8
+; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm9
+; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovaps 672(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovaps 640(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm4
; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm5
-; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %ymm8
-; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm10
-; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %ymm6
; AVX2-FAST-NEXT: vmovdqa 448(%rdi), %ymm7
-; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm11 = ymm7[2,3],ymm6[2,3]
-; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm7 = ymm7[0,1],ymm6[0,1]
-; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm13 = ymm7[2,3],ymm6[2,3]
+; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm6 = ymm7[0,1],ymm6[0,1]
; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm12 = ymm5[2,3],ymm4[2,3]
; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm14 = ymm5[0,1],ymm4[0,1]
+; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm15 = ymm5[0,1],ymm4[0,1]
; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm3[2,3],ymm2[2,3]
; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm3[0,1],ymm2[0,1]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm5 = ymm0[2,3],ymm1[2,3]
-; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm3 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm3 = ymm0[2,3],ymm1[2,3]
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = [8,9,4,5,4,5,6,7,0,1,4,5,0,1,12,13,24,25,20,21,20,21,22,23,16,17,20,21,16,17,28,29]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1],ymm10[2],ymm8[3,4],ymm10[5],ymm8[6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = <8,9,u,u,0,1,12,13,u,u,12,13,12,13,14,15>
-; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm4, %xmm0
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm1
-; AVX2-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
+; AVX2-FAST-NEXT: vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = [8,9,4,5,4,5,6,7,0,1,4,5,0,1,12,13,24,25,20,21,20,21,22,23,16,17,20,21,16,17,28,29]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1],ymm8[2],ymm9[3,4],ymm8[5],ymm9[6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <8,9,u,u,0,1,12,13,u,u,12,13,12,13,14,15>
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm2, %xmm0
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm9
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm1 = xmm9[2,2,2,2,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0],ymm7[1],ymm11[2,3,4,5],ymm7[6],ymm11[7]
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
+; AVX2-FAST-NEXT: vpshufb %ymm11, %ymm10, %ymm1
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm0, %ymm1, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm15 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm15, %xmm0
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm15, %xmm2
-; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm1 = xmm2[2,2,2,2,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm4, %xmm0
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm8
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm1 = xmm8[2,2,2,2,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0],ymm3[1],ymm5[2,3,4,5],ymm3[6],ymm5[7]
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm3, %ymm1
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0],ymm6[1],ymm13[2,3,4,5],ymm6[6],ymm13[7]
+; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufb %ymm11, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm0, %ymm1, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm5, %xmm0
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm7
-; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm1 = xmm7[2,2,2,2,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm13, %xmm0
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm13, %xmm3
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm1 = xmm3[2,2,2,2,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm12[0],ymm14[1],ymm12[2,3,4,5],ymm14[6],ymm12[7]
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm8, %ymm1
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm12[0],ymm15[1],ymm12[2,3,4,5],ymm15[6],ymm12[7]
+; AVX2-FAST-NEXT: vpshufb %ymm11, %ymm5, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm0, %ymm1, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm0, %xmm1
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm6
-; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm13 = xmm6[2,2,2,2,4,5,6,7]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm1[0],xmm13[1],xmm1[2,3],xmm13[4],xmm1[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0],ymm12[1],ymm11[2,3,4,5],ymm12[6],ymm11[7]
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm1, %ymm9
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm13, %ymm9, %ymm9
-; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} xmm9 = [10,11,6,7,10,11,6,7,10,11,6,7,10,11,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm2, %xmm13
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = <10,11,u,u,2,3,14,15,u,u,10,11,12,13,14,15>
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm15, %xmm15
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm15[0],xmm13[1],xmm15[2,3],xmm13[4],xmm15[5,6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = [10,11,6,7,4,5,6,7,6,7,6,7,2,3,14,15,26,27,22,23,20,21,22,23,22,23,22,23,18,19,30,31]
-; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm3, %ymm3
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm13, %ymm3, %ymm3
-; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm3, %xmm3
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm4, %xmm4
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm3, %ymm4, %ymm3
-; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm7, %xmm3
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm5, %xmm4
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6,7]
-; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm8, %ymm4
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm3, %ymm4, %ymm3
-; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm6, %xmm3
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2,3],xmm3[4],xmm0[5,6,7]
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm0, %xmm1
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm14
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm12 = xmm14[2,2,2,2,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm12 = xmm1[0],xmm12[1],xmm1[2,3],xmm12[4],xmm1[5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3,4,5],ymm1[6],mem[7]
+; AVX2-FAST-NEXT: vpshufb %ymm11, %ymm1, %ymm11
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm12, %ymm11, %ymm11
+; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} xmm11 = [10,11,6,7,10,11,6,7,10,11,6,7,10,11,6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm9, %xmm12
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm9 = <10,11,u,u,2,3,14,15,u,u,10,11,12,13,14,15>
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm2, %xmm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm12[1],xmm2[2,3],xmm12[4],xmm2[5,6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm12 = [10,11,6,7,4,5,6,7,6,7,6,7,2,3,14,15,26,27,22,23,20,21,22,23,22,23,22,23,18,19,30,31]
+; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm10, %ymm10
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm2, %ymm10, %ymm2
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm8, %xmm2
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm4, %xmm4
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0],xmm2[1],xmm4[2,3],xmm2[4],xmm4[5,6,7]
+; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm4, %ymm4
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm2, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm3, %xmm2
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm13, %xmm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2,3],xmm2[4],xmm3[5,6,7]
+; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm5, %ymm3
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm2, %ymm3, %ymm2
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm14, %xmm2
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm0, %xmm0
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3],xmm2[4],xmm0[5,6,7]
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm0, %ymm1, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $109, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
@@ -8207,21 +8202,20 @@
; AVX2-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = <0,1,12,13,u,u,4,5,u,u,u,u,12,13,14,15>
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm7, %xmm0
-; AVX2-FAST-NEXT: vmovdqa %xmm2, %xmm13
+; AVX2-FAST-NEXT: vmovdqa %xmm2, %xmm11
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm3 = [8,9,12,13,0,1,0,0,8,9,12,13,0,1,0,0]
; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm1, %xmm2
; AVX2-FAST-NEXT: vmovdqa %xmm3, %xmm8
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3],xmm2[4,5],xmm0[6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21>
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3,4,5],ymm1[6],mem[7]
+; AVX2-FAST-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm6[0],mem[1],ymm6[2,3,4,5],mem[6],ymm6[7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm3
; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm4
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
@@ -8243,16 +8237,16 @@
; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm0
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,1,0,3]
; AVX2-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm5, %xmm0
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm5, %xmm0
; AVX2-FAST-NEXT: vmovdqa %xmm8, %xmm6
; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm1, %xmm8
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm8[2],xmm0[3],xmm8[4,5],xmm0[6,7]
-; AVX2-FAST-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm14[0],mem[1],ymm14[2,3,4,5],mem[6],ymm14[7]
+; AVX2-FAST-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm15[0],mem[1],ymm15[2,3,4,5],mem[6],ymm15[7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm2
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm15
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm15 = ymm0[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm14
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm1
@@ -8261,23 +8255,26 @@
; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm4, %xmm0
; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm1
; AVX2-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm1, %xmm14
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm14[0,1,2],xmm0[3],xmm14[4,5],xmm0[6],xmm14[7]
+; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm1, %xmm15
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3],xmm15[4,5],xmm0[6],xmm15[7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3,4,5,6,7],ymm15[8,9,10],ymm0[11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3,4,5,6,7],ymm14[8,9,10],ymm0[11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $109, (%rsp), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpblendd $109, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5,6],ymm0[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm14
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm14[2,1,0,3]
; AVX2-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm3, %xmm15
-; AVX2-FAST-NEXT: vmovdqa %xmm13, %xmm8
+; AVX2-FAST-NEXT: vmovdqa %xmm11, %xmm15
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm3, %xmm14
; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm0, %xmm13
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm15[0,1],xmm13[2],xmm15[3],xmm13[4,5],xmm15[6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0],ymm11[1],ymm12[2,3,4,5],ymm11[6],ymm12[7]
+; AVX2-FAST-NEXT: vmovdqa %xmm6, %xmm8
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm14[0,1],xmm13[2],xmm14[3],xmm13[4,5],xmm14[6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5],mem[6],ymm0[7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm12
; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm14
@@ -8307,16 +8304,16 @@
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $146, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm10 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6],mem[7]
-; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm10, %xmm1
+; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm10, %xmm1
; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm12
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm12 = xmm12[2,1,0,3]
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm12[u,u,u,u,0,1,u,u,8,9,12,13,u,u,u,u]
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm12, %xmm8
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm8[2],xmm1[3],xmm8[4,5],xmm1[6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm8 = mem[0],ymm0[1],mem[2,3,4,5],ymm0[6],mem[7]
-; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm8, %ymm14
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm8, %ymm15
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm9 = ymm1[0,1,2],ymm9[3,4,5,6,7],ymm1[8,9,10],ymm9[11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm9[4,5,6,7]
@@ -8324,15 +8321,15 @@
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15]
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm9 = [10,11,14,15,2,3,0,0,10,11,14,15,2,3,0,0]
; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm0, %xmm14
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1],xmm14[2],xmm7[3],xmm14[4,5],xmm7[6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm0, %xmm15
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1],xmm15[2],xmm7[3],xmm15[4,5],xmm7[6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm0, %ymm15
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm0, %ymm14
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm0, %xmm0
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm0, %xmm0
; AVX2-FAST-NEXT: vpshufhw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm1 = mem[0,1,2,3,5,5,5,5]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3],xmm0[4,5],xmm1[6],xmm0[7]
@@ -8346,10 +8343,10 @@
; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm1, %xmm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3],xmm1[4,5],xmm0[6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm1, %xmm1
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3],xmm1[4,5],xmm4[6],xmm1[7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
@@ -8361,22 +8358,22 @@
; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm1, %xmm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3],xmm1[4,5],xmm0[6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm6, %xmm1
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm6, %xmm1
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm3 = xmm13[0,1,2,3,5,5,5,5]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm3[3],xmm1[4,5],xmm3[6],xmm1[7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5,6,7],ymm0[8,9,10],ymm1[11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm11, %xmm0
+; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm11, %xmm0
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,5,5,5,5]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3],xmm0[4,5],xmm1[6],xmm0[7]
; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm10, %xmm1
; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm12, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3],xmm2[4,5],xmm1[6,7]
-; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm8, %ymm2
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm8, %ymm2
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
@@ -8387,7 +8384,7 @@
; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm0[2],mem[3],ymm0[4],mem[5,6],ymm0[7]
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpblendd $219, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm0 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,1,0,3]
; AVX2-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8419,8 +8416,8 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1],ymm0[2],mem[3],ymm0[4],mem[5,6],ymm0[7]
+; AVX2-FAST-NEXT: vpblendd $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm0[0,1],mem[2],ymm0[3],mem[4],ymm0[5,6],mem[7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
@@ -8434,33 +8431,33 @@
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm12 = xmm3[0,1,2,3,6,5,6,4]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm12 = xmm2[0,1,2,3],xmm12[4],xmm2[5,6],xmm12[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $219, (%rsp), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm13
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm13[0,3,2,1]
-; AVX2-FAST-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm0, %xmm13
-; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm15 = xmm2[2,1,2,0,4,5,6,7]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm15[0],xmm13[1,2],xmm15[3],xmm13[4,5,6,7]
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm14 = xmm2[2,1,2,0,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm14[0],xmm13[1,2],xmm14[3],xmm13[4,5,6,7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm1, %ymm15
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm15[0,1,2],ymm12[3,4,5,6,7],ymm15[8,9,10],ymm12[11,12,13,14,15]
-; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,6,5,4]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4],xmm15[5,6,7]
+; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm1, %ymm14
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm12 = ymm14[0,1,2],ymm12[3,4,5,6,7],ymm14[8,9,10],ymm12[11,12,13,14,15]
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,6,5,4]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4],xmm14[5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[0,1],ymm0[2],mem[3],ymm0[4],mem[5,6],ymm0[7]
+; AVX2-FAST-NEXT: vpblendd $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm0[0,1],mem[2],ymm0[3],mem[4],ymm0[5,6],mem[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm13 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
-; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm15 = xmm13[2,1,0,3]
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm14 = xmm13[2,1,0,3]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm13, %xmm13
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,1]
-; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm15, %xmm0
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm14, %xmm0
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm11 = xmm13[0,1,2,3,6,5,6,4]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3],xmm11[4],xmm0[5,6],xmm11[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -8474,7 +8471,7 @@
; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm5 = xmm11[2,1,2,0,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0],xmm0[1,2],xmm5[3],xmm0[4,5,6,7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm14, %ymm5
+; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm15, %ymm5
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm5[0,1,2],ymm1[3,4,5,6,7],ymm5[8,9,10],ymm1[11,12,13,14,15]
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,5,4]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
@@ -8527,13 +8524,13 @@
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm4[0,1,2,3,4],xmm0[5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm15, %xmm0
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm14, %xmm0
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm1 = xmm13[0,1,2,3,7,5,6,5]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5,6],xmm1[7]
; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm6, %xmm1
; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm6 = xmm11[3,1,2,1,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm6[0],xmm1[1,2],xmm6[3],xmm1[4,5,6,7]
-; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm14, %ymm6
+; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm15, %ymm6
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3,4,5,6,7],ymm6[8,9,10],ymm0[11,12,13,14,15]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,1,3,2]
@@ -8545,7 +8542,7 @@
; AVX2-FAST-NEXT: vpshufhw $103, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm6 = mem[0,1,2,3,7,5,6,5]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm6[4],xmm0[5,6],xmm6[7]
-; AVX2-FAST-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm1, %xmm6
; AVX2-FAST-NEXT: vpshuflw $103, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm11 = mem[3,1,2,1,4,5,6,7]
@@ -8583,7 +8580,7 @@
; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = mem[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $146, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpblendd $146, (%rsp), %ymm0, %ymm9 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm9 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6],mem[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm9, %xmm10
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[0,3,2,1]
@@ -8616,25 +8613,25 @@
; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm7 = mem[0,1,2,3,4],ymm7[5,6,7]
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm8 = [10,11,0,0,2,3,14,15,10,11,0,0,2,3,14,15]
-; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm10, %xmm10
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = <0,1,2,3,2,3,6,7,u,u,14,15,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm9, %xmm9
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1,2,3],xmm10[4],xmm9[5],xmm10[6,7]
-; AVX2-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm9 = mem[0,1,2,3,4],ymm9[5,6,7]
; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm5, %xmm5
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = <0,1,2,3,2,3,6,7,u,u,14,15,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm2, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm5[4],xmm2[5],xmm5[6,7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm13, %xmm5
-; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm12, %xmm10
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm5 = xmm10[0,1,2,3],xmm5[4],xmm10[5],xmm5[6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm10, %xmm5
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm9, %xmm9
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm5 = xmm9[0,1,2,3],xmm5[4],xmm9[5],xmm5[6,7]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm5 = mem[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm13, %xmm9
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm12, %xmm10
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1,2,3],xmm9[4],xmm10[5],xmm9[6,7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
+; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,1,2,3,4],ymm9[5,6,7]
; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm0, %xmm0
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm15, %xmm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4],xmm1[5],xmm0[6,7]
@@ -8673,13 +8670,13 @@
; AVX2-FAST-NEXT: vmovdqa %ymm4, (%r8)
; AVX2-FAST-NEXT: vmovdqa %ymm7, 96(%r9)
; AVX2-FAST-NEXT: vmovdqa %ymm14, 32(%r9)
-; AVX2-FAST-NEXT: vmovdqa %ymm11, (%r9)
-; AVX2-FAST-NEXT: vmovdqa %ymm3, 64(%r9)
+; AVX2-FAST-NEXT: vmovdqa %ymm11, 64(%r9)
+; AVX2-FAST-NEXT: vmovdqa %ymm3, (%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-NEXT: vmovdqa %ymm0, 96(%rax)
-; AVX2-FAST-NEXT: vmovdqa %ymm5, 32(%rax)
-; AVX2-FAST-NEXT: vmovdqa %ymm2, 64(%rax)
-; AVX2-FAST-NEXT: vmovdqa %ymm9, (%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm9, 32(%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm5, 64(%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm2, (%rax)
; AVX2-FAST-NEXT: addq $1272, %rsp # imm = 0x4F8
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
@@ -8687,117 +8684,115 @@
; AVX2-FAST-PERLANE-LABEL: load_i16_stride6_vf64:
; AVX2-FAST-PERLANE: # %bb.0:
; AVX2-FAST-PERLANE-NEXT: subq $1272, %rsp # imm = 0x4F8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovaps 672(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovaps 640(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %ymm4
; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm10
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vmovdqa 448(%rdi), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm11 = ymm7[2,3],ymm6[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm7 = ymm7[0,1],ymm6[0,1]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm13 = ymm7[2,3],ymm6[2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm6 = ymm7[0,1],ymm6[0,1]
; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm12 = ymm5[2,3],ymm4[2,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm14 = ymm5[0,1],ymm4[0,1]
+; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm15 = ymm5[0,1],ymm4[0,1]
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm3[2,3],ymm2[2,3]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm3[0,1],ymm2[0,1]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm5 = ymm0[2,3],ymm1[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm3 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm3 = ymm0[2,3],ymm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm9 = [8,9,4,5,4,5,6,7,0,1,4,5,0,1,12,13,24,25,20,21,20,21,22,23,16,17,20,21,16,17,28,29]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1],ymm10[2],ymm8[3,4],ymm10[5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm6 = <8,9,u,u,0,1,12,13,u,u,12,13,12,13,14,15>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm4, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vperm2i128 {{.*#+}} ymm4 = ymm0[0,1],ymm1[0,1]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm11 = [8,9,4,5,4,5,6,7,0,1,4,5,0,1,12,13,24,25,20,21,20,21,22,23,16,17,20,21,16,17,28,29]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1],ymm8[2],ymm9[3,4],ymm8[5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm14 = <8,9,u,u,0,1,12,13,u,u,12,13,12,13,14,15>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm2, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm1 = xmm9[2,2,2,2,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0],ymm7[1],ymm11[2,3,4,5],ymm7[6],ymm11[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm9, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm10 = ymm3[0],ymm4[1],ymm3[2,3,4,5],ymm4[6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm11, %ymm10, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm7, %ymm0, %ymm1, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm15 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm15, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm15, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm1 = xmm2[2,2,2,2,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm4, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm1 = xmm8[2,2,2,2,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0],ymm3[1],ymm5[2,3,4,5],ymm3[6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm9, %ymm3, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0],ymm6[1],ymm13[2,3,4,5],ymm6[6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm11, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm7, %ymm0, %ymm1, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm5, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm1 = xmm7[2,2,2,2,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm13, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm13, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm1 = xmm3[2,2,2,2,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm8 = ymm12[0],ymm14[1],ymm12[2,3,4,5],ymm14[6],ymm12[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm9, %ymm8, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm12[0],ymm15[1],ymm12[2,3,4,5],ymm15[6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm11, %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm7, %ymm0, %ymm1, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm0, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm13 = xmm6[2,2,2,2,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm13 = xmm1[0],xmm13[1],xmm1[2,3],xmm13[4],xmm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0],ymm12[1],ymm11[2,3,4,5],ymm12[6],ymm11[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm9, %ymm1, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm13, %ymm9, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} xmm9 = [10,11,6,7,10,11,6,7,10,11,6,7,10,11,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm2, %xmm13
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = <10,11,u,u,2,3,14,15,u,u,10,11,12,13,14,15>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm15, %xmm15
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm13 = xmm15[0],xmm13[1],xmm15[2,3],xmm13[4],xmm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm15 = [10,11,6,7,4,5,6,7,6,7,6,7,2,3,14,15,26,27,22,23,20,21,22,23,22,23,22,23,18,19,30,31]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm13, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm4, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm3, %ymm4, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm7, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm5, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3],xmm3[4],xmm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm8, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm3, %ymm4, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm6, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2,3],xmm3[4],xmm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm0, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm14
+; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm12 = xmm14[2,2,2,2,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm12 = xmm1[0],xmm12[1],xmm1[2,3],xmm12[4],xmm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3,4,5],ymm1[6],mem[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm11, %ymm1, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm7, %ymm12, %ymm11, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} xmm11 = [10,11,6,7,10,11,6,7,10,11,6,7,10,11,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm9, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm9 = <10,11,u,u,2,3,14,15,u,u,10,11,12,13,14,15>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm2, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm12[1],xmm2[2,3],xmm12[4],xmm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm12 = [10,11,6,7,4,5,6,7,6,7,6,7,2,3,14,15,26,27,22,23,20,21,22,23,22,23,22,23,18,19,30,31]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm12, %ymm10, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm7, %ymm2, %ymm10, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm8, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0],xmm2[1],xmm4[2,3],xmm2[4],xmm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm12, %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm7, %ymm2, %ymm4, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm3, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm13, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2,3],xmm2[4],xmm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm12, %ymm5, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm7, %ymm2, %ymm3, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm12, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm14, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3],xmm2[4],xmm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm7, %ymm0, %ymm1, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendd $109, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
@@ -8807,21 +8802,20 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = <0,1,12,13,u,u,4,5,u,u,u,u,12,13,14,15>
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm7, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, %xmm13
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, %xmm11
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm3 = [8,9,12,13,0,1,0,0,8,9,12,13,0,1,0,0]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, %xmm8
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3],xmm2[4,5],xmm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,8,9,4,5,16,17,28,29,24,25,28,29,24,25,28,29,24,25,20,21>
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3,4,5],ymm1[6],mem[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm6[0],mem[1],ymm6[2,3,4,5],mem[6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm4
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
@@ -8843,16 +8837,16 @@
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm5, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm5, %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm8, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm1, %xmm8
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm8[2],xmm0[3],xmm8[4,5],xmm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm14[0],mem[1],ymm14[2,3,4,5],mem[6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm15[0],mem[1],ymm15[2,3,4,5],mem[6],ymm15[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm1, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm15 = ymm0[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm1, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm1
@@ -8861,23 +8855,26 @@
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm4, %xmm0
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm10, %xmm1, %xmm14
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm14[0,1,2],xmm0[3],xmm14[4,5],xmm0[6],xmm14[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm10, %xmm1, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3],xmm15[4,5],xmm0[6],xmm15[7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3,4,5,6,7],ymm15[8,9,10],ymm0[11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3,4,5,6,7],ymm14[8,9,10],ymm0[11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $109, (%rsp), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $109, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm14
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = xmm14[2,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm3, %xmm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm13, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm11, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm3, %xmm14
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm0, %xmm13
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm13 = xmm15[0,1],xmm13[2],xmm15[3],xmm13[4,5],xmm15[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0],ymm11[1],ymm12[2,3,4,5],ymm11[6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm6, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm13 = xmm14[0,1],xmm13[2],xmm14[3],xmm13[4,5],xmm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $66, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5],mem[6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm12
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm14
@@ -8907,16 +8904,16 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendd $146, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm10 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm10 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6],mem[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm10, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm10, %xmm1
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm10, %xmm12
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm12 = xmm12[2,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm8 = xmm12[u,u,u,u,0,1,u,u,8,9,12,13,u,u,u,u]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm12, %xmm8
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm8[2],xmm1[3],xmm8[4,5],xmm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendd $189, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0],ymm0[1],mem[2,3,4,5],ymm0[6],mem[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm8, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm8, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm9 = ymm1[0,1,2],ymm9[3,4,5,6,7],ymm1[8,9,10],ymm9[11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm9[4,5,6,7]
@@ -8924,15 +8921,15 @@
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[2,3,14,15,u,u,6,7,u,u,u,u,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm9 = [10,11,14,15,2,3,0,0,10,11,14,15,2,3,0,0]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm0, %xmm14
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1],xmm14[2],xmm7[3],xmm14[4,5],xmm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm0, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1],xmm15[2],xmm7[3],xmm15[4,5],xmm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,u,u,u,u,10,11,6,7,18,19,30,31,26,27,26,27,30,31,26,27,26,27,22,23>
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm0, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm15 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm0, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm14[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm14 = <6,7,2,3,4,5,u,u,2,3,14,15,u,u,6,7>
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpshufhw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm1 = mem[0,1,2,3,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3],xmm0[4,5],xmm1[6],xmm0[7]
@@ -8946,10 +8943,10 @@
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3],xmm1[4,5],xmm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3],xmm1[4,5],xmm4[6],xmm1[7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
@@ -8961,22 +8958,22 @@
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3],xmm1[4,5],xmm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm6, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm6, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm3 = xmm13[0,1,2,3,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm3[3],xmm1[4,5],xmm3[6],xmm1[7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5,6,7],ymm0[8,9,10],ymm1[11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm11, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm11, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3],xmm0[4,5],xmm1[6],xmm0[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm10, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm12, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3],xmm2[4,5],xmm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm8, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7],ymm1[8,9,10],ymm0[11,12,13,14,15]
@@ -8987,7 +8984,7 @@
; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1],ymm0[2],mem[3],ymm0[4],mem[5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $219, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -9019,8 +9016,8 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm0[2],mem[3],ymm0[4],mem[5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm0[0,1],mem[2],ymm0[3],mem[4],ymm0[5,6],mem[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
@@ -9034,33 +9031,33 @@
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm12 = xmm3[0,1,2,3,6,5,6,4]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm12 = xmm2[0,1,2,3],xmm12[4],xmm2[5,6],xmm12[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $219, (%rsp), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm13
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = xmm13[0,3,2,1]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm10, %xmm0, %xmm13
-; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm15 = xmm2[2,1,2,0,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm13 = xmm15[0],xmm13[1,2],xmm15[3],xmm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm14 = xmm2[2,1,2,0,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm13 = xmm14[0],xmm13[1,2],xmm14[3],xmm13[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm9, %ymm1, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm12 = ymm15[0,1,2],ymm12[3,4,5,6,7],ymm15[8,9,10],ymm12[11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,6,5,4]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4],xmm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm9, %ymm1, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm12 = ymm14[0,1,2],ymm12[3,4,5,6,7],ymm14[8,9,10],ymm12[11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,6,5,4]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4],xmm14[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm12[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $107, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[0,1],ymm0[2],mem[3],ymm0[4],mem[5,6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $148, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm0[0,1],mem[2],ymm0[3],mem[4],ymm0[5,6],mem[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm13 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm15 = xmm13[2,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm14 = xmm13[2,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm13, %xmm13
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,1]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm15, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm14, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm11 = xmm13[0,1,2,3,6,5,6,4]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3],xmm11[4],xmm0[5,6],xmm11[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -9074,7 +9071,7 @@
; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm5 = xmm11[2,1,2,0,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0],xmm0[1,2],xmm5[3],xmm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm9, %ymm14, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm9, %ymm15, %ymm5
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm5[0,1,2],ymm1[3,4,5,6,7],ymm5[8,9,10],ymm1[11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,5,4]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm5[5,6,7]
@@ -9127,13 +9124,13 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm4[0,1,2,3,4],xmm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm15, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm14, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm1 = xmm13[0,1,2,3,7,5,6,5]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5,6],xmm1[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm6, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm6 = xmm11[3,1,2,1,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm6[0],xmm1[1,2],xmm6[3],xmm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm14, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm15, %ymm6
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3,4,5,6,7],ymm6[8,9,10],ymm0[11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,1,3,2]
@@ -9145,7 +9142,7 @@
; AVX2-FAST-PERLANE-NEXT: vpshufhw $103, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm6 = mem[0,1,2,3,7,5,6,5]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm6[4],xmm0[5,6],xmm6[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm1, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpshuflw $103, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm11 = mem[3,1,2,1,4,5,6,7]
@@ -9183,7 +9180,7 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $146, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $146, (%rsp), %ymm0, %ymm9 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6],mem[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm9, %xmm10
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm9 = xmm9[0,3,2,1]
@@ -9216,25 +9213,25 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1,2,3,4],ymm7[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm8 = [10,11,0,0,2,3,14,15,10,11,0,0,2,3,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm10, %xmm10
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = <0,1,2,3,2,3,6,7,u,u,14,15,u,u,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm9, %xmm9
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm9 = xmm9[0,1,2,3],xmm10[4],xmm9[5],xmm10[6,7]
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1,2,3,4],ymm9[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = <0,1,2,3,2,3,6,7,u,u,14,15,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm5[4],xmm2[5],xmm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm13, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm12, %xmm10
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm10[0,1,2,3],xmm5[4],xmm10[5],xmm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm10, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm9, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm9[0,1,2,3],xmm5[4],xmm9[5],xmm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm13, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm12, %xmm10
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1,2,3],xmm9[4],xmm10[5],xmm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1,2,3,4],ymm9[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm15, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4],xmm1[5],xmm0[6,7]
@@ -9273,13 +9270,13 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, (%r8)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, 96(%r9)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm14, 32(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, (%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, (%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, 96(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, 32(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, 64(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, (%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, (%rax)
; AVX2-FAST-PERLANE-NEXT: addq $1272, %rsp # imm = 0x4F8
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
index dc38021..56f269c 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
@@ -1474,107 +1474,103 @@
; SSE-LABEL: load_i16_stride7_vf16:
; SSE: # %bb.0:
; SSE-NEXT: subq $216, %rsp
-; SSE-NEXT: movdqa 80(%rdi), %xmm11
-; SSE-NEXT: movdqa 64(%rdi), %xmm10
+; SSE-NEXT: movdqa 192(%rdi), %xmm14
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 176(%rdi), %xmm13
+; SSE-NEXT: movdqa (%rdi), %xmm7
+; SSE-NEXT: movdqa 16(%rdi), %xmm5
+; SSE-NEXT: movaps 32(%rdi), %xmm15
+; SSE-NEXT: movaps 48(%rdi), %xmm6
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 80(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 64(%rdi), %xmm11
+; SSE-NEXT: movdqa 96(%rdi), %xmm10
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[0,0,0,0]
; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rdi), %xmm6
-; SSE-NEXT: movdqa 128(%rdi), %xmm8
-; SSE-NEXT: movaps 160(%rdi), %xmm5
-; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 144(%rdi), %xmm7
-; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm13
-; SSE-NEXT: movdqa 176(%rdi), %xmm15
-; SSE-NEXT: movdqa 208(%rdi), %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
-; SSE-NEXT: movdqa %xmm1, %xmm14
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,65535,65535,0]
; SSE-NEXT: movdqa %xmm1, %xmm2
; SSE-NEXT: pandn %xmm0, %xmm2
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
+; SSE-NEXT: movdqa %xmm11, %xmm0
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT: pand %xmm1, %xmm0
; SSE-NEXT: por %xmm2, %xmm0
; SSE-NEXT: movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,65535,0,0,0]
; SSE-NEXT: movdqa %xmm3, %xmm2
; SSE-NEXT: movdqa %xmm3, %xmm9
; SSE-NEXT: pandn %xmm0, %xmm2
-; SSE-NEXT: movaps %xmm7, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm5[2,2]
+; SSE-NEXT: movaps %xmm15, %xmm0
+; SSE-NEXT: movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm6[2,2]
; SSE-NEXT: movaps {{.*#+}} xmm12 = [65535,65535,65535,0,0,65535,65535,65535]
; SSE-NEXT: movaps %xmm12, %xmm3
; SSE-NEXT: andnps %xmm0, %xmm3
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
-; SSE-NEXT: movdqa %xmm8, %xmm7
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm6[0,1,0,3]
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT: movdqa %xmm5, %xmm8
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm7[0,1,0,3]
+; SSE-NEXT: movdqa %xmm7, %xmm6
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: movdqa 96(%rdi), %xmm5
+; SSE-NEXT: movdqa 208(%rdi), %xmm5
; SSE-NEXT: pand %xmm12, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: movdqa %xmm9, %xmm7
; SSE-NEXT: pand %xmm9, %xmm4
; SSE-NEXT: por %xmm2, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[0,0,0,0]
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm1, %xmm2
-; SSE-NEXT: pandn %xmm0, %xmm2
-; SSE-NEXT: movdqa %xmm10, %xmm0
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm1, %xmm0
-; SSE-NEXT: por %xmm2, %xmm0
-; SSE-NEXT: movdqa %xmm9, %xmm2
-; SSE-NEXT: movdqa %xmm9, %xmm10
-; SSE-NEXT: pandn %xmm0, %xmm2
-; SSE-NEXT: movaps 32(%rdi), %xmm0
-; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 48(%rdi), %xmm3
-; SSE-NEXT: movaps %xmm3, (%rsp) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT: movaps %xmm12, %xmm3
-; SSE-NEXT: andnps %xmm0, %xmm3
-; SSE-NEXT: movdqa (%rdi), %xmm9
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[0,1,0,3]
-; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa 16(%rdi), %xmm8
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: pand %xmm12, %xmm4
-; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: pand %xmm10, %xmm4
-; SSE-NEXT: por %xmm2, %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm14, %xmm0
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
; SSE-NEXT: movdqa %xmm1, %xmm2
; SSE-NEXT: pandn %xmm0, %xmm2
; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: psrldq {{.*#+}} xmm15 = xmm15[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT: pand %xmm1, %xmm15
-; SSE-NEXT: por %xmm2, %xmm15
-; SSE-NEXT: movdqa %xmm10, %xmm0
-; SSE-NEXT: pandn %xmm15, %xmm0
-; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm2, %xmm3
-; SSE-NEXT: pandn %xmm6, %xmm3
-; SSE-NEXT: movdqa %xmm7, %xmm4
-; SSE-NEXT: pand %xmm2, %xmm4
-; SSE-NEXT: movdqa %xmm2, %xmm6
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT: pand %xmm1, %xmm0
+; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm9, %xmm2
+; SSE-NEXT: pandn %xmm0, %xmm2
+; SSE-NEXT: movaps 160(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 144(%rdi), %xmm9
+; SSE-NEXT: movaps %xmm9, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
+; SSE-NEXT: movaps %xmm12, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm3
+; SSE-NEXT: movdqa 112(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT: movdqa 128(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; SSE-NEXT: pand %xmm12, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm7, %xmm3
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,1,2,1]
+; SSE-NEXT: pand %xmm7, %xmm4
+; SSE-NEXT: por %xmm2, %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pslldq {{.*#+}} xmm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm10[0,1,2,3,4,5]
+; SSE-NEXT: movdqa %xmm1, %xmm2
+; SSE-NEXT: pandn %xmm10, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm10, %xmm0
+; SSE-NEXT: psrld $16, %xmm0
+; SSE-NEXT: psrldq {{.*#+}} xmm11 = xmm11[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT: pand %xmm1, %xmm11
+; SSE-NEXT: por %xmm2, %xmm11
+; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: pandn %xmm11, %xmm0
+; SSE-NEXT: movdqa {{.*#+}} xmm14 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: pandn %xmm6, %xmm3
+; SSE-NEXT: movdqa %xmm8, %xmm4
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pand %xmm14, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT: # xmm15 = xmm15[4],mem[4],xmm15[5],mem[5],xmm15[6],mem[6],xmm15[7],mem[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm15[2,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,7,7,7]
; SSE-NEXT: movdqa %xmm12, %xmm2
@@ -1583,230 +1579,234 @@
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,0,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm12, %xmm3
; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: pand %xmm10, %xmm2
+; SSE-NEXT: pand %xmm7, %xmm2
; SSE-NEXT: por %xmm0, %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5]
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm5, %xmm0
+; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
; SSE-NEXT: movdqa %xmm1, %xmm3
-; SSE-NEXT: pandn %xmm5, %xmm3
-; SSE-NEXT: movdqa %xmm11, %xmm0
+; SSE-NEXT: pandn %xmm0, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm2, %xmm0
; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm11, %xmm4
+; SSE-NEXT: movdqa %xmm13, %xmm4
; SSE-NEXT: psrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT: pand %xmm1, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: movdqa %xmm10, %xmm0
+; SSE-NEXT: movdqa %xmm7, %xmm0
; SSE-NEXT: pandn %xmm4, %xmm0
-; SSE-NEXT: movdqa %xmm6, %xmm3
-; SSE-NEXT: pandn %xmm9, %xmm3
-; SSE-NEXT: pand %xmm6, %xmm8
-; SSE-NEXT: por %xmm3, %xmm8
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm8[0,3,2,3]
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: movdqa (%rsp), %xmm6 # 16-byte Reload
+; SSE-NEXT: pandn %xmm6, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm15, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,0,3,3,4,5,6,7]
; SSE-NEXT: pand %xmm12, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm14, %xmm4
-; SSE-NEXT: movdqa (%rsp), %xmm15 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm15[4],xmm4[5],xmm15[5],xmm4[6],xmm15[6],xmm4[7],xmm15[7]
+; SSE-NEXT: movaps %xmm9, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm11[4],xmm4[5],xmm11[5],xmm4[6],xmm11[6],xmm4[7],xmm11[7]
; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[2,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,7,7,7]
; SSE-NEXT: pandn %xmm4, %xmm12
; SSE-NEXT: por %xmm3, %xmm12
-; SSE-NEXT: pand %xmm10, %xmm12
+; SSE-NEXT: pand %xmm7, %xmm12
; SSE-NEXT: por %xmm0, %xmm12
; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[0,1,0,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[0,1,0,1]
; SSE-NEXT: movdqa %xmm1, %xmm3
; SSE-NEXT: pandn %xmm0, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm8, %xmm0
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm13[2],xmm0[3],xmm13[3]
+; SSE-NEXT: movdqa %xmm13, %xmm0
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: movdqa %xmm2, %xmm4
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
; SSE-NEXT: pand %xmm1, %xmm0
; SSE-NEXT: por %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm10, %xmm3
+; SSE-NEXT: movdqa %xmm7, %xmm3
; SSE-NEXT: pandn %xmm0, %xmm3
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[0,1,0,3]
-; SSE-NEXT: movdqa %xmm7, %xmm6
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm0[0,1,2,3,4,5,4,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm5 = xmm5[1],xmm0[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[2,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
-; SSE-NEXT: movss {{.*#+}} xmm5 = xmm4[0],xmm5[1,2,3]
-; SSE-NEXT: andps %xmm10, %xmm5
-; SSE-NEXT: orps %xmm3, %xmm5
-; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm12[2],xmm3[3],xmm12[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,1]
-; SSE-NEXT: pand %xmm1, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm5[0,1,0,1]
-; SSE-NEXT: pandn %xmm4, %xmm1
-; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm10, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm14, %xmm4
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[0,1,0,3]
+; SSE-NEXT: movaps %xmm9, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT: movdqa %xmm11, %xmm12
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT: movdqa %xmm6, %xmm11
+; SSE-NEXT: punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm15[0],xmm11[1],xmm15[1],xmm11[2],xmm15[2],xmm11[3],xmm15[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
+; SSE-NEXT: andps %xmm7, %xmm2
+; SSE-NEXT: orps %xmm3, %xmm2
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm6, %xmm0
+; SSE-NEXT: movdqa %xmm10, %xmm15
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm10[2],xmm0[3],xmm10[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; SSE-NEXT: pand %xmm1, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[0,1,0,1]
+; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm7, %xmm10
+; SSE-NEXT: pandn %xmm1, %xmm10
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: movdqa %xmm15, %xmm14
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm1[2,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm11 = xmm11[0,3,2,3,4,5,6,7]
-; SSE-NEXT: movss {{.*#+}} xmm0 = xmm11[0],xmm0[1,2,3]
-; SSE-NEXT: andps %xmm10, %xmm0
-; SSE-NEXT: orps %xmm3, %xmm0
+; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm1[2,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm8[0,3,2,3,4,5,6,7]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm8[0],xmm0[1,2,3]
+; SSE-NEXT: andps %xmm7, %xmm0
+; SSE-NEXT: orps %xmm10, %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm0, %xmm3
-; SSE-NEXT: pandn %xmm6, %xmm3
-; SSE-NEXT: movdqa %xmm7, %xmm11
-; SSE-NEXT: pand %xmm0, %xmm11
-; SSE-NEXT: movdqa %xmm0, %xmm15
-; SSE-NEXT: por %xmm3, %xmm11
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm11[0,1,1,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,7,7,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT: movss {{.*#+}} xmm3 = xmm0[0],xmm3[1,2,3]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm13[4],xmm8[5],xmm13[5],xmm8[6],xmm13[6],xmm8[7],xmm13[7]
-; SSE-NEXT: movdqa %xmm13, %xmm7
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,7]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
-; SSE-NEXT: movdqa %xmm10, %xmm2
-; SSE-NEXT: pandn %xmm0, %xmm2
-; SSE-NEXT: andps %xmm10, %xmm3
-; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm4, %xmm0
-; SSE-NEXT: movdqa %xmm14, %xmm3
-; SSE-NEXT: pand %xmm15, %xmm3
-; SSE-NEXT: por %xmm0, %xmm3
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[0,1,1,0,4,5,6,7]
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm2, %xmm8
+; SSE-NEXT: pand %xmm14, %xmm8
+; SSE-NEXT: por %xmm0, %xmm8
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[0,1,1,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
+; SSE-NEXT: movdqa %xmm6, %xmm1
+; SSE-NEXT: movdqa %xmm6, %xmm3
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm15[4],xmm1[5],xmm15[5],xmm1[6],xmm15[6],xmm1[7],xmm15[7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT: movdqa %xmm5, %xmm13
-; SSE-NEXT: andps %xmm10, %xmm0
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
+; SSE-NEXT: movdqa %xmm9, %xmm15
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
-; SSE-NEXT: pandn %xmm1, %xmm10
-; SSE-NEXT: por %xmm0, %xmm10
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm15, %xmm11
+; SSE-NEXT: movdqa %xmm7, %xmm2
+; SSE-NEXT: pandn %xmm1, %xmm2
+; SSE-NEXT: andps %xmm7, %xmm0
+; SSE-NEXT: por %xmm0, %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm5, %xmm0
+; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm1
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,1,1,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm11[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm13, %xmm8
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT: movdqa %xmm4, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT: andps %xmm7, %xmm0
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
+; SSE-NEXT: pandn %xmm1, %xmm7
+; SSE-NEXT: por %xmm0, %xmm7
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm6, %xmm11
; SSE-NEXT: psrld $16, %xmm11
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm1[4],xmm11[5],xmm1[5],xmm11[6],xmm1[6],xmm11[7],xmm1[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm1[4],xmm15[5],xmm1[5],xmm15[6],xmm1[6],xmm15[7],xmm1[7]
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm1[4],xmm6[5],xmm1[5],xmm6[6],xmm1[6],xmm6[7],xmm1[7]
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm8, %xmm1
-; SSE-NEXT: psrlq $16, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm15
-; SSE-NEXT: psrld $16, %xmm15
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm1[4],xmm15[5],xmm1[5],xmm15[6],xmm1[6],xmm15[7],xmm1[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm1[4],xmm10[5],xmm1[5],xmm10[6],xmm1[6],xmm10[7],xmm1[7]
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: psrlq $16, %xmm1
-; SSE-NEXT: movdqa (%rsp), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm2, %xmm5
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm15[4],xmm1[5],xmm15[5],xmm1[6],xmm15[6],xmm1[7],xmm15[7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm9 = xmm1[0,1,2,3,5,6,4,7]
-; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm10 = xmm1[0,1,2,3,5,6,4,7]
+; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
+; SSE-NEXT: movdqa (%rsp), %xmm6 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm6, %xmm15
+; SSE-NEXT: psrld $16, %xmm15
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm3[4],xmm15[5],xmm3[5],xmm15[6],xmm3[6],xmm15[7],xmm3[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm3[4],xmm6[5],xmm3[5],xmm6[6],xmm3[6],xmm6[7],xmm3[7]
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: movdqa %xmm5, %xmm13
+; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm5, %xmm1
+; SSE-NEXT: psrlq $16, %xmm1
+; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT: movdqa %xmm4, %xmm3
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm1[0,1,2,3,5,6,4,7]
+; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
+; SSE-NEXT: movdqa %xmm13, %xmm0
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm8, %xmm0
+; SSE-NEXT: movdqa %xmm8, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm2
+; SSE-NEXT: por %xmm0, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE-NEXT: movdqa %xmm3, %xmm13
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[0,2]
+; SSE-NEXT: movdqa %xmm7, %xmm3
+; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: pandn %xmm8, %xmm14
+; SSE-NEXT: por %xmm0, %xmm14
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm2, %xmm0
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm1, %xmm0
-; SSE-NEXT: pandn %xmm12, %xmm0
-; SSE-NEXT: movdqa %xmm7, %xmm14
-; SSE-NEXT: movdqa %xmm7, %xmm8
-; SSE-NEXT: pand %xmm1, %xmm14
-; SSE-NEXT: por %xmm0, %xmm14
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,1],xmm0[0,2]
-; SSE-NEXT: pand %xmm1, %xmm2
-; SSE-NEXT: movdqa %xmm5, %xmm14
-; SSE-NEXT: pandn %xmm5, %xmm1
-; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: movdqa %xmm4, %xmm0
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm13[4],xmm0[5],xmm13[5],xmm0[6],xmm13[6],xmm0[7],xmm13[7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[0,2]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
@@ -1814,53 +1814,53 @@
; SSE-NEXT: # xmm2 = mem[2,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm8[1,1,1,1,4,5,6,7]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[1,1,1,1,4,5,6,7]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[0,1,0,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm12[2,2,2,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[2,2,2,2]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm0[2],xmm8[3],xmm0[3]
; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm2[0],xmm8[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
-; SSE-NEXT: psrldq {{.*#+}} xmm4 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm10[2,1,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; SSE-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[2,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[1,1,1,1,4,5,6,7]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[0,1,0,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm14[2,2,2,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm4[2,2,2,2]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm0[2],xmm13[3],xmm0[3]
; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm2[0],xmm13[1]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: movaps %xmm0, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rcx)
+; SSE-NEXT: movaps %xmm0, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%r8)
+; SSE-NEXT: movaps %xmm0, (%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%r8)
-; SSE-NEXT: movapd %xmm9, (%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%r9)
+; SSE-NEXT: movaps %xmm0, (%r8)
+; SSE-NEXT: movapd %xmm5, 16(%r9)
+; SSE-NEXT: movapd %xmm10, (%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movaps %xmm15, (%rax)
-; SSE-NEXT: movaps %xmm11, 16(%rax)
+; SSE-NEXT: movaps %xmm15, 16(%rax)
+; SSE-NEXT: movaps %xmm11, (%rax)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movapd %xmm13, (%rax)
-; SSE-NEXT: movapd %xmm8, 16(%rax)
+; SSE-NEXT: movapd %xmm13, 16(%rax)
+; SSE-NEXT: movapd %xmm8, (%rax)
; SSE-NEXT: addq $216, %rsp
; SSE-NEXT: retq
;
@@ -4026,777 +4026,792 @@
; AVX1-ONLY-LABEL: load_i16_stride7_vf32:
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: subq $680, %rsp # imm = 0x2A8
-; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm14, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa 400(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm1, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 384(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm11[2,2,3,3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm6, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa 432(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 416(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm12[0],xmm2[0],xmm12[1],xmm2[1],xmm12[2],xmm2[2],xmm12[3],xmm2[3]
+; AVX1-ONLY-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm13
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa 144(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm15
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5],xmm1[6,7]
-; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm0 = zero,xmm15[2],xmm0[2],zero
+; AVX1-ONLY-NEXT: vmovdqa 240(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm2[7]
-; AVX1-ONLY-NEXT: vmovdqa 112(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm3
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm4[2],xmm10[2],zero
-; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %xmm4, %xmm6
-; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0,1,2],xmm3[3,4],xmm0[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm0 = [65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
+; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm0[3,4],xmm2[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa 304(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm3[7]
+; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm5 = [65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm5, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 400(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa 384(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm0, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm0, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 144(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm9, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 432(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 416(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 240(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm7
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm5[2],xmm8[2],zero
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 304(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm11[0],xmm4[0],xmm11[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,0,0,0]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm4[7]
-; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm2, %ymm0
+; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[0,0,0,0]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6],xmm3[7]
+; AVX1-ONLY-NEXT: vmovdqa 112(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
+; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm4 = zero,xmm4[2],xmm0[2],zero
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3,4],xmm3[5,6,7]
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm5, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm13[4],xmm14[4],xmm13[5],xmm14[5],xmm13[6],xmm14[6],xmm13[7],xmm14[7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm11[4],xmm10[4],xmm11[5],xmm10[5],xmm11[6],xmm10[6],xmm11[7],xmm10[7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; AVX1-ONLY-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3,4,5],xmm15[6],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm6[0,1,2,3,4,5],xmm10[6],xmm6[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,0,3,2,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpslld $16, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm13[0],xmm1[0],xmm13[1],xmm1[1],xmm13[2],xmm1[2],xmm13[3],xmm1[3]
+; AVX1-ONLY-NEXT: vpslld $16, %xmm13, %xmm1
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm12[0],xmm1[0],xmm12[1],xmm1[1],xmm12[2],xmm1[2],xmm12[3],xmm1[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm14, %xmm1
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm7[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm7[0,1,2,3,4,5]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6],xmm2[7]
+; AVX1-ONLY-NEXT: vmovaps %xmm15, %xmm12
+; AVX1-ONLY-NEXT: vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm15[4],xmm13[4],xmm15[5],xmm13[5],xmm15[6],xmm13[6],xmm15[7],xmm13[7]
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm2 = [8,9,8,9,8,9,8,9,6,7,6,7,6,7,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm5
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm15[0],xmm6[1],xmm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4],xmm2[5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrld $16, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm6[4],xmm10[4],xmm6[5],xmm10[5],xmm6[6],xmm10[6],xmm6[7],xmm10[7]
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm3 = [8,9,8,9,8,9,8,9,6,7,6,7,6,7,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm6[1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,0,3,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm1[3,4],xmm3[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm3[0,1,2,3,4,5]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6],xmm3[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm3
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm2 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm2, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm8 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm8, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm9[4],xmm2[4],xmm9[5],xmm2[5],xmm9[6],xmm2[6],xmm9[7],xmm2[7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = xmm14[4],mem[4],xmm14[5],mem[5],xmm14[6],mem[6],xmm14[7],mem[7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $64, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm10[0,1,2,3,4,5],mem[6],xmm10[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,0,3,2,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpslld $16, %xmm1, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm8[0],mem[1],xmm8[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,0,3,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm12, %xmm4
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm5 = xmm11[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm11[0,1,2,3,4,5]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6],xmm5[7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm12, %xmm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm1, %ymm4
-; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm14[0,1,2,3,4,5],xmm9[6],xmm14[7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,0,3,2,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpslld $16, %xmm9, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm7, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpslldq {{.*#+}} xmm4 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm7[0,1,2,3,4,5]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6],xmm4[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm2, %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm4, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0],xmm3[1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[1,0,3,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4],xmm5[5,6,7]
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm5 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm5, %ymm1
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm4, %ymm1
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm8, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsllq $16, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm15[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1],xmm3[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm13[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm11[4],xmm0[4],xmm11[5],xmm0[5],xmm11[6],xmm0[6],xmm11[7],xmm0[7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm10[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm1 = xmm8[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1],xmm1[2],xmm11[2],xmm1[3],xmm11[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm12[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm13[2,2,3,3]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm4 = xmm1[1],xmm4[1]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm6[0],xmm15[0],xmm6[1],xmm15[1],xmm6[2],xmm15[2],xmm6[3],xmm15[3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm1[2,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm12[2,2,2,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm15[0,1,2,3,4,5],xmm5[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm10[0,1,0,1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6],xmm13[7]
+; AVX1-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm6 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm6, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm5 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm5, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm4, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsllq $16, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; AVX1-ONLY-NEXT: vpshufd $236, (%rsp), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm14[0,1],xmm4[2,3],xmm14[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm4[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,5],xmm4[6,7]
+; AVX1-ONLY-NEXT: vpshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[2,2,2,2]
; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2,3,4,5],xmm0[6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm7[0,1,0,1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm4[7]
-; AVX1-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm7[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm7[0,1,0,1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm5[7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm2[1,1,1,1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm9[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[2,2,3,3]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm5 = xmm0[1],xmm5[1]
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm6[0],mem[0],xmm6[1],mem[1],xmm6[2],mem[2],xmm6[3],mem[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm0[2,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[0,3,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm15[0,1],xmm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm1, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vorps %ymm4, %ymm5, %ymm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm1, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm4, %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpsllq $16, %xmm2, %xmm3
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm9[4],xmm3[4],xmm9[5],xmm3[5],xmm9[6],xmm3[6],xmm9[7],xmm3[7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm10[0,1],xmm4[2,3],xmm10[4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2,3],xmm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm6[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm3[0,1,2,3,4,5],xmm4[6,7]
-; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[2,2,3,3]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm5 = xmm3[1],xmm5[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm3[2,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm15 = xmm15[0,3,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm15[0,1],xmm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[2,2,2,2]
-; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,2,3,4,5],xmm15[6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm14 = xmm11[0,1,0,1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm14 = xmm15[0,1,2,3,4,5,6],xmm14[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm12[1,1,1,1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm14, %ymm14
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm8 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm8, %ymm5
-; AVX1-ONLY-NEXT: vandnps %ymm14, %ymm8, %ymm14
-; AVX1-ONLY-NEXT: vorps %ymm5, %ymm14, %ymm5
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm6[2,2,3,3]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm13 = xmm0[1],xmm13[1]
+; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = xmm3[0],mem[0],xmm3[1],mem[1],xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm14 = xmm0[2,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm14 = xmm14[0,3,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm14[0,1],xmm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm2 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm2, %ymm5
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm13, %ymm13
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm13, %ymm5
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm10 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm10, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm10, %ymm5
-; AVX1-ONLY-NEXT: vorps %ymm4, %ymm5, %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm12[0],mem[0],xmm12[1],mem[1],xmm12[2],mem[2],xmm12[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm7 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm7, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm7, %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm5, %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm3[0],mem[0],xmm3[1],mem[1],xmm3[2],mem[2],xmm3[3],mem[3]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,0,3,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $64, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = xmm5[0,1,2,3,4,5],mem[6],xmm5[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $64, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm2[0,1,2,3,4,5],mem[6],xmm2[7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,0,0,0,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0],xmm4[1,2],xmm5[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm13[4],xmm4[5],xmm13[5],xmm4[6],xmm13[6],xmm4[7],xmm13[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm14 = xmm4[2,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm14 = xmm14[0,0,0,0]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm14 = xmm5[0,1,2,3,4,5],xmm14[6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, %xmm9
-; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[0],xmm7[1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm8[4],xmm11[4],xmm8[5],xmm11[5],xmm8[6],xmm11[6],xmm8[7],xmm11[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm4[2,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,0,0,0]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm5[0,1,2,3,4,5],xmm13[6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm2[0],mem[1],xmm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,1,0,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,7,7,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm15[4],xmm12[4],xmm15[5],xmm12[5],xmm15[6],xmm12[6],xmm15[7],xmm12[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,7,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm15 = xmm5[4],xmm11[4],xmm5[5],xmm11[5],xmm5[6],xmm11[6],xmm5[7],xmm11[7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm14 = xmm5[4],xmm10[4],xmm5[5],xmm10[5],xmm5[6],xmm10[6],xmm5[7],xmm10[7]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm5 = [4,5,2,3,4,5,6,7,8,9,4,5,8,9,2,3]
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm15, %xmm15
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm8, %xmm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm15, %ymm13
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm14, %xmm14
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm10, %xmm15
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm14, %ymm14
; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm15 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm15, %ymm0
-; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm15, %ymm13
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm13, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm13
-; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm10, %ymm13
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm10, %ymm0
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm13, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm15, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm14, %ymm15, %ymm14
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm14, %ymm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm7, %ymm13
+; AVX1-ONLY-NEXT: vandps %ymm7, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm13, %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[0,0,3,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm7[0],xmm2[0],xmm7[1],xmm2[1],xmm7[2],xmm2[2],xmm7[3],xmm2[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,0,3,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $191, (%rsp), %xmm2, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,1,2,3,4,5],xmm2[6],mem[7]
+; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[0,1,2,3,4,5],xmm7[6],mem[7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[0,0,0,0,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,7,6,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0],xmm0[1,2],xmm13[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm6[4],xmm1[4],xmm6[5],xmm1[5],xmm6[6],xmm1[6],xmm6[7],xmm1[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm14 = xmm0[2,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0],xmm1[1,2],xmm13[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm14 = xmm1[2,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm14 = xmm14[0,0,0,0]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5],xmm14[6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm8[0],mem[1],xmm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm14 = xmm6[0],xmm9[1],xmm6[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm14 = xmm14[0,1,1,0,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,7,7,7,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm10[4],mem[4],xmm10[5],mem[5],xmm10[6],mem[6],xmm10[7],mem[7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm14 = xmm14[0,1,2,1]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,4,7,7]
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm14[4],mem[4],xmm14[5],mem[5],xmm14[6],mem[6],xmm14[7],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm14 = xmm14[4],xmm11[4],xmm14[5],xmm11[5],xmm14[6],xmm11[6],xmm14[7],xmm11[7]
; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm14, %xmm5
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm14
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm8, %xmm14
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm15, %ymm3
-; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm15, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm3, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm15 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm15, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm15, %ymm1
-; AVX1-ONLY-NEXT: vmovaps %ymm15, %ymm5
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0],xmm1[1,2],xmm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm4[0,1,2,1]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm3[6,7]
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm15, %ymm0
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm15, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm6 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm6, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm14[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1,2],xmm2[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm4[0,1,2,1]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm2[6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm2, %xmm3
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm2, %xmm2
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm6[2,3,2,3]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm9, %xmm4
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm3, %xmm4
; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1],xmm4[2],mem[2],xmm4[3],mem[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm14[0,1,0,3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm11[4],xmm4[5],xmm11[5],xmm4[6],xmm11[6],xmm4[7],xmm11[7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[12,13,14,15,4,5,6,7,0,1,4,5,8,9,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm15[2,3,2,3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm10[2,3,2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm5, %ymm1
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm3, %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm9[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0],xmm1[1,2],xmm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5],xmm0[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm11, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm7[2,3,2,3]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm5, %xmm3
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm10[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm8[2],xmm3[3],xmm8[3]
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[12,13,14,15,4,5,6,7,0,1,4,5,8,9,6,7]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm3 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm3, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm5, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vmovaps %ymm5, %ymm13
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm2, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm12[2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm7[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1,2],xmm2[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm7, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm9, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm9[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm10[2],xmm2[3],xmm10[3]
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm11[4],xmm2[5],xmm11[5],xmm2[6],xmm11[6],xmm2[7],xmm11[7]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[12,13,14,15,4,5,6,7,0,1,4,5,8,9,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm13, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm13, %ymm1
+; AVX1-ONLY-NEXT: vmovaps %ymm13, %ymm8
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm12, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm0
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm10[0],mem[0],xmm10[1],mem[1],xmm10[2],mem[2],xmm10[3],mem[3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm15[0],xmm14[0],xmm15[1],xmm14[1],xmm15[2],xmm14[2],xmm15[3],xmm14[3]
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[8,9,8,9,8,9,8,9,6,7,6,7,6,7,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3,4],xmm0[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $64, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm12[0,1,2,3,4,5],mem[6],xmm12[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $64, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm14[0,1,2,3,4,5],mem[6],xmm14[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm1[5,6,7]
; AVX1-ONLY-NEXT: vpsrld $16, %xmm6, %xmm1
; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1],xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm14[1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,4,7,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm14[4],xmm3[5],xmm14[5],xmm3[6],xmm14[6],xmm3[7],xmm14[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm15[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm15[0],xmm13[0],xmm15[1],xmm13[1],xmm15[2],xmm13[2],xmm15[3],xmm13[3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,4,7,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,2,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm3 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm3, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm8, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm8, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = xmm12[0],mem[0],xmm12[1],mem[1],xmm12[2],mem[2],xmm12[3],mem[3]
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[8,9,8,9,8,9,8,9,6,7,6,7,6,7,6,7]
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm13, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm3 = xmm15[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm12[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3,4],xmm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3,4,5],xmm9[6],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm8[0,1,2,3,4,5],xmm11[6],xmm8[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm1[5,6,7]
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm7, %xmm1
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm11[4],xmm1[5],xmm11[5],xmm1[6],xmm11[6],xmm1[7],xmm11[7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm8[0],xmm11[1],xmm8[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,4,7,7]
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm5, %xmm1
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm10[0],xmm9[1],xmm10[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,4,7,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm7[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm3 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm3, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vinsertps $41, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = zero,xmm10[1],mem[0],zero
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm7[4],xmm2[5],xmm7[5],xmm2[6],xmm7[6],xmm2[7],xmm7[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,2,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3,4],xmm3[5,6,7]
-; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm12[1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm3[5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm2 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm2, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vinsertps $41, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1],mem[0],zero
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3,4],xmm2[5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0],xmm14[1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm2[5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm13[1,1,1,1]
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm15[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3],xmm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3],xmm4[4,5,6,7]
; AVX1-ONLY-NEXT: vpmovzxwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm14[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm5[6,7]
; AVX1-ONLY-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = mem[3,3,3,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm4 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm4, %ymm1
-; AVX1-ONLY-NEXT: vandps %ymm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vmovaps %ymm4, %ymm10
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm3, %ymm1
-; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps $41, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = zero,xmm3[1],mem[0],zero
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm4 = xmm15[2],xmm13[2],xmm15[3],xmm13[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
-; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm9[0],mem[1],xmm9[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vmovaps %ymm4, %ymm13
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vinsertps $41, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = zero,xmm2[1],mem[0],zero
+; AVX1-ONLY-NEXT: vpunpckhdq (%rsp), %xmm12, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm12[2],mem[2],xmm12[3],mem[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3,4],xmm4[5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm11[0],xmm8[1],xmm11[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm4[5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[1,1,1,1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm5 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm4[5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm5 = xmm6[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3],xmm5[4,5,6,7]
-; AVX1-ONLY-NEXT: vpmovzxwd {{.*#+}} xmm5 = xmm8[0],zero,xmm8[1],zero,xmm8[2],zero,xmm8[3],zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
+; AVX1-ONLY-NEXT: vpmovzxwd {{.*#+}} xmm5 = xmm10[0],zero,xmm10[1],zero,xmm10[2],zero,xmm10[3],zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm7[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5],xmm6[6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm7[3,3,3,3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm3[3,3,3,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm5
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm10, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm4, %ymm10, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm13, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm13, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
; AVX1-ONLY-NEXT: addq $680, %rsp # imm = 0x2A8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-SLOW-LABEL: load_i16_stride7_vf32:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $520, %rsp # imm = 0x208
-; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm13
-; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm8
-; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm10
-; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm11
-; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm3
-; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm9
-; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm9[1],ymm1[2,3,4],ymm9[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm1, %ymm7
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,2,2,3]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm0 = [0,1,14,15,12,13,14,15,2,3,6,7,12,13,2,3,16,17,30,31,28,29,30,31,18,19,22,23,28,29,18,19]
-; AVX2-SLOW-NEXT: vpshufb %ymm0, %ymm1, %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1],ymm4[2],ymm3[3,4,5],ymm4[6],ymm3[7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm5
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm6
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm3
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4],xmm1[5],xmm3[6],xmm1[7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,14,15,12,13,10,11,8,9,8,9,8,9,8,9]
-; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm1, %xmm4
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm4, %ymm2, %ymm14
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1],ymm10[2],ymm11[3,4,5],ymm10[6],ymm11[7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm4
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm4[4],xmm2[5],xmm4[6],xmm2[7]
-; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm2, %xmm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0],ymm13[1],ymm8[2,3,4],ymm13[5],ymm8[6,7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-SLOW-NEXT: vpshufb %ymm0, %ymm3, %ymm0
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1],ymm13[2],ymm8[3,4],ymm13[5],ymm8[6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm8, %ymm12
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3],ymm0[4,5,6,7,8,9,10],ymm2[11],ymm0[12,13,14,15]
+; AVX2-SLOW-NEXT: subq $552, %rsp # imm = 0x228
+; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm5
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm10
+; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm3
+; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm1[2],ymm2[3,4,5],ymm1[6],ymm2[7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm11
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, %ymm9
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5],xmm1[6],xmm0[7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,14,15,12,13,10,11,8,9,8,9,8,9,8,9]
+; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm0, %xmm2
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0],ymm10[1],ymm3[2,3,4],ymm10[5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm13
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,14,15,12,13,14,15,2,3,6,7,12,13,2,3,16,17,30,31,28,29,30,31,18,19,22,23,28,29,18,19]
+; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm0, %ymm4
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm2, %ymm4, %ymm2
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0],ymm7[1],ymm8[2,3,4],ymm7[5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
+; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm2, %ymm2
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm6[2],ymm5[3,4,5],ymm6[6],ymm5[7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
+; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1],ymm10[2],ymm13[3,4],ymm10[5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7,8,9,10],ymm2[11],ymm1[12,13,14,15]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = [2,3,2,3,2,3,2,3,8,9,8,9,6,7,4,5,18,19,18,19,18,19,18,19,24,25,24,25,22,23,20,21]
-; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1,2],ymm10[3],ymm11[4,5],ymm10[6],ymm11[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1,2],ymm9[3],ymm11[4,5],ymm9[6],ymm11[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3,4,5],xmm3[6],xmm4[7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = [2,3,0,1,14,15,12,13,10,11,10,11,10,11,10,11]
; AVX2-SLOW-NEXT: vpshufb %xmm4, %xmm3, %xmm3
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm3, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1],ymm9[2],ymm7[3,4],ymm9[5],ymm7[6,7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3],ymm0[4,5,6,7,8,9,10],ymm3[11],ymm0[12,13,14,15]
-; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2],ymm5[3],ymm6[4,5],ymm5[6],ymm6[7]
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3],ymm1[4,5,6,7,8,9,10],ymm3[11],ymm1[12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2],ymm6[3],ymm5[4,5],ymm6[6],ymm5[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2,3,4,5],xmm2[6],xmm3[7]
; AVX2-SLOW-NEXT: vpshufb %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm9[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm8
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0],ymm11[1],ymm9[2,3],ymm11[4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm11, %ymm14
+; AVX2-SLOW-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm15
; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4],ymm2[5,6,7,8,9,10,11],ymm0[12],ymm2[13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0],ymm6[1],ymm5[2,3],ymm6[4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3,4,5],xmm2[6],xmm1[7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm10, %ymm11
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1],ymm10[2,3],ymm13[4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm10[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7,8,9,10,11],ymm3[12],ymm2[13,14,15]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = [4,5,2,3,0,1,14,15,12,13,12,13,12,13,12,13]
+; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm4 = [4,5,4,5,4,5,4,5,8,9,10,11,8,9,6,7,20,21,20,21,20,21,20,21,24,25,26,27,24,25,22,23]
+; AVX2-SLOW-NEXT: vpshufb %ymm4, %ymm2, %ymm2
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm10
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm7[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm8, %ymm12
+; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm7, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4],ymm2[5,6,7,8,9,10,11],ymm1[12],ymm2[13,14,15]
+; AVX2-SLOW-NEXT: vpshufb %ymm4, %ymm1, %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3,4,5],xmm3[6],xmm2[7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = [4,5,4,5,4,5,4,5,8,9,10,11,8,9,6,7,20,21,20,21,20,21,20,21,24,25,26,27,24,25,22,23]
-; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = [4,5,2,3,0,1,14,15,12,13,12,13,12,13,12,13]
-; AVX2-SLOW-NEXT: vpshufb %xmm4, %xmm2, %xmm2
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm11[1],ymm10[2,3],ymm11[4],ymm10[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm10, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3,4,5],xmm2[6],xmm0[7]
-; AVX2-SLOW-NEXT: vpshufb %xmm4, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1],ymm13[2,3],ymm12[4,5],ymm13[6,7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm13[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4],ymm2[5,6,7,8,9,10,11],ymm4[12],ymm2[13,14,15]
-; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm0, %ymm2, %ymm9
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0],ymm6[1],ymm5[2,3,4],ymm6[5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2],xmm0[3],xmm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1,2],ymm8[3],ymm7[4,5],ymm8[6],ymm7[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0],ymm5[1],ymm6[2,3],ymm5[4],ymm6[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm4[1],xmm2[2,3,4,5],xmm4[6],xmm2[7]
+; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm2, %xmm2
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm9
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0],ymm14[1],ymm15[2,3,4],ymm14[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,2,1,0,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2],ymm11[3],ymm13[4,5],ymm11[6],ymm13[7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[1,3,2,3]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = [6,7,6,7,6,7,6,7,8,9,4,5,10,11,0,1,22,23,22,23,22,23,22,23,24,25,20,21,26,27,16,17]
; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm0, %ymm2, %ymm8
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm11[1],ymm10[2,3,4],ymm11[5],ymm10[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2],xmm0[3],xmm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1,2],ymm13[3],ymm12[4,5],ymm13[6],ymm12[7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm13, %ymm6
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[1,3,2,3]
-; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[3,2,1,0,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm0, %ymm2, %ymm7
-; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm15
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2],ymm5[3],ymm15[4,5],ymm5[6],ymm15[7]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,6,4,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm4[0,1,0,2]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm2[0,1,2,1,4,5,6,5]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3,4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm14
-; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm13
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm14[3],ymm13[4,5],ymm14[6],ymm13[7]
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm8
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1,2],ymm7[3],ymm12[4,5],ymm7[6],ymm12[7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,3,2,3]
+; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0],ymm5[1],ymm6[2,3,4],ymm5[5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3],xmm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm7
+; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm14
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm1[3],ymm14[4,5],ymm1[6],ymm14[7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, %ymm6
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,6,4,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %ymm3
+; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm4[0,1,0,2]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm15 = ymm2[0,1,2,1,4,5,6,5]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm15 = ymm15[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm0
+; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1,2],ymm0[3],ymm15[4,5],ymm0[6],ymm15[7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,6,4,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm1
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %ymm3
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm3[0,1,0,2]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm10 = ymm0[0,1,2,1,4,5,6,5]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm10[7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm12 = ymm0[0,1,2,1,4,5,6,5]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm12[7]
; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1],ymm13[2],ymm14[3,4,5],ymm13[6],ymm14[7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm10
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm10[4],xmm1[5],xmm10[6],xmm1[7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm10 = [0,1,0,1,14,15,12,13,10,11,8,9,12,13,14,15]
-; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm14[2],ymm6[3,4,5],ymm14[6],ymm6[7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm12
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm12[4],xmm1[5],xmm12[6],xmm1[7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = [0,1,0,1,14,15,12,13,10,11,8,9,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[0,1,1,3,4,5,5,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,5,6,8,9,10,11,12,13,13,14]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3,4,5,6,7],ymm2[8],ymm1[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm15[2],ymm5[3,4,5],ymm15[6],ymm5[7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4],xmm1[5],xmm2[6],xmm1[7]
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,5,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,5,6,8,9,10,11,12,13,13,14]
@@ -4805,19 +4820,7 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1],ymm15[2],ymm5[3,4,5],ymm15[6],ymm5[7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5],xmm1[6],xmm0[7]
-; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm2[0,1,1,3,4,5,5,7]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,5,6,8,9,10,11,12,13,13,14]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm15[3],ymm5[4,5],ymm15[6],ymm5[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2],ymm14[3],ymm6[4,5],ymm14[6],ymm6[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3,4,5],xmm0[6],xmm1[7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,0,1,14,15,12,13,10,11,12,13,14,15]
@@ -4827,11 +4830,10 @@
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[0,1,0,3,4,5,4,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm10[0],ymm0[1,2,3,4,5,6,7],ymm10[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm13[3],ymm14[4,5],ymm13[6],ymm14[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm15[3],ymm5[4,5],ymm15[6],ymm5[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm0, %xmm0
@@ -4843,8 +4845,8 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm9[0],ymm0[1,2,3,4,5,6,7],ymm9[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0],ymm5[1],ymm15[2,3],ymm5[4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0],ymm6[1],ymm14[2,3],ymm6[4],ymm14[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,0,4,5,6,7]
@@ -4857,7 +4859,8 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm0[1,2,3,4,5,6,7],ymm8[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0],ymm14[1],ymm13[2,3],ymm14[4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0],ymm5[1],ymm15[2,3],ymm5[4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm12
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3,4,5],xmm2[6],xmm0[7]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,0,4,5,6,7]
@@ -4870,52 +4873,54 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm7[0],ymm0[1,2,3,4,5,6,7],ymm7[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm8
-; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1],ymm12[2],ymm6[3,4,5],ymm12[6],ymm6[7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm13, %ymm10
+; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %ymm11, %ymm8
+; AVX2-SLOW-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1],ymm13[2],ymm11[3,4,5],ymm13[6],ymm11[7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27>
; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm0, %ymm4
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6],ymm4[7,8,9,10,11,12,13],ymm0[14],ymm4[15]
-; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1],ymm11[2],ymm9[3,4],ymm11[5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1],ymm13[2],ymm9[3,4],ymm13[5],ymm9[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm7
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0],ymm6[1],ymm14[2,3,4],ymm6[5],ymm14[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm7
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm7[0],xmm4[1],xmm7[2],xmm4[3],xmm7[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm7 = [0,1,6,7,4,5,2,3,0,1,14,15,12,13,14,15]
; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm4, %xmm4
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[0,1,2,1,4,5,6,5]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm0[0],ymm2[1,2,3,4,5,6,7],ymm0[8],ymm2[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0],ymm5[1],ymm15[2,3,4],ymm5[5],ymm15[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2],xmm0[3],xmm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,1,4,5,6,5]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3,4,5,6,7],ymm0[8],ymm1[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0],ymm5[1],ymm15[2,3,4],ymm5[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2],xmm0[3],xmm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm2[0,1,2,1,4,5,6,5]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $187, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm10[2],mem[3,4,5],ymm10[6],mem[7]
+; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm4[2],ymm5[3,4,5],ymm4[6],ymm5[7]
; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm1, %ymm2
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6],ymm2[7,8,9,10,11,12,13],ymm1[14],ymm2[15]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm5[2],ymm6[3,4],ymm5[5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1],ymm6[2],ymm11[3,4],ymm6[5],ymm11[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
@@ -4925,10 +4930,10 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm12[3],ymm8[4,5],ymm12[6],ymm8[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm10[3],ymm8[4,5],ymm10[6],ymm8[7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7,8],ymm0[9,10,11,12,13,14],ymm1[15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm11[2,3],ymm9[4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm13[2,3],ymm9[4,5],ymm13[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,2,2,3,4,5,6,7]
@@ -4938,11 +4943,12 @@
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29>
; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1],ymm14[2],ymm13[3,4],ymm14[5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1],ymm9[2],ymm14[3,4],ymm9[5],ymm14[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2],xmm2[3],xmm1[4],xmm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = [8,9,8,9,6,7,4,5,2,3,2,3,2,3,2,3]
-; AVX2-SLOW-NEXT: vpshufb %xmm4, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm10 = [8,9,8,9,6,7,4,5,2,3,2,3,2,3,2,3]
+; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm7
; AVX2-SLOW-NEXT: vmovdqa 432(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %xmm1
@@ -4954,25 +4960,23 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm0[0],ymm7[1,2,3,4,5,6,7],ymm0[8],ymm7[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm9[2],ymm15[3,4],ymm9[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm12[2],ymm15[3,4],ymm12[5],ymm15[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm7
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm7[0,1],xmm0[2],xmm7[3],xmm0[4],xmm7[5,6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm4, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm4
+; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm8
; AVX2-SLOW-NEXT: vmovdqa 208(%rdi), %xmm12
-; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm7 = xmm0[0,1,2,3,4,5],xmm12[6],xmm0[7]
+; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %xmm10
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm7 = xmm10[0,1,2,3,4,5],xmm12[6],xmm10[7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,1,0,3]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,7,6]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm7[5,6,7],ymm4[8,9,10,11,12],ymm7[13,14,15]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1,2],ymm10[3],ymm11[4,5],ymm10[6],ymm11[7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm7[5,6,7],ymm8[8,9,10,11,12],ymm7[13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm5[0,1,2],ymm4[3],ymm5[4,5],ymm4[6],ymm5[7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm7[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm7 = ymm8[0],ymm7[1,2,3,4,5,6],ymm8[7,8],ymm7[9,10,11,12,13,14],ymm8[15]
; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm7, %ymm3
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm7 = ymm11[0,1],ymm6[2,3],ymm11[4,5],ymm6[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm7, %xmm8
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[2,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[0,2,2,3,4,5,6,7]
@@ -4980,16 +4984,15 @@
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm3[0],ymm4[1,2,3,4,5,6,7],ymm3[8],ymm4[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm3[0],ymm0[1,2,3,4,5,6,7],ymm3[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3],mem[4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5,6,7,8],ymm4[9],ymm3[10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5],mem[6],ymm4[7]
+; AVX2-SLOW-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,2],ymm13[3],mem[4,5],ymm13[6],mem[7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm7 = xmm4[0,1,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm4
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
@@ -4998,7 +5001,7 @@
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31>
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm3, %ymm3
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0,1],ymm14[2,3],ymm13[4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1],ymm9[2,3],ymm14[4,5],ymm9[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3],xmm5[4],xmm4[5],xmm5[6,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2,3,4,5,6,7]
@@ -5011,22 +5014,24 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm4[0,1,2,3,4],ymm1[5,6,7],ymm4[8,9,10,11,12],ymm1[13,14,15]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0],ymm1[1,2,3,4,5,6,7],ymm3[8],ymm1[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm9[2,3],ymm15[4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3],xmm4[4],xmm3[5],xmm4[6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm2
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm12[0],xmm0[1],xmm12[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm12[0],xmm10[1],xmm12[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7],ymm2[8,9,10,11,12],ymm0[13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm11[1],ymm10[2,3],ymm11[4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0],ymm2[1],mem[2,3],ymm2[4],mem[5,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4,5,6,7,8],ymm3[9],ymm2[10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm2, %ymm2
-; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm6[0,1,2],mem[3],ymm6[4,5],mem[6],ymm6[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1,2],ymm6[3],ymm11[4,5],ymm6[6],ymm11[7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm3[0,1,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm3
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
@@ -5036,174 +5041,176 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rsi)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rdx)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rsi)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rdx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%r8)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm2, (%r8)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%r9)
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%r8)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm2, (%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX2-SLOW-NEXT: vmovdqa %ymm8, (%rax)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm2, 32(%rax)
-; AVX2-SLOW-NEXT: vmovdqa %ymm8, (%rax)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovdqa %ymm1, 32(%rax)
; AVX2-SLOW-NEXT: vmovdqa %ymm0, (%rax)
-; AVX2-SLOW-NEXT: addq $520, %rsp # imm = 0x208
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, 32(%rax)
+; AVX2-SLOW-NEXT: addq $552, %rsp # imm = 0x228
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: load_i16_stride7_vf32:
; AVX2-FAST: # %bb.0:
-; AVX2-FAST-NEXT: subq $584, %rsp # imm = 0x248
-; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm5
-; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm6
-; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm7
-; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm8
-; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm4
-; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm10
+; AVX2-FAST-NEXT: subq $680, %rsp # imm = 0x2A8
+; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm5
+; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm8
+; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm2
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm1[2],ymm2[3,4,5],ymm1[6],ymm2[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm11
; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm12
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [0,1,14,15,12,13,14,15,2,3,6,7,12,13,2,3,16,17,30,31,28,29,30,31,18,19,22,23,28,29,18,19]
-; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm4[2],ymm3[3,4,5],ymm4[6],ymm3[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5],xmm1[6],xmm0[7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,14,15,12,13,10,11,8,9,8,9,8,9,8,9]
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm2
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0],ymm3[1],ymm4[2,3,4],ymm3[5],ymm4[6,7]
; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm9
-; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm11
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm3
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4],xmm0[5],xmm3[6],xmm0[7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,14,15,12,13,10,11,8,9,8,9,8,9,8,9]
-; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm0, %xmm4
+; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm10
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,14,15,12,13,14,15,2,3,6,7,12,13,2,3,16,17,30,31,28,29,30,31,18,19,22,23,28,29,18,19]
+; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm0, %ymm4
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm4, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1],ymm7[2],ymm8[3,4,5],ymm7[6],ymm8[7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm4
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm4[4],xmm2[5],xmm4[6],xmm2[7]
-; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm2, %xmm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0],ymm5[1],ymm6[2,3,4],ymm5[5],ymm6[6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm13
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm1
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm12[2],ymm10[3,4],ymm12[5],ymm10[6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <2,5,1,u,4,u,u,u>
-; AVX2-FAST-NEXT: vpermd %ymm1, %ymm2, %ymm1
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [2,3,2,3,2,3,2,3,8,9,0,1,6,7,8,9,18,19,18,19,18,19,18,19,24,25,16,17,22,23,24,25]
-; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm11[0,1,2],ymm9[3],ymm11[4,5],ymm9[6],ymm11[7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = [2,3,0,1,14,15,12,13,10,11,10,11,10,11,10,11]
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm4, %xmm4
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm4, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2],ymm7[3],ymm8[4,5],ymm7[6],ymm8[7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm4
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm4[0],xmm1[1],xmm4[2,3,4,5],xmm1[6],xmm4[7]
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm1, %xmm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1],ymm13[2],ymm6[3,4],ymm13[5],ymm6[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm4, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm2, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0],ymm7[1],ymm8[2,3,4],ymm7[5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm6[2],ymm5[3,4,5],ymm6[6],ymm5[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm13
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm3, %xmm1
; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1,2],ymm12[3],ymm11[4,5],ymm12[6],ymm11[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2,3,4,5],xmm1[6],xmm2[7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [2,3,0,1,14,15,12,13,10,11,10,11,10,11,10,11]
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm10[2],ymm9[3,4],ymm10[5],ymm9[6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <2,5,1,u,4,u,u,u>
+; AVX2-FAST-NEXT: vpermd %ymm3, %ymm4, %ymm3
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = [2,3,2,3,2,3,2,3,8,9,0,1,6,7,8,9,18,19,18,19,18,19,18,19,24,25,16,17,22,23,24,25]
+; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm3, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm12[2,3],ymm10[4,5],ymm12[6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm15
-; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm14
-; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0],ymm11[1],ymm9[2,3],ymm11[4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3,4,5],xmm3[6],xmm2[7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <2,6,1,u,5,u,u,u>
-; AVX2-FAST-NEXT: vpermd %ymm1, %ymm3, %ymm1
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [0,1,2,3,4,5,2,3,8,9,2,3,4,5,10,11,16,17,18,19,20,21,18,19,24,25,18,19,20,21,26,27]
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = [4,5,2,3,0,1,14,15,12,13,12,13,12,13,12,13]
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm2, %xmm2
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm1, %ymm4, %ymm1
+; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1,2],ymm6[3],ymm13[4,5],ymm6[6],ymm13[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3,4,5],xmm3[6],xmm4[7]
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm2
; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0],ymm8[1],ymm7[2,3],ymm8[4],ymm7[5,6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm8, %ymm12
-; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0],ymm11[1],ymm12[2,3],ymm11[4],ymm12[5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3,4,5],xmm2[6],xmm1[7]
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [4,5,2,3,0,1,14,15,12,13,12,13,12,13,12,13]
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm10[2,3],ymm9[4,5],ymm10[6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <2,6,1,u,5,u,u,u>
+; AVX2-FAST-NEXT: vpermd %ymm3, %ymm4, %ymm3
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = [0,1,2,3,4,5,2,3,8,9,2,3,4,5,10,11,16,17,18,19,20,21,18,19,24,25,18,19,20,21,26,27]
+; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm3, %ymm1
+; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm8, %ymm14
+; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermd %ymm1, %ymm4, %ymm1
+; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm1, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm13[2,3],ymm6[4,5],ymm13[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm2, %ymm3, %ymm2
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm8
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0],ymm11[1],ymm9[2,3,4],ymm11[5],ymm9[6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm10
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0],ymm13[1],ymm6[2,3],ymm13[4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5],xmm4[6],xmm3[7]
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm8
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0],ymm11[1],ymm12[2,3,4],ymm11[5],ymm12[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0,1,2],ymm15[3],ymm14[4,5],ymm15[6],ymm14[7]
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm2 = ymm2[1,3,2,3]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [6,7,6,7,6,7,6,7,8,9,4,5,10,11,0,1,22,23,22,23,22,23,22,23,24,25,20,21,26,27,16,17]
-; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = [6,7,4,5,2,3,0,1,14,15,14,15,14,15,14,15]
-; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm1, %xmm1
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm9
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0],ymm12[1],ymm7[2,3,4],ymm12[5],ymm7[6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm1, %xmm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2],ymm13[3],ymm6[4,5],ymm13[6],ymm6[7]
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm2 = ymm2[1,3,2,3]
-; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm6
-; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm5
-; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2],ymm5[3],ymm1[4,5],ymm5[6],ymm1[7]
-; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm3
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [6,7,4,5,2,3,0,1,14,15,14,15,14,15,14,15]
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2],ymm10[3],ymm9[4,5],ymm10[6],ymm9[7]
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,3,2,3]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [6,7,6,7,6,7,6,7,8,9,4,5,10,11,0,1,22,23,22,23,22,23,22,23,24,25,20,21,26,27,16,17]
+; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm3, %ymm9
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1,2],ymm7[3],ymm14[4,5],ymm7[6],ymm14[7]
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,3,2,3]
+; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0],ymm13[1],ymm6[2,3,4],ymm13[5],ymm6[6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2],xmm3[3],xmm4[4,5,6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm10
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1,2],ymm0[3],ymm10[4,5],ymm0[6],ymm10[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm5
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm15 = [3,6,2,5,3,6,2,5]
; AVX2-FAST-NEXT: # ymm15 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpermd %ymm2, %ymm15, %ymm2
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [0,1,2,3,0,1,6,7,8,9,14,15,12,13,14,15,16,17,18,19,16,17,22,23,24,25,30,31,28,29,30,31]
; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm14
-; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %ymm7
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm7[0,1,0,2]
-; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm2 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm13
+; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm6 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm0, %ymm13
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5,6],ymm13[7]
-; AVX2-FAST-NEXT: vpblendd $31, (%rsp), %ymm13, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3,4],ymm13[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm13
-; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm14
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm14[0,1,2],ymm13[3],ymm14[4,5],ymm13[6],ymm14[7]
+; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm2
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm2[0,1,2],ymm1[3],ymm2[4,5],ymm1[6],ymm2[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm14
+; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm3
; AVX2-FAST-NEXT: vpermd %ymm12, %ymm15, %ymm12
; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm12, %ymm4
-; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %ymm1
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm12 = ymm1[0,1,0,2]
-; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm15
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm12, %ymm11
+; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm13
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm12, %ymm11
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm11[7]
-; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpblendd $31, (%rsp), %ymm4, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3,4],ymm4[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm3[2],ymm5[3,4,5],ymm3[6],ymm5[7]
+; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm15
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm10[2],ymm5[3,4,5],ymm10[6],ymm5[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm11
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm11[4],xmm4[5],xmm11[6],xmm4[7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm11 = [0,1,0,1,14,15,12,13,10,11,8,9,12,13,14,15]
@@ -5216,7 +5223,7 @@
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1],ymm14[2],ymm13[3,4,5],ymm14[6],ymm13[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm14[2],ymm3[3,4,5],ymm14[6],ymm3[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm4
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm4[4],xmm0[5],xmm4[6],xmm0[7]
; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm0, %xmm0
@@ -5227,7 +5234,7 @@
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm10[3],ymm5[4,5],ymm10[6],ymm5[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,0,1,14,15,12,13,10,11,12,13,14,15]
@@ -5242,19 +5249,19 @@
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm14[3],ymm13[4,5],ymm14[6],ymm13[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2],ymm14[3],ymm3[4,5],ymm14[6],ymm3[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm11
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm11[0],xmm0[1],xmm11[2,3,4,5],xmm0[6],xmm11[7]
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd %ymm15, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermd %ymm13, %ymm4, %ymm2
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm0[1,2,3,4,5,6,7],ymm8[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0],ymm5[1],ymm3[2,3],ymm5[4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm5[1],ymm10[2,3],ymm5[4],ymm10[5,6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3,4,5],xmm2[6],xmm0[7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,4,5,2,3,0,1,14,15,12,13,12,13,14,15]
@@ -5267,561 +5274,568 @@
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm9[0],ymm0[1,2,3,4,5,6,7],ymm9[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0],ymm13[1],ymm14[2,3],ymm13[4],ymm14[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0],ymm3[1],ymm14[2,3],ymm3[4],ymm14[5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm7
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm9
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm9[1],xmm0[2,3,4,5],xmm9[6],xmm0[7]
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm2 = ymm15[0,1,1,3]
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm2 = ymm13[0,1,1,3]
; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm2, %ymm8
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm0[1,2,3,4,5,6,7],ymm6[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0],ymm5[1],ymm3[2,3,4],ymm5[5],ymm3[6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm15
-; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm11
-; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm5[1],ymm10[2,3,4],ymm5[5],ymm10[6,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm11
+; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm6
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm6[0],xmm0[1],xmm6[2],xmm0[3],xmm6[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = [0,1,6,7,4,5,2,3,0,1,14,15,12,13,14,15]
-; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm0, %xmm0
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm13 = [0,1,6,7,4,5,2,3,0,1,14,15,12,13,14,15]
+; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm0, %xmm0
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm7 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
-; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm4, %ymm4
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0,1,2,3,4,5,6],ymm4[7]
-; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm1
-; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm10[0,1],ymm3[2],ymm10[3,4],ymm3[5],ymm10[6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm5[0,1],ymm3[2],ymm5[3,4],ymm3[5],ymm5[6,7]
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm8[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm8
; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm10[0,1],ymm0[2],ymm10[3,4,5],ymm0[6],ymm10[7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = <0,3,7,2,6,u,u,u>
-; AVX2-FAST-NEXT: vpermd %ymm9, %ymm5, %ymm12
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm0[0,1],ymm10[2],ymm0[3,4,5],ymm10[6],ymm0[7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <0,3,7,2,6,u,u,u>
+; AVX2-FAST-NEXT: vpermd %ymm9, %ymm6, %ymm12
; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm12 = ymm12[u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1],ymm12[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm8[0],ymm4[1,2,3,4,5,6,7],ymm8[8],ymm4[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0],ymm13[1],ymm14[2,3,4],ymm13[5],ymm14[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0],ymm7[1],ymm14[2,3,4],ymm7[5],ymm14[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm8
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm8[0],xmm4[1],xmm8[2],xmm4[3],xmm8[4,5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm4, %xmm4
-; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm4, %xmm4
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm7[2],ymm5[3,4],ymm7[5],ymm5[6,7]
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm4[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm12[0,1],ymm9[2],ymm12[3,4,5],ymm9[6],ymm12[7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = <0,3,7,2,6,u,u,u>
-; AVX2-FAST-NEXT: vpermd %ymm6, %ymm8, %ymm6
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1],ymm14[2],ymm9[3,4,5],ymm14[6],ymm9[7]
+; AVX2-FAST-NEXT: vpermd %ymm4, %ymm6, %ymm4
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm1[0,1],ymm12[2],ymm1[3,4],ymm12[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm7 = xmm6[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm6
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm2[1,2,3,4,5,6,7],ymm4[8],ymm2[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,3,3,3,0,3,7,7]
-; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm15[0,1],ymm11[2],ymm15[3,4],ymm11[5],ymm15[6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermd %ymm13, %ymm2, %ymm4
+; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm7 = [0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25]
+; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm4, %ymm4
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm11[0,1],ymm15[2],ymm11[3,4],ymm15[5],ymm11[6,7]
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm11 = [2,5,1,4,2,5,1,4]
; AVX2-FAST-NEXT: # ymm11 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpermd %ymm8, %ymm11, %ymm8
; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31,u,u,u,u,u,u]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm8[0,1,2,3,4],ymm4[5,6,7],ymm8[8,9,10,11,12],ymm4[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm5[0,1],ymm3[2,3],ymm5[4,5],ymm3[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm6
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm6, %xmm6
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = [10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm8, %xmm8
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm8 = xmm8[10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm10[0,1,2],ymm0[3],ymm10[4,5],ymm0[6],ymm10[7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <0,4,7,3,6,u,u,u>
-; AVX2-FAST-NEXT: vpermd %ymm8, %ymm10, %ymm8
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm0[0,1,2],ymm10[3],ymm0[4,5],ymm10[6],ymm0[7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <0,4,7,3,6,u,u,u>
+; AVX2-FAST-NEXT: vpermd %ymm8, %ymm0, %ymm8
; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm8 = ymm8[u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1],ymm8[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm6[0],ymm4[1,2,3,4,5,6,7],ymm6[8],ymm4[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm4[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermd %ymm10, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm2, %ymm2
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermd %ymm4, %ymm2, %ymm2
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm14[0,1],ymm13[2],ymm14[3,4],ymm13[5],ymm14[6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm5[0,1],ymm4[2],ymm5[3,4],ymm4[5],ymm5[6,7]
; AVX2-FAST-NEXT: vpermd %ymm6, %ymm11, %ymm6
; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31,u,u,u,u,u,u]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm6[0,1,2,3,4],ymm2[5,6,7],ymm6[8,9,10,11,12],ymm2[13,14,15]
-; AVX2-FAST-NEXT: vmovdqa %ymm7, %ymm11
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm5[0,1],ymm7[2,3],ymm5[4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm1[0,1],ymm12[2,3],ymm1[4,5],ymm12[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm7
; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm7, %xmm0
-; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm6, %xmm6
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm12[0,1,2],ymm9[3],ymm12[4,5],ymm9[6],ymm12[7]
-; AVX2-FAST-NEXT: vpermd %ymm6, %ymm10, %ymm6
+; AVX2-FAST-NEXT: vmovdqa %ymm14, %ymm11
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0,1,2],ymm14[3],ymm9[4,5],ymm14[6],ymm9[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm14
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <0,4,7,3,6,u,u,u>
+; AVX2-FAST-NEXT: vpermd %ymm6, %ymm1, %ymm6
; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm0[0],ymm2[1,2,3,4,5,6,7],ymm0[8],ymm2[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = mem[0,1],ymm15[2,3],mem[4,5],ymm15[6,7]
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [0,4,7,0,0,4,7,0]
; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm5 = [0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27]
-; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpermd %ymm13, %ymm1, %ymm3
+; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm13 = [0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27]
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm3, %ymm3
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm6 = [2,6,1,5,2,6,1,5]
; AVX2-FAST-NEXT: # ymm6 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm6, %ymm0
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm7 = [28,29,18,19,20,21,26,27,28,29,18,19,20,21,26,27,28,29,18,19,20,21,26,27,28,29,18,19,20,21,26,27]
; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7],ymm0[8,9,10,11,12],ymm3[13,14,15]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3],mem[4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm8 = mem[0,1,2],ymm8[3],mem[4,5],ymm8[6],mem[7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm9
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm10 = [0,1,2,3,4,5,6,7,10,11,6,7,4,5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm9, %xmm9
-; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,6,4,6,7]
-; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm8 = xmm8[4],xmm9[4],xmm8[5],xmm9[5],xmm8[6],xmm9[6],xmm8[7],xmm9[7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = <1,4,0,3,7,u,u,u>
-; AVX2-FAST-NEXT: vpermd %ymm3, %ymm9, %ymm3
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm2[0,1,2],mem[3],ymm2[4,5],mem[6],ymm2[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm8
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm9 = [0,1,2,3,4,5,6,7,10,11,6,7,4,5,6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm8, %xmm8
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,4,6,7]
+; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0],ymm2[1],mem[2,3],ymm2[4],mem[5,6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <1,4,0,3,7,u,u,u>
+; AVX2-FAST-NEXT: vpermd %ymm8, %ymm2, %ymm8
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31>
-; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm3, %ymm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm8, %ymm8
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm8[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4,5,6,7],ymm3[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm10, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm6, %ymm3
; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm3, %ymm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5,6,7],ymm3[8,9,10,11,12],ymm1[13,14,15]
-; AVX2-FAST-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1,2],ymm11[3],mem[4,5],ymm11[6],mem[7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm5
-; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm5, %xmm5
-; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,4,6,7]
-; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
-; AVX2-FAST-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0],ymm12[1],mem[2,3],ymm12[4],mem[5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm9, %ymm5
-; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0],ymm14[1],ymm11[2,3],ymm14[4],ymm11[5,6,7]
+; AVX2-FAST-NEXT: vpermd %ymm3, %ymm2, %ymm3
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,1,2],ymm12[3],mem[4,5],ymm12[6],mem[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm6, %xmm6
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,4,6,7]
+; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0],ymm1[1,2,3,4,5,6,7],ymm3[8],ymm1[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rsi)
; AVX2-FAST-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm3, (%rsi)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rdx)
+; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rsi)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm3, (%rdx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rdx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm3, (%rcx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%r8)
+; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%rcx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm3, (%r8)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%r9)
+; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%r8)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm3, (%r9)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm3, 32(%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovdqa %ymm2, 32(%rax)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm2, (%rax)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm2, 32(%rax)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovdqa %ymm1, 32(%rax)
-; AVX2-FAST-NEXT: vmovdqa %ymm0, (%rax)
-; AVX2-FAST-NEXT: addq $584, %rsp # imm = 0x248
+; AVX2-FAST-NEXT: vmovdqa %ymm1, (%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm0, 32(%rax)
+; AVX2-FAST-NEXT: addq $680, %rsp # imm = 0x2A8
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
;
; AVX2-FAST-PERLANE-LABEL: load_i16_stride7_vf32:
; AVX2-FAST-PERLANE: # %bb.0:
; AVX2-FAST-PERLANE-NEXT: subq $520, %rsp # imm = 0x208
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm8
; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %ymm10
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm11
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm6[1],ymm1[2,3,4],ymm6[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm1[2],ymm2[3,4,5],ymm1[6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5],xmm1[6],xmm0[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,14,15,12,13,10,11,8,9,8,9,8,9,8,9]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0],ymm10[1],ymm3[2,3,4],ymm10[5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm11
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = [0,1,14,15,12,13,14,15,2,3,6,7,12,13,2,3,16,17,30,31,28,29,30,31,18,19,22,23,28,29,18,19]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1],ymm4[2],ymm3[3,4,5],ymm4[6],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4],xmm1[5],xmm3[6],xmm1[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,14,15,12,13,10,11,8,9,8,9,8,9,8,9]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm4, %ymm0, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0,1],ymm5[2],ymm12[3,4,5],ymm5[6],ymm12[7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm5[4],xmm4[5],xmm5[6],xmm4[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm4, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm11[0],ymm10[1],ymm11[2,3,4],ymm10[5],ymm11[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm4, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm3, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1],ymm10[2],ymm11[3,4],ymm10[5],ymm11[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, %ymm13
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3],ymm2[4,5,6,7,8,9,10],ymm3[11],ymm2[12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm3 = [2,3,2,3,2,3,2,3,8,9,8,9,6,7,4,5,18,19,18,19,18,19,18,19,24,25,24,25,22,23,20,21]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,14,15,12,13,14,15,2,3,6,7,12,13,2,3,16,17,30,31,28,29,30,31,18,19,22,23,28,29,18,19]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm0, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm2, %ymm4, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0],ymm7[1],ymm8[2,3,4],ymm7[5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0,1,2],ymm0[3],ymm12[4,5],ymm0[6],ymm12[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm5 = [2,3,0,1,14,15,12,13,10,11,10,11,10,11,10,11]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm5, %xmm4, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm4, %ymm2, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1],ymm6[2],ymm9[3,4],ymm6[5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7,8,9,10],ymm4[11],ymm2[12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1,2],ymm7[3],ymm8[4,5],ymm7[6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm6[2],ymm5[3,4,5],ymm6[6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm3, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm10[2],ymm11[3,4],ymm10[5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7,8,9,10],ymm2[11],ymm1[12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = [2,3,2,3,2,3,2,3,8,9,8,9,6,7,4,5,18,19,18,19,18,19,18,19,24,25,24,25,22,23,20,21]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1,2],ymm12[3],ymm13[4,5],ymm12[6],ymm13[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3,4,5],xmm3[6],xmm4[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm5, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm3, %ymm2, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm6[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm6[2,3],ymm9[4,5],ymm6[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4],ymm3[5,6,7,8,9,10,11],ymm2[12],ymm3[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0],ymm8[1],ymm7[2,3],ymm8[4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm8, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5],xmm4[6],xmm3[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = [2,3,0,1,14,15,12,13,10,11,10,11,10,11,10,11]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm3, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm3, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3],ymm1[4,5,6,7,8,9,10],ymm3[11],ymm1[12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2],ymm6[3],ymm5[4,5],ymm6[6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2,3,4,5],xmm2[6],xmm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm2, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0],ymm13[1],ymm12[2,3],ymm13[4],ymm12[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3,4,5],xmm2[6],xmm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm10[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7,8,9,10,11],ymm3[12],ymm2[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [4,5,2,3,0,1,14,15,12,13,12,13,12,13,12,13]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm5 = [4,5,4,5,4,5,4,5,8,9,10,11,8,9,6,7,20,21,20,21,20,21,20,21,24,25,26,27,24,25,22,23]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm8 = [4,5,2,3,0,1,14,15,12,13,12,13,12,13,12,13]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm3, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0],ymm12[1],ymm14[2,3],ymm12[4],ymm14[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3,4,5],xmm3[6],xmm2[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm2, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1],ymm10[2,3],ymm13[4,5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm8 = ymm10[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4],ymm3[5,6,7,8,9,10,11],ymm8[12],ymm3[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm2, %ymm3, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0],ymm0[1],ymm7[2,3,4],ymm0[5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3],xmm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1,2],ymm6[3],ymm11[4,5],ymm6[6],ymm11[7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm7[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm8, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4],ymm2[5,6,7,8,9,10,11],ymm1[12],ymm2[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0],ymm9[1],ymm6[2,3],ymm9[4],ymm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm5[1],xmm2[2,3,4,5],xmm5[6],xmm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm2, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [6,7,4,5,2,3,0,1,14,15,14,15,14,15,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1,2],ymm14[3],ymm11[4,5],ymm14[6],ymm11[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[1,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm5 = [6,7,6,7,6,7,6,7,8,9,4,5,10,11,0,1,22,23,22,23,22,23,22,23,24,25,20,21,26,27,16,17]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm8 = [6,7,4,5,2,3,0,1,14,15,14,15,14,15,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm2, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm2, %ymm3, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0],ymm12[1],ymm14[2,3,4],ymm12[5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm12
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm12[0],xmm3[1],xmm12[2],xmm3[3],xmm12[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm8 = ymm13[0,1,2],ymm10[3],ymm13[4,5],ymm10[6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm8 = ymm8[1,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm8, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm3, %ymm5, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1,2],ymm5[3],ymm6[4,5],ymm5[6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,2,3,4,5,6,7,8,9,10,11,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
-; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm8[0,1,0,2]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm7 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm3, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm13
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1,2],ymm13[3],ymm12[4,5],ymm13[6],ymm12[7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm15
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm15, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm2, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm7[3],ymm10[4,5],ymm7[6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0],ymm4[1],ymm6[2,3,4],ymm4[5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm7[0],xmm5[1],xmm7[2],xmm5[3],xmm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm5, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm3, %ymm1, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1,2],ymm0[3],ymm10[4,5],ymm0[6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,2,3,4,5,6,7,8,9,10,11,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm5, %xmm7, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,4,7]
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm7[4],xmm3[5],xmm7[5],xmm3[6],xmm7[6],xmm3[7],xmm7[7]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm7[0,1,0,2]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm3 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm1, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $31, (%rsp), %ymm14, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2],ymm14[3],ymm15[4,5],ymm14[6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm13
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm5, %xmm13, %xmm5
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
-; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm15 = ymm4[0,1,0,2]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm15, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm13 = ymm5[0,1,0,2]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm13, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm12[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1],ymm12[2],ymm13[3,4,5],ymm12[6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm14
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm14[4],xmm0[5],xmm14[6],xmm0[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm14 = [0,1,0,1,14,15,12,13,10,11,8,9,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm10[2],ymm4[3,4,5],ymm10[6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm12[4],xmm0[5],xmm12[6],xmm0[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm12 = [0,1,0,1,14,15,12,13,10,11,8,9,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm1 = [22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm15, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1],ymm6[2],ymm5[3,4,5],ymm6[6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm15
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm15[4],xmm0[5],xmm15[6],xmm0[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm3 = [22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm6[3],ymm5[4,5],ymm6[6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1],ymm15[2],ymm14[3,4,5],ymm15[6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5],xmm1[6],xmm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm13, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm10[3],ymm4[4,5],ymm10[6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3,4,5],xmm0[6],xmm1[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,0,1,14,15,12,13,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm7[0,1,1,2]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm12 = [16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm12, %ymm3, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4,5,6,7],ymm3[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm15[3],ymm14[4,5],ymm15[6],ymm14[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0],xmm0[1],xmm3[2,3,4,5],xmm0[6],xmm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,2,3,0,1,14,15,12,13,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm5[0,1,1,2]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm12, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm8[0,1,1,2]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm14 = [16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm5, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm12[3],ymm13[4,5],ymm12[6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0],xmm0[1],xmm5[2,3,4,5],xmm0[6],xmm5[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm4[0,1,1,2]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm9[0],ymm0[1,2,3,4,5,6,7],ymm9[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0],ymm1[1],ymm6[2,3],ymm1[4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2,3,4,5],xmm3[6],xmm0[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,4,5,2,3,0,1,14,15,12,13,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm4[1],ymm10[2,3],ymm4[4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,4,5,2,3,0,1,14,15,12,13,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm8[0,1,1,3]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm8 = [18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm8, %ymm5, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm7[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm7 = [18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm3, %ymm9
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm9[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0],ymm0[1,2,3,4,5,6,7],ymm11[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm0[1,2,3,4,5,6,7],ymm8[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0],ymm13[1],ymm12[2,3],ymm13[4],ymm12[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3,4,5],xmm7[6],xmm0[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0],ymm14[1],ymm15[2,3],ymm14[4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm8[1],xmm0[2,3,4,5],xmm8[6],xmm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm4[0,1,1,3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm8, %ymm3, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm4[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm10[0],ymm0[1,2,3,4,5,6,7],ymm10[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm5[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm7, %ymm1, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm0[1,2,3,4,5,6,7],ymm6[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm14[2],ymm15[3,4,5],ymm14[6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm4 = <u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm0, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1],ymm11[2],ymm13[3,4,5],ymm11[6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6],ymm6[7,8,9,10,11,12,13],ymm0[14],ymm6[15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1],ymm11[2],ymm8[3,4],ymm11[5],ymm8[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm7 = xmm6[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm10[0],ymm4[1],ymm10[2,3,4],ymm4[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm7
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0],xmm6[1],xmm7[2],xmm6[3],xmm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm7 = [0,1,6,7,4,5,2,3,0,1,14,15,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm6, %xmm6
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm10 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm10, %ymm3, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm2 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm3, %ymm3
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm0[0],ymm3[1,2,3,4,5,6,7],ymm0[8],ymm3[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0],ymm14[1],ymm15[2,3,4],ymm14[5],ymm15[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0],xmm0[1],xmm3[2],xmm0[3],xmm3[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm10, %ymm5, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $187, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1],ymm1[2],mem[3,4,5],ymm1[6],mem[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm3, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6],ymm4[7,8,9,10,11,12,13],ymm3[14],ymm4[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $187, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm7[2],mem[3,4,5],ymm7[6],mem[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm1, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6],ymm3[7,8,9,10,11,12,13],ymm1[14],ymm3[15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm7 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm4, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4,5,6,7],ymm3[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm6 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm3, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2],ymm14[3],ymm15[4,5],ymm14[6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm0[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4,5,6],ymm3[7,8],ymm0[9,10,11,12,13,14],ymm3[15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1],ymm11[2,3],ymm8[4,5],ymm11[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm4, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm12[3],ymm13[4,5],ymm12[6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7,8],ymm0[9,10,11,12,13,14],ymm1[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm11[2,3],ymm8[4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm3, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm6 = [10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0,1],ymm13[2],ymm12[3,4],ymm13[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm3[2],xmm4[3],xmm3[4],xmm4[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1],ymm9[2],ymm10[3,4],ymm9[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0,1],xmm0[2],xmm3[3],xmm0[4],xmm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm5 = [8,9,8,9,6,7,4,5,2,3,2,3,2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm5, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 432(%rdi), %xmm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm10 = xmm3[0,1,2,3,4,5],xmm4[6],xmm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm14 = [0,1,2,3,4,5,6,7,0,1,0,1,14,15,12,13]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm10, %xmm10
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm10[5,6,7],ymm7[8,9,10,11,12],ymm10[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm7 = ymm0[0],ymm7[1,2,3,4,5,6,7],ymm0[8],ymm7[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1],ymm6[2],ymm9[3,4],ymm6[5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm7[0,1],xmm0[2],xmm7[3],xmm0[4],xmm7[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm5, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 432(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm10 = xmm3[0,1,2,3,4,5],xmm0[6],xmm3[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm12 = [0,1,2,3,4,5,6,7,0,1,0,1,14,15,12,13]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm10, %xmm10
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5,6,7],ymm8[8,9,10,11,12],ymm10[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm8 = ymm1[0],ymm8[1,2,3,4,5,6,7],ymm1[8],ymm8[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm14[2],ymm15[3,4],ymm14[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm8[0,1],xmm1[2],xmm8[3],xmm1[4],xmm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm5, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa 208(%rdi), %xmm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %xmm15
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm7 = xmm15[0,1,2,3,4,5],xmm5[6],xmm15[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm7, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5,6,7],ymm0[8,9,10,11,12],ymm7[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm7, %xmm10
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %xmm13
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm8 = xmm13[0,1,2,3,4,5],xmm5[6],xmm13[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm8, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5,6,7],ymm1[8,9,10,11,12],ymm8[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm8 = ymm2[0,1],ymm4[2,3],ymm2[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm8, %xmm10
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm9 = xmm10[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm6 = xmm7[10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm8, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm7 = ymm14[0,1,2],ymm10[3],ymm14[4,5],ymm10[6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm9 = ymm7[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm7 = ymm9[0],ymm7[1,2,3,4,5,6],ymm9[7,8],ymm7[9,10,11,12,13,14],ymm9[15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm8 = ymm10[0,1,2],ymm7[3],ymm10[4,5],ymm7[6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm9 = ymm8[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm8 = ymm9[0],ymm8[1,2,3,4,5,6],ymm9[7,8],ymm8[9,10,11,12,13,14],ymm9[15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm7 = ymm8[u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm0[1,2,3,4,5,6,7],ymm6[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm9 = ymm6[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $237, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm0[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm6[1],ymm0[2,3,4,5,6,7,8],ymm6[9],ymm0[10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm8[0,1,2],mem[3],ymm8[4,5],mem[6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm6[0],ymm1[1,2,3,4,5,6,7],ymm6[8],ymm1[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm9 = ymm6[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3],ymm1[4],mem[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2,3,4,5,6,7,8],ymm6[9],ymm1[10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1,2],ymm2[3],mem[4,5],ymm2[6],mem[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm7, %xmm8
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm6 = [0,1,2,3,4,5,6,7,10,11,6,7,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm8, %xmm8
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm7 = xmm7[4],xmm8[4],xmm7[5],xmm8[5],xmm7[6],xmm8[6],xmm7[7],xmm8[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm8, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm13[2,3],ymm12[4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3],xmm2[4],xmm1[5],xmm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0],xmm3[1],xmm4[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm8, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm11[0,1],mem[2,3],ymm11[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2],xmm2[3],xmm4[4],xmm2[5],xmm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = [8,9,10,11,8,9,6,7,4,5,4,5,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,2,3,4,5,6,7,0,1,2,3,0,1,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm2, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7],ymm1[8,9,10,11,12],ymm2[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3,4,5,6,7],ymm0[8],ymm1[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm11[0,1],mem[2,3],ymm11[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,2,3,4,5,6,7,0,1,2,3,0,1,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7],ymm2[8,9,10,11,12],ymm0[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm14[2,3],ymm15[4,5],ymm14[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3],xmm2[4],xmm1[5],xmm2[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm5[0],xmm15[1],xmm5[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm5[0],xmm13[1],xmm5[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7],ymm1[8,9,10,11,12],ymm2[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm14[1],ymm10[2,3],ymm14[4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0],ymm10[1],ymm12[2,3],ymm10[4],ymm12[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4,5,6,7,8],ymm3[9],ymm2[10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm8, %ymm2, %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm3[0,1,2],mem[3],ymm3[4,5],mem[6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2],ymm3[3],mem[4,5],ymm3[6],mem[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm4, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,4,6,7]
@@ -5830,32 +5844,32 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3,4,5,6,7],ymm2[8],ymm1[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rsi)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rdx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%r8)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%r8)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, (%rax)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 32(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, (%rax)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, 32(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, (%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, 32(%rax)
; AVX2-FAST-PERLANE-NEXT: addq $520, %rsp # imm = 0x208
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
@@ -7339,152 +7353,84 @@
define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind {
; SSE-LABEL: load_i16_stride7_vf64:
; SSE: # %bb.0:
-; SSE-NEXT: subq $1352, %rsp # imm = 0x548
-; SSE-NEXT: movdqa 640(%rdi), %xmm5
+; SSE-NEXT: subq $1368, %rsp # imm = 0x558
+; SSE-NEXT: movdqa 752(%rdi), %xmm5
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 624(%rdi), %xmm11
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rdi), %xmm12
-; SSE-NEXT: movdqa 128(%rdi), %xmm14
-; SSE-NEXT: movaps 160(%rdi), %xmm7
+; SSE-NEXT: movdqa 736(%rdi), %xmm6
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 224(%rdi), %xmm10
+; SSE-NEXT: movdqa 240(%rdi), %xmm12
+; SSE-NEXT: movaps 272(%rdi), %xmm7
; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 144(%rdi), %xmm6
-; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm9
-; SSE-NEXT: movdqa 176(%rdi), %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 208(%rdi), %xmm10
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[0,0,0,0]
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 256(%rdi), %xmm8
+; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 304(%rdi), %xmm9
+; SSE-NEXT: movdqa 288(%rdi), %xmm13
+; SSE-NEXT: movdqa 320(%rdi), %xmm11
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[0,0,0,0]
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,65535,65535,0]
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm13, %xmm0
+; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa {{.*#+}} xmm15 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movaps %xmm6, %xmm0
+; SSE-NEXT: movaps %xmm8, %xmm0
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm7[2,2]
-; SSE-NEXT: movaps {{.*#+}} xmm6 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT: movaps %xmm6, %xmm3
+; SSE-NEXT: movaps {{.*#+}} xmm7 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT: movaps %xmm7, %xmm3
; SSE-NEXT: andnps %xmm0, %xmm3
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm12[0,1,0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm10[0,1,0,3]
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: movdqa 656(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: movdqa 768(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: pand %xmm7, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
; SSE-NEXT: por %xmm1, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm11, %xmm0
+; SSE-NEXT: movdqa %xmm6, %xmm0
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movaps 608(%rdi), %xmm3
+; SSE-NEXT: movaps 720(%rdi), %xmm3
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 592(%rdi), %xmm0
-; SSE-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: movaps 704(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT: movaps %xmm6, %xmm3
+; SSE-NEXT: movaps %xmm7, %xmm3
; SSE-NEXT: andnps %xmm0, %xmm3
-; SSE-NEXT: movdqa 560(%rdi), %xmm7
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[0,1,0,3]
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 672(%rdi), %xmm6
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,1,0,3]
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa 576(%rdi), %xmm5
+; SSE-NEXT: movdqa 688(%rdi), %xmm5
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: pand %xmm7, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
; SSE-NEXT: por %xmm1, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 96(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa 80(%rdi), %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 64(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: pand %xmm2, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movaps 32(%rdi), %xmm0
-; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 48(%rdi), %xmm3
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT: movaps %xmm6, %xmm3
-; SSE-NEXT: andnps %xmm0, %xmm3
-; SSE-NEXT: movdqa (%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa 16(%rdi), %xmm8
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
+; SSE-NEXT: movdqa 432(%rdi), %xmm8
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[0,0,0,0]
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: pand %xmm6, %xmm4
-; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
-; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 544(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa 528(%rdi), %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 512(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: pand %xmm2, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movaps 496(%rdi), %xmm3
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 480(%rdi), %xmm0
-; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT: movaps %xmm6, %xmm3
-; SSE-NEXT: andnps %xmm0, %xmm3
-; SSE-NEXT: movdqa 448(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa 464(%rdi), %xmm11
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: pand %xmm6, %xmm4
-; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
-; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 432(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa 416(%rdi), %xmm3
@@ -7494,14 +7440,14 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movaps 384(%rdi), %xmm3
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 368(%rdi), %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT: movaps %xmm6, %xmm3
+; SSE-NEXT: movaps %xmm7, %xmm3
; SSE-NEXT: andnps %xmm0, %xmm3
; SSE-NEXT: movdqa 336(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7511,9 +7457,9 @@
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: pand %xmm7, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
; SSE-NEXT: por %xmm1, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 880(%rdi), %xmm0
@@ -7528,14 +7474,14 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm15, %xmm13
-; SSE-NEXT: pandn %xmm0, %xmm13
+; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movaps 832(%rdi), %xmm3
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 816(%rdi), %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT: movaps %xmm6, %xmm3
+; SSE-NEXT: movaps %xmm7, %xmm3
; SSE-NEXT: andnps %xmm0, %xmm3
; SSE-NEXT: movdqa 784(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7545,111 +7491,179 @@
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: pand %xmm7, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
-; SSE-NEXT: por %xmm13, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
+; SSE-NEXT: por %xmm1, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 320(%rdi), %xmm0
+; SSE-NEXT: movdqa 96(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa 304(%rdi), %xmm3
+; SSE-NEXT: movdqa 80(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 288(%rdi), %xmm0
+; SSE-NEXT: movdqa 64(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movaps 272(%rdi), %xmm3
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 256(%rdi), %xmm0
+; SSE-NEXT: movaps 32(%rdi), %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 48(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT: movaps %xmm6, %xmm3
+; SSE-NEXT: movaps %xmm7, %xmm3
; SSE-NEXT: andnps %xmm0, %xmm3
-; SSE-NEXT: movdqa 224(%rdi), %xmm0
+; SSE-NEXT: movdqa (%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa 240(%rdi), %xmm0
+; SSE-NEXT: movdqa 16(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: pand %xmm7, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
; SSE-NEXT: por %xmm1, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 768(%rdi), %xmm0
+; SSE-NEXT: movdqa 544(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa 752(%rdi), %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 736(%rdi), %xmm0
+; SSE-NEXT: movdqa 528(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 512(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movaps 720(%rdi), %xmm3
+; SSE-NEXT: movaps 496(%rdi), %xmm3
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 704(%rdi), %xmm0
+; SSE-NEXT: movaps 480(%rdi), %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT: movaps %xmm6, %xmm3
+; SSE-NEXT: movaps %xmm7, %xmm3
; SSE-NEXT: andnps %xmm0, %xmm3
-; SSE-NEXT: movdqa 672(%rdi), %xmm0
+; SSE-NEXT: movdqa 448(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT: movdqa 688(%rdi), %xmm0
+; SSE-NEXT: movdqa 464(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: pand %xmm6, %xmm4
+; SSE-NEXT: pand %xmm7, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
; SSE-NEXT: por %xmm1, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pslldq {{.*#+}} xmm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm10[0,1,2,3,4,5]
+; SSE-NEXT: movdqa 208(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm10, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movdqa 192(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 176(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movaps 160(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 144(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
+; SSE-NEXT: movaps %xmm7, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm3
+; SSE-NEXT: movdqa 112(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT: movdqa 128(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; SSE-NEXT: pand %xmm7, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
+; SSE-NEXT: por %xmm1, %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 656(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movdqa 640(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 624(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movaps 608(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 592(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
+; SSE-NEXT: movaps %xmm7, %xmm3
+; SSE-NEXT: andnps %xmm0, %xmm3
+; SSE-NEXT: movdqa 560(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT: movdqa 576(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; SSE-NEXT: pand %xmm7, %xmm4
+; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
+; SSE-NEXT: por %xmm1, %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pslldq {{.*#+}} xmm11 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm11[0,1,2,3,4,5]
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: pandn %xmm11, %xmm1
; SSE-NEXT: psrld $16, %xmm9
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1]
-; SSE-NEXT: pand %xmm2, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: movdqa {{.*#+}} xmm13 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm13, %xmm1
-; SSE-NEXT: pandn %xmm12, %xmm1
-; SSE-NEXT: pand %xmm13, %xmm14
-; SSE-NEXT: por %xmm1, %xmm14
+; SSE-NEXT: psrldq {{.*#+}} xmm13 = xmm13[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm9[0],xmm13[1],xmm9[1]
+; SSE-NEXT: pand %xmm2, %xmm13
+; SSE-NEXT: por %xmm1, %xmm13
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm13, %xmm0
+; SSE-NEXT: movdqa {{.*#+}} xmm4 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm4, %xmm1
+; SSE-NEXT: pandn %xmm10, %xmm1
+; SSE-NEXT: pand %xmm4, %xmm12
+; SSE-NEXT: movdqa %xmm4, %xmm15
+; SSE-NEXT: por %xmm1, %xmm12
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT: movdqa %xmm6, %xmm4
+; SSE-NEXT: movdqa %xmm7, %xmm4
; SSE-NEXT: pandn %xmm1, %xmm4
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[0,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
+; SSE-NEXT: pand %xmm7, %xmm1
; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
; SSE-NEXT: por %xmm0, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
@@ -7660,31 +7674,30 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT: pand %xmm2, %xmm3
; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: movdqa %xmm14, %xmm0
; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm13, %xmm1
-; SSE-NEXT: pandn %xmm7, %xmm1
-; SSE-NEXT: pand %xmm13, %xmm5
+; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: pandn %xmm6, %xmm1
+; SSE-NEXT: pand %xmm15, %xmm5
; SSE-NEXT: por %xmm1, %xmm5
-; SSE-NEXT: movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT: movdqa %xmm6, %xmm4
+; SSE-NEXT: movdqa %xmm7, %xmm4
; SSE-NEXT: pandn %xmm1, %xmm4
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
+; SSE-NEXT: pand %xmm7, %xmm1
; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
; SSE-NEXT: por %xmm0, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
+; SSE-NEXT: pslldq {{.*#+}} xmm8 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm8[0,1,2,3,4,5]
; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm8, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
@@ -7692,97 +7705,66 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT: pand %xmm2, %xmm3
; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: movdqa %xmm14, %xmm0
; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm15, %xmm1
; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: pand %xmm13, %xmm8
-; SSE-NEXT: por %xmm1, %xmm8
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pand %xmm15, %xmm3
+; SSE-NEXT: por %xmm1, %xmm3
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT: movdqa %xmm6, %xmm4
-; SSE-NEXT: pandn %xmm1, %xmm4
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
-; SSE-NEXT: por %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
-; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: pand %xmm2, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm13, %xmm1
-; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: pand %xmm13, %xmm11
-; SSE-NEXT: por %xmm1, %xmm11
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT: movdqa %xmm6, %xmm4
-; SSE-NEXT: pandn %xmm1, %xmm4
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
-; SSE-NEXT: por %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
-; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: pand %xmm2, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm13, %xmm1
-; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm14, %xmm1
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT: movdqa %xmm6, %xmm4
+; SSE-NEXT: movdqa %xmm7, %xmm4
; SSE-NEXT: pandn %xmm1, %xmm4
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
+; SSE-NEXT: pand %xmm7, %xmm1
; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
+; SSE-NEXT: pand %xmm14, %xmm4
; SSE-NEXT: por %xmm0, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: psrld $16, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT: pand %xmm2, %xmm3
+; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pand %xmm15, %xmm3
+; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
+; SSE-NEXT: movdqa %xmm7, %xmm5
+; SSE-NEXT: pandn %xmm1, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm7, %xmm1
+; SSE-NEXT: por %xmm1, %xmm5
+; SSE-NEXT: pand %xmm14, %xmm5
+; SSE-NEXT: por %xmm0, %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
@@ -7790,12 +7772,12 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT: pand %xmm2, %xmm3
; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: movdqa %xmm14, %xmm0
; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm15, %xmm1
; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm3
+; SSE-NEXT: pand %xmm15, %xmm3
; SSE-NEXT: por %xmm1, %xmm3
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -7803,88 +7785,120 @@
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT: movdqa %xmm6, %xmm4
-; SSE-NEXT: pandn %xmm1, %xmm4
+; SSE-NEXT: movdqa %xmm7, %xmm5
+; SSE-NEXT: pandn %xmm1, %xmm5
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: pand %xmm15, %xmm4
-; SSE-NEXT: por %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pand %xmm7, %xmm1
+; SSE-NEXT: por %xmm1, %xmm5
+; SSE-NEXT: pand %xmm14, %xmm5
+; SSE-NEXT: por %xmm0, %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT: pand %xmm2, %xmm3
+; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pand %xmm15, %xmm3
+; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
+; SSE-NEXT: movdqa %xmm7, %xmm5
+; SSE-NEXT: pandn %xmm1, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm7, %xmm1
+; SSE-NEXT: por %xmm1, %xmm5
+; SSE-NEXT: pand %xmm14, %xmm5
+; SSE-NEXT: por %xmm0, %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: psrld $16, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT: pand %xmm2, %xmm3
+; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pand %xmm15, %xmm3
+; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
+; SSE-NEXT: movdqa %xmm7, %xmm5
+; SSE-NEXT: pandn %xmm1, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm7, %xmm1
+; SSE-NEXT: por %xmm1, %xmm5
+; SSE-NEXT: pand %xmm14, %xmm5
+; SSE-NEXT: por %xmm0, %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm10, %xmm0
; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT: movdqa %xmm11, %xmm0
; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm8, %xmm3
-; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: pand %xmm2, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm13, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: pandn %xmm9, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm3
-; SSE-NEXT: pand %xmm13, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT: movdqa %xmm6, %xmm7
-; SSE-NEXT: pandn %xmm1, %xmm7
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
-; SSE-NEXT: por %xmm1, %xmm7
-; SSE-NEXT: pand %xmm15, %xmm7
-; SSE-NEXT: por %xmm0, %xmm7
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
-; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm9, %xmm3
; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT: pand %xmm2, %xmm3
; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: movdqa %xmm14, %xmm0
; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm13, %xmm1
-; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: pand %xmm13, %xmm3
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT: pand %xmm6, %xmm1
+; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: pandn %xmm12, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT: movdqa %xmm13, %xmm3
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
+; SSE-NEXT: pand %xmm15, %xmm3
+; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
+; SSE-NEXT: pand %xmm7, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm5, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,1,2,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,7,7,7]
-; SSE-NEXT: pandn %xmm3, %xmm6
-; SSE-NEXT: por %xmm1, %xmm6
-; SSE-NEXT: pand %xmm15, %xmm6
-; SSE-NEXT: por %xmm0, %xmm6
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: pandn %xmm3, %xmm7
+; SSE-NEXT: por %xmm1, %xmm7
+; SSE-NEXT: pand %xmm14, %xmm7
+; SSE-NEXT: por %xmm0, %xmm7
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $68, (%rsp), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,0,1]
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
@@ -7894,343 +7908,366 @@
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: pandn %xmm0, %xmm3
; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,4,7]
; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[2,2,3,3]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm0[1]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
-; SSE-NEXT: movss {{.*#+}} xmm3 = xmm0[0],xmm3[1,2,3]
-; SSE-NEXT: andps %xmm15, %xmm3
-; SSE-NEXT: orps %xmm1, %xmm3
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
+; SSE-NEXT: andps %xmm14, %xmm1
+; SSE-NEXT: orps %xmm3, %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,0,1]
-; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: pandn %xmm0, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: por %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm14, %xmm4
+; SSE-NEXT: pandn %xmm0, %xmm4
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,2,3,3]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT: movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
+; SSE-NEXT: andps %xmm14, %xmm1
+; SSE-NEXT: orps %xmm4, %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,0,1]
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: pandn %xmm0, %xmm4
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
; SSE-NEXT: pand %xmm2, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm0[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[2,2,3,3]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm0[1]
+; SSE-NEXT: por %xmm4, %xmm0
+; SSE-NEXT: movdqa %xmm14, %xmm4
+; SSE-NEXT: pandn %xmm0, %xmm4
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[2,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT: movss {{.*#+}} xmm6 = xmm3[0],xmm6[1,2,3]
-; SSE-NEXT: andps %xmm15, %xmm6
-; SSE-NEXT: orps %xmm1, %xmm6
-; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[0,1,0,1]
-; SSE-NEXT: movdqa %xmm2, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm11[2],xmm8[3],xmm11[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[0,1,0,1]
-; SSE-NEXT: pand %xmm2, %xmm1
-; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm15, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm9, %xmm6
-; SSE-NEXT: punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm10[0],xmm6[1],xmm10[1],xmm6[2],xmm10[2],xmm6[3],xmm10[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[2,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
-; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
-; SSE-NEXT: andps %xmm15, %xmm0
-; SSE-NEXT: orps %xmm3, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm0[2,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[0,3,2,3,4,5,6,7]
+; SSE-NEXT: movss {{.*#+}} xmm1 = xmm6[0],xmm1[1,2,3]
+; SSE-NEXT: andps %xmm14, %xmm1
+; SSE-NEXT: orps %xmm4, %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm10[0,1,0,1]
+; SSE-NEXT: movdqa %xmm2, %xmm6
+; SSE-NEXT: pandn %xmm4, %xmm6
+; SSE-NEXT: punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm9[0,1,0,1]
+; SSE-NEXT: pand %xmm2, %xmm4
+; SSE-NEXT: por %xmm6, %xmm4
+; SSE-NEXT: movdqa %xmm14, %xmm6
+; SSE-NEXT: pandn %xmm4, %xmm6
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm5[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm8[2,2,3,3]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm4[1]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm13[0],xmm12[1],xmm13[1],xmm12[2],xmm13[2],xmm12[3],xmm13[3]
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm12[2,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,3,2,3,4,5,6,7]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm7[0],xmm0[1,2,3]
+; SSE-NEXT: andps %xmm14, %xmm0
+; SSE-NEXT: orps %xmm6, %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,0,1]
-; SSE-NEXT: movdqa %xmm2, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
-; SSE-NEXT: pand %xmm2, %xmm1
-; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm15, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: pshufd $196, (%rsp), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,2,3,3]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,0,1]
+; SSE-NEXT: movdqa %xmm2, %xmm7
+; SSE-NEXT: pandn %xmm6, %xmm7
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,0,1]
+; SSE-NEXT: pand %xmm2, %xmm6
+; SSE-NEXT: por %xmm7, %xmm6
+; SSE-NEXT: movdqa %xmm14, %xmm7
+; SSE-NEXT: pandn %xmm6, %xmm7
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[2,2,3,3]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm6[1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[2,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,3,2,3,4,5,6,7]
-; SSE-NEXT: movss {{.*#+}} xmm0 = xmm7[0],xmm0[1,2,3]
-; SSE-NEXT: andps %xmm15, %xmm0
-; SSE-NEXT: orps %xmm3, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm1[2,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm8[0,3,2,3,4,5,6,7]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm8[0],xmm0[1,2,3]
+; SSE-NEXT: andps %xmm14, %xmm0
+; SSE-NEXT: orps %xmm7, %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,0,1]
-; SSE-NEXT: movdqa %xmm2, %xmm7
-; SSE-NEXT: pandn %xmm3, %xmm7
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,1]
-; SSE-NEXT: pand %xmm2, %xmm3
-; SSE-NEXT: por %xmm7, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm8
-; SSE-NEXT: pandn %xmm3, %xmm8
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[2,2,3,3]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; SSE-NEXT: pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[0,1,0,1]
+; SSE-NEXT: movdqa %xmm2, %xmm8
+; SSE-NEXT: pandn %xmm7, %xmm8
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: # xmm7 = xmm7[0],mem[0],xmm7[1],mem[1],xmm7[2],mem[2],xmm7[3],mem[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[2,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT: movss {{.*#+}} xmm0 = xmm3[0],xmm0[1,2,3]
-; SSE-NEXT: andps %xmm15, %xmm0
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,0,1]
+; SSE-NEXT: pand %xmm2, %xmm7
+; SSE-NEXT: por %xmm8, %xmm7
+; SSE-NEXT: movdqa %xmm14, %xmm8
+; SSE-NEXT: pandn %xmm7, %xmm8
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,4,5,4,7]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[2,2,3,3]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm7[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm1[2,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm9[0,3,2,3,4,5,6,7]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm9[0],xmm0[1,2,3]
+; SSE-NEXT: andps %xmm14, %xmm0
; SSE-NEXT: orps %xmm8, %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,0,1]
-; SSE-NEXT: movdqa %xmm2, %xmm8
-; SSE-NEXT: pandn %xmm3, %xmm8
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,1]
-; SSE-NEXT: pand %xmm2, %xmm3
-; SSE-NEXT: por %xmm8, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm8
-; SSE-NEXT: pandn %xmm3, %xmm8
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[2,2,3,3]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1],xmm3[2],mem[2],xmm3[3],mem[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm3[2,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm9[0,3,2,3,4,5,6,7]
-; SSE-NEXT: movss {{.*#+}} xmm1 = xmm9[0],xmm1[1,2,3]
-; SSE-NEXT: andps %xmm15, %xmm1
-; SSE-NEXT: orps %xmm8, %xmm1
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT: # xmm8 = mem[0,1,0,1]
; SSE-NEXT: movdqa %xmm2, %xmm9
; SSE-NEXT: pandn %xmm8, %xmm9
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm13, %xmm8
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,1,0,1]
; SSE-NEXT: pand %xmm2, %xmm8
; SSE-NEXT: por %xmm9, %xmm8
-; SSE-NEXT: movdqa %xmm15, %xmm9
-; SSE-NEXT: pandn %xmm8, %xmm9
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm13[0,1,0,3]
+; SSE-NEXT: movdqa %xmm14, %xmm11
+; SSE-NEXT: pandn %xmm8, %xmm11
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm6[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,4,5,4,7]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = mem[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm5[2,2,3,3]
; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm8[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = xmm8[0],mem[0],xmm8[1],mem[1],xmm8[2],mem[2],xmm8[3],mem[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm8[2,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm11 = xmm11[0,3,2,3,4,5,6,7]
-; SSE-NEXT: movss {{.*#+}} xmm0 = xmm11[0],xmm0[1,2,3]
-; SSE-NEXT: andps %xmm15, %xmm0
-; SSE-NEXT: orps %xmm9, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT: # xmm9 = xmm9[0],mem[0],xmm9[1],mem[1],xmm9[2],mem[2],xmm9[3],mem[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm9[2,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm8[0,3,2,3,4,5,6,7]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm8[0],xmm0[1,2,3]
+; SSE-NEXT: andps %xmm14, %xmm0
+; SSE-NEXT: orps %xmm11, %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm9
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm12[2],xmm9[3],xmm12[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm9[0,1,0,1]
-; SSE-NEXT: pand %xmm2, %xmm9
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm14[0,1,0,1]
+; SSE-NEXT: movdqa %xmm12, %xmm8
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,1,0,1]
+; SSE-NEXT: pand %xmm2, %xmm8
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm10[0,1,0,1]
; SSE-NEXT: pandn %xmm11, %xmm2
-; SSE-NEXT: por %xmm9, %xmm2
-; SSE-NEXT: movdqa %xmm15, %xmm9
-; SSE-NEXT: pandn %xmm2, %xmm9
+; SSE-NEXT: por %xmm8, %xmm2
+; SSE-NEXT: movdqa %xmm14, %xmm8
+; SSE-NEXT: pandn %xmm2, %xmm8
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,5,4,7]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,2,3,3]
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm2[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,2,3,3]
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm2[2,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm11 = xmm11[0,3,2,3,4,5,6,7]
-; SSE-NEXT: movss {{.*#+}} xmm4 = xmm11[0],xmm4[1,2,3]
-; SSE-NEXT: andps %xmm15, %xmm4
-; SSE-NEXT: orps %xmm9, %xmm4
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{.*#+}} xmm4 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm4, %xmm9
-; SSE-NEXT: pandn %xmm1, %xmm9
-; SSE-NEXT: pand %xmm4, %xmm0
-; SSE-NEXT: por %xmm9, %xmm0
-; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm0[0,1,1,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,7,7,7,7]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm11[0],xmm0[1,2,3]
+; SSE-NEXT: andps %xmm14, %xmm0
+; SSE-NEXT: orps %xmm8, %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm15, %xmm8
+; SSE-NEXT: pandn %xmm3, %xmm8
+; SSE-NEXT: pand %xmm15, %xmm1
+; SSE-NEXT: por %xmm8, %xmm1
+; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm1[0,1,1,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,7,7,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
-; SSE-NEXT: movss {{.*#+}} xmm9 = xmm2[0],xmm9[1,2,3]
-; SSE-NEXT: movdqa %xmm10, %xmm2
-; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm12[4],xmm2[5],xmm12[5],xmm2[6],xmm12[6],xmm2[7],xmm12[7]
+; SSE-NEXT: movss {{.*#+}} xmm8 = xmm2[0],xmm8[1,2,3]
+; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,7]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm14[4],xmm2[5],xmm14[5],xmm2[6],xmm14[6],xmm2[7],xmm14[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,4,7]
-; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: movdqa %xmm14, %xmm0
; SSE-NEXT: pandn %xmm2, %xmm0
-; SSE-NEXT: andps %xmm15, %xmm9
-; SSE-NEXT: por %xmm9, %xmm0
+; SSE-NEXT: andps %xmm14, %xmm8
+; SSE-NEXT: por %xmm8, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm4, %xmm2
+; SSE-NEXT: movdqa %xmm15, %xmm2
+; SSE-NEXT: pandn %xmm6, %xmm2
+; SSE-NEXT: pand %xmm15, %xmm5
+; SSE-NEXT: por %xmm2, %xmm5
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm5[0,1,1,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm9[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[2,2,2,2]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm8[0],xmm2[1,2,3]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm7[4],xmm13[5],xmm7[5],xmm13[6],xmm7[6],xmm13[7],xmm7[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm13[0,1,2,1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,7,7]
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = xmm8[4],mem[4],xmm8[5],mem[5],xmm8[6],mem[6],xmm8[7],mem[7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm8[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,6,4,7]
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm8, %xmm0
+; SSE-NEXT: andps %xmm14, %xmm2
+; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm15, %xmm2
; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm12, %xmm8
+; SSE-NEXT: pand %xmm15, %xmm8
+; SSE-NEXT: por %xmm2, %xmm8
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm8[0,1,1,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; SSE-NEXT: pshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[2,2,2,2]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm6[0],xmm2[1,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,7,7]
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,4,7]
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm6, %xmm0
+; SSE-NEXT: andps %xmm14, %xmm2
+; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm15, %xmm2
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pand %xmm15, %xmm6
+; SSE-NEXT: por %xmm2, %xmm6
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm6[0,1,1,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; SSE-NEXT: pshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[2,2,2,2]
+; SSE-NEXT: movss {{.*#+}} xmm2 = xmm6[0],xmm2[1,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm11, %xmm6
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: pand %xmm4, %xmm9
-; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: por %xmm2, %xmm9
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm9[0,1,1,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT: pshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[2,2,2,2]
-; SSE-NEXT: movss {{.*#+}} xmm2 = xmm4[0],xmm2[1,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,7,7]
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,6,4,7]
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm4, %xmm0
-; SSE-NEXT: andps %xmm15, %xmm2
+; SSE-NEXT: punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,7,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm8[4],xmm6[5],xmm8[5],xmm6[6],xmm8[6],xmm6[7],xmm8[7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,4,7]
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: pandn %xmm6, %xmm0
+; SSE-NEXT: andps %xmm14, %xmm2
; SSE-NEXT: por %xmm2, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm1, %xmm2
+; SSE-NEXT: movdqa %xmm15, %xmm2
; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: pand %xmm1, %xmm4
-; SSE-NEXT: movdqa %xmm1, %xmm12
-; SSE-NEXT: por %xmm2, %xmm4
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm4[0,1,1,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm6[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[2,2,2,2]
-; SSE-NEXT: movss {{.*#+}} xmm2 = xmm4[0],xmm2[1,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,7,7]
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,6,4,7]
-; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: pandn %xmm4, %xmm0
-; SSE-NEXT: andps %xmm15, %xmm2
-; SSE-NEXT: por %xmm2, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm1, %xmm2
-; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: pand %xmm1, %xmm4
-; SSE-NEXT: por %xmm2, %xmm4
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm4[0,1,1,0,4,5,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pand %xmm15, %xmm6
+; SSE-NEXT: por %xmm2, %xmm6
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm6[0,1,1,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
; SSE-NEXT: pshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
; SSE-NEXT: movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm11, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm6, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm14[4],xmm0[5],xmm14[5],xmm0[6],xmm14[6],xmm0[7],xmm14[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: andps %xmm15, %xmm2
+; SSE-NEXT: andps %xmm14, %xmm2
; SSE-NEXT: por %xmm2, %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm12, %xmm1
-; SSE-NEXT: movdqa %xmm12, %xmm0
-; SSE-NEXT: pandn %xmm5, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm12, %xmm2
-; SSE-NEXT: pand %xmm1, %xmm2
-; SSE-NEXT: movdqa %xmm1, %xmm4
+; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: pand %xmm15, %xmm2
; SSE-NEXT: por %xmm0, %xmm2
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,1,1,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm7[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
; SSE-NEXT: movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm7, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm4, %xmm2
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,4,7]
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: andps %xmm15, %xmm0
+; SSE-NEXT: andps %xmm14, %xmm0
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm4, %xmm0
-; SSE-NEXT: pandn (%rsp), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: pand %xmm4, %xmm2
+; SSE-NEXT: pand %xmm15, %xmm2
; SSE-NEXT: por %xmm0, %xmm2
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,1,1,0,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
@@ -8243,76 +8280,52 @@
; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: punpckhwd (%rsp), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
-; SSE-NEXT: movdqa %xmm15, %xmm2
+; SSE-NEXT: movdqa %xmm14, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: andps %xmm15, %xmm0
+; SSE-NEXT: andps %xmm14, %xmm0
; SSE-NEXT: por %xmm0, %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm4, %xmm0
-; SSE-NEXT: pandn %xmm13, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm4, %xmm1
-; SSE-NEXT: por %xmm0, %xmm1
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,1,1,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm8[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
-; SSE-NEXT: movdqa %xmm15, %xmm2
-; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: andps %xmm15, %xmm0
-; SSE-NEXT: por %xmm0, %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm4, %xmm2
-; SSE-NEXT: movdqa %xmm4, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: pandn %xmm4, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
-; SSE-NEXT: por %xmm0, %xmm1
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,1,1,0,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: andps %xmm15, %xmm0
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
-; SSE-NEXT: pandn %xmm1, %xmm15
-; SSE-NEXT: por %xmm0, %xmm15
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; SSE-NEXT: movdqa %xmm15, %xmm0
-; SSE-NEXT: psrld $16, %xmm0
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT: pand %xmm15, %xmm1
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,1,1,0,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
+; SSE-NEXT: pshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: andps %xmm14, %xmm0
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
+; SSE-NEXT: pandn %xmm1, %xmm14
+; SSE-NEXT: por %xmm0, %xmm14
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: psrld $16, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm1[4],xmm15[5],xmm1[5],xmm15[6],xmm1[6],xmm15[7],xmm1[7]
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE-NEXT: movdqa %xmm1, %xmm13
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm2[4],xmm13[5],xmm2[5],xmm13[6],xmm2[6],xmm13[7],xmm2[7]
+; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -8321,7 +8334,39 @@
; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[0,1,0,3]
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: psrld $16, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE-NEXT: movdqa %xmm1, %xmm10
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm2[4],xmm10[5],xmm2[5],xmm10[6],xmm2[6],xmm10[7],xmm2[7]
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $48, %xmm1
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: psrlq $16, %xmm1
+; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
@@ -8344,11 +8389,9 @@
; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: psrlq $16, %xmm1
-; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -8364,49 +8407,15 @@
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm14, %xmm0
; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT: movdqa %xmm1, %xmm8
-; SSE-NEXT: movdqa %xmm3, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm3[4],xmm8[5],xmm3[5],xmm8[6],xmm3[6],xmm8[7],xmm3[7]
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: psrlq $16, %xmm1
-; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm1, %xmm0
-; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT: movdqa %xmm1, %xmm8
-; SSE-NEXT: movdqa %xmm3, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm3[4],xmm8[5],xmm3[5],xmm8[6],xmm3[6],xmm8[7],xmm3[7]
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,3,2,3]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm1[4],xmm14[5],xmm1[5],xmm14[6],xmm1[6],xmm14[7],xmm1[7]
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -8418,7 +8427,7 @@
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm9[2],xmm1[3],xmm9[3]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm14[4],xmm1[5],xmm14[5],xmm1[6],xmm14[6],xmm1[7],xmm14[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
@@ -8426,27 +8435,28 @@
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm1, %xmm15
-; SSE-NEXT: psrld $16, %xmm15
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm3[4],xmm15[5],xmm3[5],xmm15[6],xmm3[6],xmm15[7],xmm3[7]
+; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: psrld $16, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE-NEXT: movdqa %xmm1, %xmm8
-; SSE-NEXT: movdqa %xmm3, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm3[4],xmm8[5],xmm3[5],xmm8[6],xmm3[6],xmm8[7],xmm3[7]
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm2[4],xmm8[5],xmm2[5],xmm8[6],xmm2[6],xmm8[7],xmm2[7]
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: psrlq $16, %xmm1
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3]
+; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,1,0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
@@ -8456,28 +8466,26 @@
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movdqa %xmm1, %xmm14
; SSE-NEXT: psrld $16, %xmm14
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm3[4],xmm14[5],xmm3[5],xmm14[6],xmm3[6],xmm14[7],xmm3[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm2[4],xmm14[5],xmm2[5],xmm14[6],xmm2[6],xmm14[7],xmm2[7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT: movdqa %xmm1, %xmm6
-; SSE-NEXT: movdqa %xmm3, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm3[4],xmm6[5],xmm3[5],xmm6[6],xmm3[6],xmm6[7],xmm3[7]
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm1, %xmm5
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: psrlq $16, %xmm1
; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[0,1,0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
@@ -8487,27 +8495,29 @@
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movdqa %xmm1, %xmm11
; SSE-NEXT: psrld $16, %xmm11
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm3[4],xmm11[5],xmm3[5],xmm11[6],xmm3[6],xmm11[7],xmm3[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm2[4],xmm11[5],xmm2[5],xmm11[6],xmm2[6],xmm11[7],xmm2[7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT: movdqa %xmm1, %xmm6
-; SSE-NEXT: movdqa %xmm3, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm3[4],xmm6[5],xmm3[5],xmm6[6],xmm3[6],xmm6[7],xmm3[7]
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm1, %xmm3
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm12, %xmm1
; SSE-NEXT: psrlq $16, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[0,1,0,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm12[2],xmm1[3],xmm12[3]
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm13[2],xmm1[3],xmm13[3]
+; SSE-NEXT: punpckhwd (%rsp), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
@@ -8517,19 +8527,21 @@
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movdqa %xmm1, %xmm10
; SSE-NEXT: psrld $16, %xmm10
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm3[4],xmm10[5],xmm3[5],xmm10[6],xmm3[6],xmm10[7],xmm3[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm2[4],xmm10[5],xmm2[5],xmm10[6],xmm2[6],xmm10[7],xmm2[7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT: movdqa %xmm1, %xmm5
-; SSE-NEXT: movdqa %xmm3, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm1, %xmm3
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: psrlq $16, %xmm4
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; SSE-NEXT: movdqa %xmm5, %xmm1
+; SSE-NEXT: psrlq $16, %xmm1
+; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,1,0,3]
@@ -8544,17 +8556,39 @@
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm3, %xmm0
+; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: pand %xmm3, %xmm1
+; SSE-NEXT: movdqa %xmm3, %xmm15
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,1,0,3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm15, %xmm2
+; SSE-NEXT: movdqa %xmm15, %xmm0
; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: pand %xmm15, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
@@ -8570,10 +8604,10 @@
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movdqa %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm15, %xmm0
; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: pand %xmm15, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
@@ -8589,10 +8623,10 @@
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movdqa %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm15, %xmm0
; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: pand %xmm15, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
@@ -8606,28 +8640,11 @@
; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT: movdqa %xmm2, %xmm0
-; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
-; SSE-NEXT: por %xmm0, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,1,0,3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[0,2]
-; SSE-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
-; SSE-NEXT: movdqa %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm15, %xmm0
; SSE-NEXT: pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: pand %xmm15, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
@@ -8635,29 +8652,28 @@
; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,1],xmm0[0,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm9[0],xmm12[1],xmm9[1],xmm12[2],xmm9[2],xmm12[3],xmm9[3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm12[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
-; SSE-NEXT: movdqa %xmm2, %xmm0
-; SSE-NEXT: pandn %xmm13, %xmm0
-; SSE-NEXT: pand %xmm2, %xmm12
-; SSE-NEXT: por %xmm0, %xmm12
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[0,1,0,3]
+; SSE-NEXT: movdqa %xmm15, %xmm0
+; SSE-NEXT: pandn %xmm4, %xmm0
+; SSE-NEXT: pand %xmm15, %xmm13
+; SSE-NEXT: por %xmm0, %xmm13
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE-NEXT: punpckhwd (%rsp), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,1],xmm0[0,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm12, %xmm0
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm13[0],xmm5[1],xmm13[1],xmm5[2],xmm13[2],xmm5[3],xmm13[3]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; SSE-NEXT: movdqa %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm15, %xmm0
; SSE-NEXT: pandn %xmm7, %xmm0
+; SSE-NEXT: movdqa %xmm7, %xmm15
; SSE-NEXT: pand %xmm2, %xmm6
; SSE-NEXT: por %xmm0, %xmm6
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,1,0,3]
@@ -8665,28 +8681,28 @@
; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[0,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm6, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm4, %xmm0
; SSE-NEXT: pand %xmm2, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: pandn %xmm8, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pandn %xmm3, %xmm2
; SSE-NEXT: por %xmm0, %xmm2
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movdqa %xmm1, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[0,2]
-; SSE-NEXT: movaps %xmm3, %xmm13
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
+; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[0,2]
+; SSE-NEXT: movaps %xmm6, %xmm12
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
@@ -8694,10 +8710,10 @@
; SSE-NEXT: # xmm1 = mem[2,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm6[1,1,1,1,4,5,6,7]
-; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm4[1,1,1,1,4,5,6,7]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,1,0,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm8[2,2,2,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm3[2,2,2,2]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm0[2],xmm9[3],xmm0[3]
; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
@@ -8783,7 +8799,7 @@
; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[1,1,1,1]
@@ -8815,14 +8831,14 @@
; SSE-NEXT: # xmm0 = mem[1,1,1,1,4,5,6,7]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
+; SSE-NEXT: pshufd $196, (%rsp), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[0,1,0,3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,1,1,1]
-; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm13[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
@@ -8832,7 +8848,7 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; SSE-NEXT: pshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[1,1,1,1,4,5,6,7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm15, %xmm0
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
; SSE-NEXT: pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = mem[0,1,0,3]
@@ -8841,53 +8857,53 @@
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 96(%rsi)
+; SSE-NEXT: movaps %xmm3, 80(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 112(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 48(%rsi)
+; SSE-NEXT: movaps %xmm3, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 64(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 80(%rsi)
+; SSE-NEXT: movaps %xmm3, 112(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 16(%rsi)
+; SSE-NEXT: movaps %xmm3, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 96(%rdx)
+; SSE-NEXT: movaps %xmm3, 96(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 32(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 112(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, (%rdx)
+; SSE-NEXT: movaps %xmm3, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 80(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 96(%rcx)
+; SSE-NEXT: movaps %xmm3, 64(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 112(%rcx)
+; SSE-NEXT: movaps %xmm3, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 64(%rcx)
+; SSE-NEXT: movaps %xmm3, 112(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 80(%rcx)
+; SSE-NEXT: movaps %xmm3, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 32(%rcx)
+; SSE-NEXT: movaps %xmm3, 96(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 48(%rcx)
+; SSE-NEXT: movaps %xmm3, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, (%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 16(%rcx)
+; SSE-NEXT: movaps %xmm3, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 80(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 64(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 112(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 96(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 112(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
@@ -8924,14 +8940,15 @@
; SSE-NEXT: movaps %xmm10, 112(%rax)
; SSE-NEXT: movaps %xmm11, 96(%rax)
; SSE-NEXT: movaps %xmm14, 80(%rax)
-; SSE-NEXT: movaps %xmm15, 64(%rax)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 64(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 48(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 32(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 16(%rax)
-; SSE-NEXT: movaps %xmm13, (%rax)
+; SSE-NEXT: movaps %xmm12, (%rax)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT: movapd %xmm0, 112(%rax)
; SSE-NEXT: movapd %xmm1, 96(%rax)
@@ -8941,12 +8958,12 @@
; SSE-NEXT: movapd %xmm7, 32(%rax)
; SSE-NEXT: movapd %xmm8, 16(%rax)
; SSE-NEXT: movapd %xmm9, (%rax)
-; SSE-NEXT: addq $1352, %rsp # imm = 0x548
+; SSE-NEXT: addq $1368, %rsp # imm = 0x558
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i16_stride7_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1544, %rsp # imm = 0x608
+; AVX1-ONLY-NEXT: subq $1512, %rsp # imm = 0x5E8
; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpsrld $16, %xmm0, %xmm0
@@ -8961,26 +8978,26 @@
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5],xmm1[6,7]
-; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm14
; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6],xmm2[7]
-; AVX1-ONLY-NEXT: vmovdqa 112(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm15[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 112(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm10[0,3,2,3]
+; AVX1-ONLY-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm3
@@ -8991,184 +9008,183 @@
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm6[2],xmm7[2],zero
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm3[2],xmm7[2],zero
; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0,1,2],xmm3[3,4],xmm0[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm0 = [65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm5 = [65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm5, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm3, %ymm3
; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 624(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa 608(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa 624(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm0, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa 608(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX1-ONLY-NEXT: vmovdqa 576(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa 592(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; AVX1-ONLY-NEXT: vmovdqa 576(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm0, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 592(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 656(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 656(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 640(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 640(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vmovdqa 448(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; AVX1-ONLY-NEXT: vmovdqa 448(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 464(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm5[2,2,3,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 464(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 496(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm4[2],xmm3[2],zero
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm0[2],xmm3[2],zero
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 528(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 528(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 512(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovdqa 544(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,0,0,0]
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovdqa 544(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,0,0,0]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm4[7]
-; AVX1-ONLY-NEXT: vmovdqa 560(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
+; AVX1-ONLY-NEXT: vmovdqa 560(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm5, %ymm3
; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 400(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa 384(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa 400(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm0, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa 384(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm14[4],xmm2[5],xmm14[5],xmm2[6],xmm14[6],xmm2[7],xmm14[7]
-; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm0, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 432(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 432(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 416(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 240(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm12[2,2,3,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 240(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm3[2],xmm4[2],zero
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm6[2],xmm11[2],zero
+; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 304(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; AVX1-ONLY-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,0,0,0]
+; AVX1-ONLY-NEXT: vmovdqa 304(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,0,0,0]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm4[7]
-; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm11[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm5, %ymm3
; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 848(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa 832(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa 848(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm15, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 832(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm12[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; AVX1-ONLY-NEXT: vmovdqa 800(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa 816(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; AVX1-ONLY-NEXT: vmovdqa 800(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm0, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 816(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 880(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 880(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 864(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vmovdqa 672(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; AVX1-ONLY-NEXT: vmovdqa 672(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 688(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa 688(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm13[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 720(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm4[2],xmm3[2],zero
-; AVX1-ONLY-NEXT: vmovaps %xmm4, %xmm10
-; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovaps 720(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm9[2],xmm8[2],zero
+; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa 752(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 752(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 736(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovdqa 768(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,0,0,0]
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovdqa 768(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,0,0,0]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm4[7]
-; AVX1-ONLY-NEXT: vmovdqa 784(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
+; AVX1-ONLY-NEXT: vmovdqa 784(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm5, %ymm0
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -9189,33 +9205,34 @@
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm14, %xmm1
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT: vpslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0,1,2,3,4,5]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6],xmm2[7]
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm15, %xmm2
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm10, %xmm2
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7]
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm13 = [8,9,8,9,8,9,8,9,6,7,6,7,6,7,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm6[1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm3 = [8,9,8,9,8,9,8,9,6,7,6,7,6,7,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm7[1],mem[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,0,3,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm1[3,4],xmm3[5,6,7]
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm14 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm14, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm14, %ymm3
; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm7 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm7, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm10 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm10, %ymm3
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm10, %ymm2
; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -9237,9 +9254,55 @@
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm5[0],mem[1],xmm5[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm0[0],mem[1],xmm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,0,3,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm0, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm5 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6],xmm5[7]
+; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm0, %xmm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm14, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm14, %ymm4
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm10, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2,3,4,5],xmm0[6],mem[7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,0,3,2,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpslld $16, %xmm0, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm3[6,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm6[4],xmm11[4],xmm6[5],xmm11[5],xmm6[6],xmm11[6],xmm6[7],xmm11[7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm6
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm0[0],mem[1],xmm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,0,3,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
@@ -9254,76 +9317,32 @@
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrld $16, %xmm0, %xmm5
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm1, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm14, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm14, %ymm4
; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm3
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm7, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm10, %ymm3
; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm12[4],xmm15[4],xmm12[5],xmm15[5],xmm12[6],xmm15[6],xmm12[7],xmm15[7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
-; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2,3,4,5],xmm14[6],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm11[0,1,2,3,4,5],xmm15[6],xmm11[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,0,3,2,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpslld $16, %xmm0, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpslld $16, %xmm1, %xmm3
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm12[0],mem[1],xmm12[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,0,3,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm8, %xmm4
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm5 = xmm9[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6],xmm5[7]
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm11, %xmm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm1, %ymm4
-; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm7, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm9[4],xmm8[4],xmm9[5],xmm8[5],xmm9[6],xmm8[6],xmm9[7],xmm8[7]
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm3, %xmm3
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm12[0,1,2,3,4,5],xmm8[6],xmm12[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,0,3,2,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpslld $16, %xmm0, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm10[4],xmm15[4],xmm10[5],xmm15[5],xmm10[6],xmm15[6],xmm10[7],xmm15[7]
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm14[0],xmm11[1],xmm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm13[0],xmm12[1],xmm13[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,0,3,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
@@ -9332,19 +9351,18 @@
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm8[0,1,2,3,4,5]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6],xmm5[7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm10, %xmm5
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm9, %xmm5
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm0 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm14, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm14, %ymm4
; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm3
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm7, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm10, %ymm3
; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
@@ -9378,17 +9396,17 @@
; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = mem[2,2,3,3]
; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm5 = xmm3[1],xmm5[1]
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm6[0],mem[0],xmm6[1],mem[1],xmm6[2],mem[2],xmm6[3],mem[3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm7[0],mem[0],xmm7[1],mem[1],xmm7[2],mem[2],xmm7[3],mem[3]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm3[2,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm14, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm14, %ymm5
; AVX1-ONLY-NEXT: vorps %ymm4, %ymm5, %ymm4
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm7, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm10, %ymm4
; AVX1-ONLY-NEXT: vorps %ymm2, %ymm4, %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
@@ -9398,8 +9416,8 @@
; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = mem[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw $243, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1],xmm4[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3],xmm6[4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2,3],xmm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
@@ -9415,275 +9433,182 @@
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1],xmm5[2],mem[2],xmm5[3],mem[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm5[2,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,3,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[2,2,2,2]
-; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3,4,5],xmm6[6,7]
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[2,2,2,2]
+; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2,3,4,5],xmm5[6,7]
; AVX1-ONLY-NEXT: vpshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,6],xmm7[7]
-; AVX1-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6],xmm7[7]
+; AVX1-ONLY-NEXT: vpermilps $85, (%rsp), %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vorps %ymm6, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm14, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm14, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm4, %ymm4
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm0 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm10, %ymm4
; AVX1-ONLY-NEXT: vorps %ymm2, %ymm4, %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpsllq $16, %xmm9, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsllq $16, %xmm2, %xmm2
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm8[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm15[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm12[0,1],xmm4[2,3],xmm12[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm11[0,1],xmm4[2,3],xmm11[4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1],xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm4[6,7]
; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm15[2,2,3,3]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm6[1]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm11[0],xmm14[0],xmm11[1],xmm14[1],xmm11[2],xmm14[2],xmm11[3],xmm14[3]
+; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[2,2,3,3]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm5[1]
+; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = xmm12[0],mem[0],xmm12[1],mem[1],xmm12[2],mem[2],xmm12[3],mem[3]
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm0[2,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,3,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm13[2,2,2,2]
-; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3,4,5],xmm6[6,7]
-; AVX1-ONLY-NEXT: vpshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,6],xmm7[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm10[1,1,1,1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm15 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandps %ymm4, %ymm15, %ymm4
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm15, %ymm6
-; AVX1-ONLY-NEXT: vmovaps %ymm15, %ymm10
-; AVX1-ONLY-NEXT: vorps %ymm6, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm0[2,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm13[2,2,2,2]
+; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2,3,4,5],xmm5[6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm8[0,1,0,1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6],xmm7[7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm9[1,1,1,1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm14, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm14, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm4, %ymm4
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm13 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm13, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm4, %ymm13, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm10, %ymm4
; AVX1-ONLY-NEXT: vorps %ymm2, %ymm4, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsllq $16, %xmm12, %xmm2
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; AVX1-ONLY-NEXT: vpsllq $16, %xmm0, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm8[0,3,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm1[0,1],xmm4[2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2,3],xmm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1],xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm0[0,1],xmm4[2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm11[0],xmm4[1],xmm11[1],xmm4[2],xmm11[2],xmm4[3],xmm11[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm2[0,1,2,3,4,5],xmm4[6,7]
-; AVX1-ONLY-NEXT: vpshufd $196, (%rsp), %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[2,2,3,3]
-; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm6 = xmm2[1],xmm6[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm2[2,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,3,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm9[2,2,2,2]
-; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,2,3,4,5],xmm7[6,7]
-; AVX1-ONLY-NEXT: vpshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6],xmm15[7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm11[1,1,1,1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm7, %ymm7
-; AVX1-ONLY-NEXT: vandps %ymm6, %ymm10, %ymm6
-; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm10, %ymm7
-; AVX1-ONLY-NEXT: vorps %ymm7, %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm13, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm6, %ymm13, %ymm6
-; AVX1-ONLY-NEXT: vmovaps %ymm13, %ymm7
-; AVX1-ONLY-NEXT: vorps %ymm4, %ymm6, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,0,3,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3,4,5],xmm0[6],mem[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,0,0,0,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0],xmm4[1,2],xmm6[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm0[2,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm6[6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm0[0],mem[1],xmm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,1,0,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,7,7,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,7,7]
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm6 = [4,5,2,3,4,5,6,7,8,9,4,5,8,9,2,3]
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm15, %xmm15
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm15, %ymm13
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm10, %ymm3
-; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm10, %ymm13
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm13, %ymm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm7, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,0,3,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $64, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm0[0,1,2,3,4,5],mem[6],xmm0[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,6,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm3[1,2],xmm4[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm3[2,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,0,0,0]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm13[6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[2,2,2,2]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm10[0],mem[1],xmm10[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[0,1,1,0,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,7,7,7,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1],xmm13[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,1]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,4,7,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm13 = xmm13[4],xmm7[4],xmm13[5],xmm7[5],xmm13[6],xmm7[6],xmm13[7],xmm7[7]
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm14, %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm13, %ymm13
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm15 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm15, %ymm5
-; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm15, %ymm13
-; AVX1-ONLY-NEXT: vorps %ymm5, %ymm13, %ymm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm13 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm13, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm13, %ymm5
-; AVX1-ONLY-NEXT: vorps %ymm4, %ymm5, %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,0,3,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2,3,4,5],xmm12[6],mem[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,0,0,0,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0],xmm4[1,2],xmm5[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm4[2,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,0,0,0]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5],xmm13[6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0],xmm1[1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[0,1,1,0,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,7,7,7,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm13[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm13 = xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,1]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,4,7,7]
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm13[4],mem[4],xmm13[5],mem[5],xmm13[6],mem[6],xmm13[7],mem[7]
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm11, %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm13, %ymm13
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,0,0,0,0,65535,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm1, %ymm13
-; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm9
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm13, %ymm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm12 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm12, %ymm5
-; AVX1-ONLY-NEXT: vandps %ymm2, %ymm12, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm5, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[2,2,3,3]
+; AVX1-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm5 = xmm2[1],xmm5[1]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,3,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm2[2,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm7 = xmm7[0,3,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm9[2,2,2,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm12[0,1,2,3,4,5],xmm7[6,7]
+; AVX1-ONLY-NEXT: vpshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6],xmm15[7]
+; AVX1-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm7, %ymm7
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm14, %ymm5
+; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm14, %ymm7
+; AVX1-ONLY-NEXT: vorps %ymm7, %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm10, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm10, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm5, %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2,3,4,5],xmm1[6],mem[7]
+; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,0,3,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $64, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm1[0,1,2,3,4,5],mem[6],xmm1[7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,0,0,0,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm5[0],xmm2[1,2],xmm5[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1,2],xmm5[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm1[2,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,0,0,0]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm13[6,7]
-; AVX1-ONLY-NEXT: vpshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,2,3,6,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[2,2,2,2]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm1[2,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,0,0,0]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm4[0,1,2,3,4,5],xmm5[6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm7[0],mem[1],xmm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,1,0,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,7,7,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm1[0],mem[1],xmm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,7,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm4 = [4,5,2,3,4,5,6,7,8,9,4,5,8,9,2,3]
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm15, %xmm15
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm13
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm15, %ymm13
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm14, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm14, %ymm13
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm13, %ymm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm10, %ymm5
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm10, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm3, %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm6[0],mem[0],xmm6[1],mem[1],xmm6[2],mem[2],xmm6[3],mem[3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,0,3,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $64, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm1[0,1,2,3,4,5],mem[6],xmm1[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,0,0,0,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm5[0],xmm3[1,2],xmm5[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm5[2,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,0,0,0]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm13[6,7]
+; AVX1-ONLY-NEXT: vpshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3,6,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[2,2,2,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[0],xmm1[1],mem[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[0,1,1,0,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,7,7,7,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1],xmm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1],xmm13[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm13 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm13 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
@@ -9691,158 +9616,243 @@
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm13 = xmm13[4],mem[4],xmm13[5],mem[5],xmm13[6],mem[6],xmm13[7],mem[7]
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm13, %xmm6
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vandps %ymm9, %ymm8, %ymm8
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm9, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm8, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm12, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm12, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm15
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm13, %ymm13
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm14, %ymm6
+; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm14, %ymm13
+; AVX1-ONLY-NEXT: vorps %ymm6, %ymm13, %ymm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm10, %ymm3
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm10, %ymm6
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm6, %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,0,3,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3,4,5],xmm0[6],mem[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,0,0,0,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0],xmm3[1,2],xmm6[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm3[2,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,0,0,0]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5],xmm13[6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[0],xmm0[1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[0,1,1,0,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,7,7,7,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm13 = xmm12[4],xmm9[4],xmm12[5],xmm9[5],xmm12[6],xmm9[6],xmm12[7],xmm9[7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,1]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,4,7,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[4],mem[4],xmm13[5],mem[5],xmm13[6],mem[6],xmm13[7],mem[7]
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm15
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm13, %ymm13
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm14, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm14, %ymm13
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm13, %ymm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm10, %ymm6
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm6, %ymm2, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,0,3,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3,4,5],xmm0[6],mem[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,0,0,0,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0],xmm2[1,2],xmm6[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm2[2,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,0,0,0]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5],xmm13[6,7]
+; AVX1-ONLY-NEXT: vpshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,2,3,6,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[2,2,2,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[0],xmm0[1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm13 = xmm13[0,1,1,0,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,7,7,7,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm8 = xmm8[0,1],xmm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm13 = xmm13[0,1,2,1]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,4,7,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[4],mem[4],xmm13[5],mem[5],xmm13[6],mem[6],xmm13[7],mem[7]
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm13, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm13
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm14, %ymm8, %ymm8
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm14, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm8, %ymm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm10, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm10, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[1,1,1,1]
; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[0,1],mem[2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1,2],xmm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm3[0,1,2,1]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0,1],xmm3[2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[0,1,0,3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm15[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm4[0],xmm1[1,2],xmm4[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $100, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,2,1]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,5],xmm4[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm1
+; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm0, %xmm6
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm1[0,1],xmm6[2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm1 = xmm13[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm1 = [12,13,14,15,4,5,6,7,0,1,4,5,8,9,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm8 = xmm14[2,3,2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm12, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm3, %ymm12, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm0
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm7, %xmm7
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm7, %ymm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm10, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm10, %ymm6
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm6, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm10[2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm13[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1,2],xmm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $100, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2,1]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm3
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm4[0,1],mem[2,3],xmm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm12[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm11[0],xmm6[1],xmm11[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0],xmm4[1,2],xmm6[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm5[6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm5[2,3,2,3]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm5, %xmm5
+; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrlq $16, %xmm8, %xmm6
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm6[2,3],xmm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm11[0,1,0,3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3],xmm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm6 = xmm6[2],xmm15[2],xmm6[3],xmm15[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
+; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm14[2,3,2,3]
+; AVX1-ONLY-NEXT: vpermilps $238, (%rsp), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm7 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm7, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1],mem[2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1,2],xmm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm4[0,1,2,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm10, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm10, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm5, %ymm4
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm4[0,1],mem[2,3],xmm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm14[0],xmm5[1],xmm14[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1,2],xmm5[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1],xmm4[2],mem[2],xmm4[3],mem[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm4, %xmm4
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,5],xmm3[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm4, %xmm4
+; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm5, %xmm5
+; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1],xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3],xmm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm5, %xmm5
; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm7, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1],mem[2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,3,2,3]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1,2],xmm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $100, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2,1]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm3[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm10, %ymm3
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm10, %ymm4
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm4, %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpblendw $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[0,1],mem[2,3],xmm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm6[0,3,2,3]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1,2],xmm4[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,5],xmm2[6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrlq $48, %xmm3, %xmm3
; AVX1-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $16, %xmm4, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $16, %xmm7, %xmm4
; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1],xmm4[2],mem[2],xmm4[3],mem[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6,7]
@@ -9859,187 +9869,181 @@
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm7, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm10, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm2, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm10, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm1
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm9[0],xmm13[0],xmm9[1],xmm13[1],xmm9[2],xmm13[2],xmm9[3],xmm13[3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm9[0],xmm15[0],xmm9[1],xmm15[1],xmm9[2],xmm15[2],xmm9[3],xmm15[3]
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm9 = [8,9,8,9,8,9,8,9,6,7,6,7,6,7,6,7]
; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm2, %xmm2
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3,4],xmm1[5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $64, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1,2,3,4,5],mem[6],xmm2[7]
+; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3,4,5],xmm2[6],mem[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm5, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1],xmm3[2],mem[2],xmm3[3],mem[3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm13[1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,4,7,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm10, %ymm1
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm15, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm13[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3]
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3,4],xmm1[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm12[0,1,2,3,4,5],xmm11[6],xmm12[7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm2[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm2, %xmm2
; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm15[0],xmm11[1],xmm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,4,7,7]
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm12[4],xmm3[5],xmm12[5],xmm3[6],xmm12[6],xmm3[7],xmm12[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm14[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm7, %ymm1
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3,4],xmm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3,4,5],xmm0[6],mem[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm0, %xmm2
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm0[1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0],xmm8[1],xmm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,4,7,7]
; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm7, %ymm1
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm0, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm10, %ymm1
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3,4],xmm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3,4,5],xmm0[6],mem[7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm0, %xmm2
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm0[0],mem[1],xmm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
-; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,4,7,7]
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm7, %ymm1
-; AVX1-ONLY-NEXT: vandps %ymm7, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX1-ONLY-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = xmm14[0],mem[0],xmm14[1],mem[1],xmm14[2],mem[2],xmm14[3],mem[3]
; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrlq $48, %xmm12, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm11[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3,4],xmm1[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3,4,5],xmm2[6],mem[7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm2[5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm14[0],xmm3[1],xmm14[1],xmm3[2],xmm14[2],xmm3[3],xmm14[3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,4,7,7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
+; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm10, %ymm1
+; AVX1-ONLY-NEXT: vandps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrlq $48, %xmm1, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm4 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3,4],xmm3[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm9[0,1,2,3,4,5],xmm7[6],xmm9[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $191, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2,3,4,5],xmm1[6],mem[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,7,6]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4],xmm3[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrld $16, %xmm15, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrld $16, %xmm1, %xmm3
; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,2,2,2]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3],xmm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0],xmm13[1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[0],xmm4[1],mem[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,4,7,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm10[4],xmm4[5],xmm10[5],xmm4[6],xmm10[6],xmm4[7],xmm10[7]
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,2,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,0,2]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm5 = xmm8[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm5 = xmm5[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm4, %ymm4
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm4 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm4, %ymm2
-; AVX1-ONLY-NEXT: vandps %ymm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm10, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm10, %ymm3
; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm3[1],xmm6[1],zero
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; AVX1-ONLY-NEXT: vinsertps $41, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = zero,xmm3[1],mem[0],zero
+; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm4 = xmm13[2],xmm15[2],xmm13[3],xmm15[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2],xmm3[3,4],xmm4[5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm2[0],mem[1],xmm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm11[0],xmm12[1],xmm11[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4],xmm4[5,6,7]
@@ -10048,57 +10052,65 @@
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm5 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3],xmm5[4,5,6,7]
-; AVX1-ONLY-NEXT: vpmovzxwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; AVX1-ONLY-NEXT: vpmovzxwd {{.*#+}} xmm5 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3]
; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5],xmm6[6,7]
-; AVX1-ONLY-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vpermilps $255, (%rsp), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[3,3,3,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm5
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm2 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm2, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm2, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm10, %ymm3
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm10, %ymm4
; AVX1-ONLY-NEXT: vorps %ymm3, %ymm4, %ymm3
-; AVX1-ONLY-NEXT: vinsertps $41, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = zero,xmm14[1],mem[0],zero
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm5 = xmm11[2],xmm12[2],xmm11[3],xmm12[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vinsertps $41, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = zero,xmm0[1],mem[0],zero
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3,4],xmm5[5,6,7]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm7[0],xmm9[1],xmm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm0[0],mem[1],xmm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm5[5,6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm0[1,1,1,1]
-; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm6 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm6 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm15[4],mem[4],xmm15[5],mem[5],xmm15[6],mem[6],xmm15[7],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[2,1,2,3]
; AVX1-ONLY-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[0,3,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3],xmm6[4,5,6,7]
; AVX1-ONLY-NEXT: vpmovzxwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm13[0],xmm6[0],xmm13[1],xmm6[1],xmm13[2],xmm6[2],xmm13[3],xmm6[3]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm10[0,1,0,3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
+; AVX1-ONLY-NEXT: vpshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5],xmm7[6,7]
-; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm8[3,3,3,3]
+; AVX1-ONLY-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[3,3,3,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm2, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm2, %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm10, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm5, %ymm10, %ymm5
; AVX1-ONLY-NEXT: vorps %ymm4, %ymm5, %ymm4
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vinsertps $41, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
@@ -10108,13 +10120,12 @@
; AVX1-ONLY-NEXT: # xmm6 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3,4],xmm6[5,6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0],xmm0[1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm0[0],mem[1],xmm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm6[5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[1,1,1,1]
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm6 = xmm14[1,1,1,1]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm7 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
@@ -10137,8 +10148,8 @@
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm7, %ymm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm7[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm2, %ymm5
-; AVX1-ONLY-NEXT: vandps %ymm2, %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm10, %ymm5
+; AVX1-ONLY-NEXT: vandps %ymm6, %ymm10, %ymm6
; AVX1-ONLY-NEXT: vorps %ymm5, %ymm6, %ymm5
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vinsertps $41, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm6 # 16-byte Folded Reload
@@ -10153,9 +10164,8 @@
; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,1,0,3]
; AVX1-ONLY-NEXT: vpshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,4,7]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4],xmm7[5,6,7]
-; AVX1-ONLY-NEXT: vpshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[1,1,1,1]
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm1[1,1,1,1]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vpsrldq {{.*#+}} xmm8 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -10177,11 +10187,11 @@
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm8, %ymm8
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm2, %ymm6
-; AVX1-ONLY-NEXT: vandps %ymm2, %ymm7, %ymm0
+; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm10, %ymm6
+; AVX1-ONLY-NEXT: vandps %ymm7, %ymm10, %ymm0
; AVX1-ONLY-NEXT: vorps %ymm6, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
@@ -10194,8 +10204,8 @@
; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -10217,12 +10227,11 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r9)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%r9)
-; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r9)
+; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -10230,29 +10239,30 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rax)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rax)
-; AVX1-ONLY-NEXT: addq $1544, %rsp # imm = 0x608
+; AVX1-ONLY-NEXT: addq $1512, %rsp # imm = 0x5E8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-SLOW-LABEL: load_i16_stride7_vf64:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $1464, %rsp # imm = 0x5B8
-; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm15
-; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm9
-; AVX2-SLOW-NEXT: vmovdqa 512(%rdi), %ymm14
+; AVX2-SLOW-NEXT: subq $1480, %rsp # imm = 0x5C8
+; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vmovdqa 512(%rdi), %ymm6
; AVX2-SLOW-NEXT: vmovdqa 544(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %ymm10
-; AVX2-SLOW-NEXT: vmovdqa 448(%rdi), %ymm12
+; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %ymm14
+; AVX2-SLOW-NEXT: vmovdqa 448(%rdi), %ymm9
; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm4
; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
@@ -10264,213 +10274,206 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4],xmm0[5],xmm2[6],xmm0[7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,14,15,12,13,10,11,8,9,8,9,8,9,8,9]
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm0, %xmm4
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
-; AVX2-SLOW-NEXT: vpblendvb %ymm6, %ymm4, %ymm3, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm13
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm13 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm4, %ymm3, %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0,1],ymm10[2],ymm12[3,4,5],ymm10[6],ymm12[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm14[2],ymm9[3,4,5],ymm14[6],ymm9[7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm14[1],ymm5[2,3,4],ymm14[5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm14, %ymm6
-; AVX2-SLOW-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm14
; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm6[1],ymm5[2,3,4],ymm6[5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm3, %ymm4, %ymm0
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm15[2],ymm9[3,4,5],ymm15[6],ymm9[7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm3, %ymm4, %ymm3
+; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm8[2],ymm15[3,4,5],ymm8[6],ymm15[7]
; AVX2-SLOW-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm8
-; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm8[1],ymm5[2,3,4],ymm8[5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm7
-; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm10
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm3, %ymm4, %ymm0
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 704(%rdi), %ymm3
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 672(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0,1],ymm3[2],ymm0[3,4,5],ymm3[6],ymm0[7]
+; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm3, %ymm4, %ymm3
+; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 704(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vmovdqa 672(%rdi), %ymm12
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0,1],ymm8[2],ymm12[3,4,5],ymm8[6],ymm12[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm2
-; AVX2-SLOW-NEXT: vmovdqa 736(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vmovdqa 768(%rdi), %ymm11
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0],ymm0[1],ymm11[2,3,4],ymm0[5],ymm11[6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 736(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovdqa 768(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0],ymm6[1],ymm4[2,3,4],ymm6[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm11
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm3, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa %ymm13, %ymm5
; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm2, %ymm1, %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1],ymm6[2],ymm14[3,4],ymm6[5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm0[2],ymm5[3,4],ymm0[5],ymm5[6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7,8,9,10],ymm2[11],ymm1[12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1,2],ymm10[3],ymm12[4,5],ymm10[6],ymm12[7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm10, %ymm14
-; AVX2-SLOW-NEXT: vmovdqa %ymm12, %ymm13
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1,2],ymm14[3],ymm9[4,5],ymm14[6],ymm9[7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm14
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm2[1],xmm3[2,3,4,5],xmm2[6],xmm3[7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = [2,3,2,3,2,3,2,3,8,9,8,9,6,7,4,5,18,19,18,19,18,19,18,19,24,25,24,25,22,23,20,21]
; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm1, %ymm4
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = [2,3,0,1,14,15,12,13,10,11,10,11,10,11,10,11]
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm3
-; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm3
-; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm6
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm8[2],ymm7[3,4],ymm8[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm3, %ymm4, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1],ymm7[2],ymm10[3,4],ymm7[5],ymm10[6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3],ymm3[4,5,6,7,8,9,10],ymm4[11],ymm3[12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1,2],ymm15[3],ymm9[4,5],ymm15[6],ymm9[7]
+; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm15[0,1,2],mem[3],ymm15[4,5],mem[6],ymm15[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm3, %ymm3
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm4
-; AVX2-SLOW-NEXT: vpblendvb %ymm6, %ymm4, %ymm3, %ymm3
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm11, %ymm12
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm0[2],ymm11[3,4],ymm0[5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm4, %ymm3, %ymm0
+; AVX2-SLOW-NEXT: vmovdqa %ymm13, %ymm15
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm6[2],ymm11[3,4],ymm6[5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm10
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3],ymm3[4,5,6,7,8,9,10],ymm4[11],ymm3[12,13,14,15]
-; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0,1,2],ymm15[3],ymm10[4,5],ymm15[6],ymm10[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0,1,2],ymm8[3],ymm12[4,5],ymm8[6],ymm12[7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm8, %ymm13
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm3, %ymm3
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm4, %xmm4
-; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm0
-; AVX2-SLOW-NEXT: vpblendvb %ymm6, %ymm4, %ymm3, %ymm3
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm6[2],ymm5[3,4],ymm6[5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm4, %ymm3, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm6[2],ymm9[3,4],ymm6[5],ymm9[6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3],ymm3[4,5,6,7,8,9,10],ymm4[11],ymm3[12,13,14,15]
; AVX2-SLOW-NEXT: vpshufb %ymm2, %ymm3, %ymm2
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2],ymm7[3],ymm9[4,5],ymm7[6],ymm9[7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm7[3],ymm5[4,5],ymm7[6],ymm5[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3,4,5],xmm3[6],xmm4[7]
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm3, %xmm1
-; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm1
-; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm1, %ymm2, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm6[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm6[2,3],ymm5[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1],ymm6[2,3],ymm9[4,5],ymm6[6,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4],ymm2[5,6,7,8,9,10,11],ymm1[12],ymm2[13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0],ymm9[1],ymm7[2,3],ymm9[4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0],ymm5[1],ymm7[2,3],ymm5[4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm7, %ymm9
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm3
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0],xmm3[1],xmm1[2,3,4,5],xmm3[6],xmm1[7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = [4,5,4,5,4,5,4,5,8,9,10,11,8,9,6,7,20,21,20,21,20,21,20,21,24,25,26,27,24,25,22,23]
; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm2, %ymm4
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = [4,5,2,3,0,1,14,15,12,13,12,13,12,13,12,13]
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm3, %ymm4, %ymm3
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm3, %ymm4, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0],ymm13[1],ymm14[2,3],ymm13[4],ymm14[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm13, %ymm8
-; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5],xmm4[6],xmm3[7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1],ymm11[2,3],ymm6[4,5],ymm11[6,7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm11[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7,8,9,10,11],ymm5[12],ymm4[13,14,15]
-; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm3, %ymm4, %ymm3
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0],ymm10[1],ymm15[2,3],ymm10[4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0],ymm14[1],ymm0[2,3],ymm14[4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5],xmm4[6],xmm3[7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0,1],ymm5[2,3],ymm12[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm12, %ymm15
-; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1],ymm5[2,3],ymm8[4,5],ymm5[6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm5[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7,8,9,10,11],ymm5[12],ymm4[13,14,15]
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm3
; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vmovdqa %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm3, %ymm4, %ymm0
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm3, %ymm4, %ymm3
+; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0],ymm12[1],ymm13[2,3],ymm12[4],ymm13[5,6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5],xmm4[6],xmm3[7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %ymm11, %ymm6
+; AVX2-SLOW-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm10[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7,8,9,10,11],ymm5[12],ymm4[13,14,15]
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm3
+; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm4, %ymm4
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm3, %ymm4, %ymm3
+; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0],ymm10[1],ymm0[2,3],ymm10[4],ymm0[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0],ymm10[1],ymm11[2,3],ymm10[4],ymm11[5,6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5],xmm4[6],xmm3[7]
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm3, %xmm2
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1],ymm12[2,3],ymm13[4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm5[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4],ymm3[5,6,7,8,9,10,11],ymm4[12],ymm3[13,14,15]
; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm3, %ymm1
-; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm2, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm2, %ymm1, %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0],ymm9[1],ymm7[2,3,4],ymm9[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2],ymm1[3],mem[4,5],ymm1[6],mem[7]
+; AVX2-SLOW-NEXT: vpblendd $72, (%rsp), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm1[0,1,2],mem[3],ymm1[4,5],mem[6],ymm1[7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm1[1,3,2,3]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = [6,7,6,7,6,7,6,7,8,9,4,5,10,11,0,1,22,23,22,23,22,23,22,23,24,25,20,21,26,27,16,17]
; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm3, %ymm3
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm2, %ymm3, %ymm2
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm2, %ymm3, %ymm2
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0],ymm8[1],ymm14[2,3,4],ymm8[5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm0[0],ymm14[1],ymm0[2,3,4],ymm14[5],ymm0[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3],xmm3[4,5,6,7]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2],ymm11[3],ymm6[4,5],ymm11[6],ymm6[7]
+; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm8[0,1,2],mem[3],ymm8[4,5],mem[6],ymm8[7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,3,2,3]
; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm3, %ymm3
-; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm2, %ymm3, %ymm2
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3],xmm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1,2],ymm12[3],ymm13[4,5],ymm12[6],ymm13[7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,3,2,3]
-; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm3, %ymm3
-; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm2, %ymm3, %ymm0
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm2, %ymm3, %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0],ymm10[1],ymm11[2,3,4],ymm10[5],ymm11[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3],xmm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm15[0,1,2],mem[3],ymm15[4,5],mem[6],ymm15[7]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1,2],ymm5[3],ymm7[4,5],ymm5[6],ymm7[7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,3,2,3]
+; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm3, %ymm3
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm2, %ymm3, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0],ymm12[1],ymm13[2,3,4],ymm12[5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3],xmm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm6[0,1,2],mem[3],ymm6[4,5],mem[6],ymm6[7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,3,2,3]
; AVX2-SLOW-NEXT: vpshufb %ymm1, %ymm3, %ymm1
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,2,1,0,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm2, %ymm1, %ymm0
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm2, %ymm1, %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm10
-; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1,2],ymm10[3],ymm12[4,5],ymm10[6],ymm12[7]
+; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm13
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm8[3],ymm13[4,5],ymm8[6],ymm13[7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,6,4,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
@@ -10479,18 +10482,17 @@
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm9 = ymm0[0,1,0,2]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm9[0,1,2,1,4,5,6,5]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm10 = ymm0[0,1,0,2]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm10[0,1,2,1,4,5,6,5]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %ymm6
-; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %ymm8
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2],ymm6[3],ymm8[4,5],ymm6[6],ymm8[7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1,2],ymm15[3],ymm7[4,5],ymm15[6],ymm7[7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,6,4,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
@@ -10506,49 +10508,49 @@
; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2],ymm0[3],ymm4[4,5],ymm0[6],ymm4[7]
+; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1,2],ymm4[3],ymm6[4,5],ymm4[6],ymm6[7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm0, %ymm11
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,6,4,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,7,7]
; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %ymm15
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm15[0,1,0,2]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm7 = ymm2[0,1,2,1,4,5,6,5]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %ymm14
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm14[0,1,0,2]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm9 = ymm2[0,1,2,1,4,5,6,5]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm9[7]
; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 800(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovdqa 832(%rdi), %ymm5
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2],ymm0[3],ymm5[4,5],ymm0[6],ymm5[7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm0, %ymm7
+; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm11
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, %ymm9
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,6,4,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,7,7]
; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa 864(%rdi), %ymm14
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm14[0,1,0,2]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm13 = ymm0[0,1,2,1,4,5,6,5]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT: vmovdqa 864(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm5[0,1,0,2]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm12 = ymm0[0,1,2,1,4,5,6,5]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm12[7]
; AVX2-SLOW-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm8[2],ymm6[3,4,5],ymm8[6],ymm6[7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm13
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm13[4],xmm1[5],xmm13[6],xmm1[7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm13 = [0,1,0,1,14,15,12,13,10,11,8,9,12,13,14,15]
-; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm7[2],ymm15[3,4,5],ymm7[6],ymm15[7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm12
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm12[4],xmm1[5],xmm12[6],xmm1[7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = [0,1,0,1,14,15,12,13,10,11,8,9,12,13,14,15]
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[0,1,1,3,4,5,5,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,5,6,8,9,10,11,12,13,13,14]
@@ -10557,10 +10559,10 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0],ymm1[1,2,3,4,5,6,7],ymm3[8],ymm1[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm4[2],ymm11[3,4,5],ymm4[6],ymm11[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm6[2],ymm4[3,4,5],ymm6[6],ymm4[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm3
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4],xmm1[5],xmm3[6],xmm1[7]
-; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[0,1,1,3,4,5,5,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,5,6,8,9,10,11,12,13,13,14]
@@ -10569,11 +10571,10 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3,4,5,6,7],ymm2[8],ymm1[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm11
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm5[2],ymm7[3,4,5],ymm5[6],ymm7[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm11[2],ymm9[3,4,5],ymm11[6],ymm9[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4],xmm1[5],xmm2[6],xmm1[7]
-; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,5,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,5,6,8,9,10,11,12,13,13,14]
@@ -10582,29 +10583,28 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1],ymm12[2],ymm10[3,4,5],ymm12[6],ymm10[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1],ymm13[2],ymm8[3,4,5],ymm13[6],ymm8[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4],xmm0[5],xmm1[6],xmm0[7]
-; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm9[0,1,1,3,4,5,5,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm10[0,1,1,3,4,5,5,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,5,6,8,9,10,11,12,13,13,14]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm12[3],ymm10[4,5],ymm12[6],ymm10[7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm12, %ymm6
-; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm13[3],ymm8[4,5],ymm13[6],ymm8[7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3,4,5],xmm0[6],xmm1[7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,0,1,14,15,12,13,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm8[0,1,1,2]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm7[0,1,1,2]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[0,1,0,3,4,5,4,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
@@ -10612,9 +10612,9 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %ymm15, %ymm3
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm13[3],ymm4[4,5],ymm13[6],ymm4[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2],ymm4[3],ymm15[4,5],ymm4[6],ymm15[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm0, %xmm0
@@ -10628,15 +10628,14 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2],ymm5[3],ymm7[4,5],ymm5[6],ymm7[7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm7, %ymm5
-; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2],ymm11[3],ymm9[4,5],ymm11[6],ymm9[7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm14[0,1,1,2]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm5[0,1,1,2]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[0,1,0,3,4,5,4,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
@@ -10644,14 +10643,14 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2],ymm3[3],ymm9[4,5],ymm3[6],ymm9[7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2],ymm10[3],ymm15[4,5],ymm10[6],ymm15[7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm15[0,1,1,2]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm14[0,1,1,2]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,1,0,3,4,5,4,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
@@ -10659,13 +10658,13 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0],ymm10[1],ymm6[2,3],ymm10[4],ymm6[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0],ymm8[1],ymm13[2,3],ymm8[4],ymm13[5,6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,0,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,6,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm8[0,1,1,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm7[0,1,1,3]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,7,4,8,9,10,11,12,13,15,12]
@@ -10674,9 +10673,9 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0],ymm4[1],ymm13[2,3],ymm4[4],ymm13[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm13, %ymm7
-; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm8
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm7
+; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,0,4,5,6,7]
@@ -10690,28 +10689,27 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0],ymm9[1],ymm3[2,3],ymm9[4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm12
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm15[1],ymm10[2,3],ymm15[4],ymm10[5,6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,0,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,6,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm15[0,1,1,3]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm3[0,1,2,0,4,5,6,4]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm8 = ymm14[0,1,1,3]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm8[0,1,2,0,4,5,6,4]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,7,4,8,9,10,11,12,13,15,12]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0],ymm5[1],ymm11[2,3],ymm5[4],ymm11[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0],ymm9[1],ymm11[2,3],ymm9[4],ymm11[5,6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,2,1,0,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,6,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm14[0,1,1,3]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm5[0,1,1,3]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm4 = ymm1[0,1,2,0,4,5,6,4]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,7,4,8,9,10,11,12,13,15,12]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm4[7]
@@ -10719,24 +10717,24 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2,3,4,5,6,7],ymm4[8],ymm0[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1],ymm10[2],ymm14[3,4,5],ymm10[6],ymm14[7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0,1],ymm11[2],ymm13[3,4,5],ymm11[6],ymm13[7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm0 = <u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27>
; AVX2-SLOW-NEXT: vpshufb %ymm0, %ymm4, %ymm5
-; AVX2-SLOW-NEXT: vmovdqa %ymm0, %ymm9
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, %ymm10
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7,8,9,10,11,12,13],ymm4[14],ymm5[15]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm11[0,1],ymm13[2],ymm11[3,4],ymm13[5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm12[0,1],ymm9[2],ymm12[3,4],ymm9[5],ymm12[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0],ymm8[1],ymm7[2,3,4],ymm8[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0],ymm3[1],ymm7[2,3,4],ymm3[5],ymm7[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0],xmm5[1],xmm6[2],xmm5[3],xmm6[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,6,7,4,5,2,3,0,1,14,15,12,13,14,15]
@@ -10749,29 +10747,27 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $187, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm6[2],mem[3,4,5],ymm6[6],mem[7]
-; AVX2-SLOW-NEXT: vpshufb %ymm9, %ymm2, %ymm4
-; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm7
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm7[2],ymm6[3,4,5],ymm7[6],ymm6[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm10, %ymm2, %ymm4
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6],ymm4[7,8,9,10,11,12,13],ymm2[14],ymm4[15]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm15[0,1],ymm9[2],ymm15[3,4],ymm9[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1],ymm10[2],ymm14[3,4],ymm10[5],ymm14[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0],ymm12[1],mem[2,3,4],ymm12[5],mem[6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm12, %ymm8
+; AVX2-SLOW-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0],ymm15[1],mem[2,3,4],ymm15[5],mem[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2],xmm4[3],xmm5[4,5,6,7]
; AVX2-SLOW-NEXT: vpshufb %xmm0, %xmm4, %xmm4
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[0,1,2,1,4,5,6,5]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm8[0,1,2,1,4,5,6,5]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5,6],ymm3[7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm2[0],ymm3[1,2,3,4,5,6,7],ymm2[8],ymm3[9,10,11,12,13,14,15]
@@ -10780,12 +10776,13 @@
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vpblendd $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm2 = ymm2[0,1],mem[2],ymm2[3,4,5],mem[6],ymm2[7]
-; AVX2-SLOW-NEXT: vpshufb %ymm7, %ymm2, %ymm3
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27>
+; AVX2-SLOW-NEXT: vpshufb %ymm5, %ymm2, %ymm3
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7,8,9,10,11,12,13],ymm2[14],ymm3[15]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $219, (%rsp), %ymm5, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,1],ymm5[2],mem[3,4],ymm5[5],mem[6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm8[2],ymm15[3,4],ymm8[5],ymm15[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
@@ -10816,15 +10813,15 @@
; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,1,4,5,6,5]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6,5,8,9,10,11,12,13,14,13]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1],ymm7[2],ymm12[3,4,5],ymm7[6],ymm12[7]
-; AVX2-SLOW-NEXT: vpshufb {{.*#+}} ymm0 = ymm2[u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27]
+; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm0[0,1],mem[2],ymm0[3,4,5],mem[6],ymm0[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm5, %ymm2, %ymm0
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6],ymm0[7,8,9,10,11,12,13],ymm2[14],ymm0[15]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = ymm2[0,1],mem[2],ymm2[3,4],mem[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm2[2],mem[3,4],ymm2[5],mem[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
@@ -10834,139 +10831,139 @@
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3,4,5,6,7],ymm0[8],ymm1[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm10[3],ymm14[4,5],ymm10[6],ymm14[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm11[3],ymm13[4,5],ymm11[6],ymm13[7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7,8],ymm0[9,10,11,12,13,14],ymm1[15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm13[2,3],ymm11[4,5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm9[2,3],ymm12[4,5],ymm9[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[0,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm11 = <u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29>
-; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = ymm1[0,1],mem[2],ymm1[3,4],mem[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2],xmm2[3],xmm1[4],xmm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm10 = [8,9,8,9,6,7,4,5,2,3,2,3,2,3,2,3]
-; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm1, %xmm1
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa 656(%rdi), %xmm3
-; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 640(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm2[0,1,2,3,4,5],xmm3[6],xmm2[7]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,7,6]
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5,6,7],ymm1[8,9,10,11,12],ymm4[13,14,15]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3,4,5,6,7],ymm0[8],ymm1[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,2],ymm6[3],mem[4,5],ymm6[6],mem[7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7,8],ymm0[9,10,11,12,13,14],ymm1[15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm9[2,3],ymm15[4,5],ymm9[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm4
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,2,2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm8[2],mem[3,4],ymm8[5],mem[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm4
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm1 = xmm4[0,1],xmm1[2],xmm4[3],xmm1[4],xmm4[5,6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm1, %xmm1
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa 432(%rdi), %xmm3
-; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %xmm4
-; AVX2-SLOW-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm3[6],xmm4[7]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,7,6]
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5,6,7],ymm1[8,9,10,11,12],ymm4[13,14,15]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3,4,5,6,7],ymm0[8],ymm1[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm14[3],ymm13[4,5],ymm14[6],ymm13[7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7,8],ymm0[9,10,11,12,13,14],ymm1[15]
-; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm5[2,3],ymm9[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm4
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,2,2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1],ymm3[2],ymm8[3,4],ymm3[5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm12 = <u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29>
+; AVX2-SLOW-NEXT: vpshufb %ymm12, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = mem[0,1],ymm11[2],mem[3,4],ymm11[5],mem[6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4],xmm1[5,6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vmovdqa 880(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vmovdqa 864(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm6 = xmm0[0,1,2,3,4,5],xmm1[6],xmm0[7]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,7,6]
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7],ymm5[8,9,10,11,12],ymm6[13,14,15]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0],ymm5[1,2,3,4,5,6,7],ymm4[8],ymm5[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,1],ymm15[2],mem[3,4],ymm15[5],mem[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2],xmm5[3],xmm4[4],xmm5[5,6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm13 = [8,9,8,9,6,7,4,5,2,3,2,3,2,3,2,3]
+; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm3
+; AVX2-SLOW-NEXT: vmovdqa 656(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovdqa 640(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,5],xmm1[6],xmm0[7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,7,6]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-SLOW-NEXT: vmovdqa 208(%rdi), %xmm6
-; AVX2-SLOW-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %xmm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7],ymm3[8,9,10,11,12],ymm4[13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm2[0],ymm3[1,2,3,4,5,6,7],ymm2[8],ymm3[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2],ymm7[3],ymm6[4,5],ymm7[6],ymm6[7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3,4,5,6],ymm3[7,8],ymm2[9,10,11,12,13,14],ymm3[15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1],ymm10[2,3],ymm14[4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[0,2,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX2-SLOW-NEXT: vpshufb %ymm12, %ymm2, %ymm2
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm9[2],ymm7[3,4],ymm9[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1],xmm3[2],xmm4[3],xmm3[4],xmm4[5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm3, %xmm3
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm4
+; AVX2-SLOW-NEXT: vmovdqa 432(%rdi), %xmm5
; AVX2-SLOW-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5],xmm6[6],xmm5[7]
+; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %xmm3
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm3[0,1,2,3,4,5],xmm5[6],xmm3[7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,0,3]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,7,6]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm12[0,1,2],ymm7[3],ymm12[4,5],ymm7[6],ymm12[7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm2[0],ymm4[1,2,3,4,5,6,7],ymm2[8],ymm4[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm10[0,1,2],mem[3],ymm10[4,5],mem[6],ymm10[7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm2[1,2,3,4,5,6],ymm4[7,8],ymm2[9,10,11,12,13,14],ymm4[15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm15[0,1],ymm8[2,3],ymm15[4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm5 = xmm5[0,2,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; AVX2-SLOW-NEXT: vpshufb %ymm12, %ymm2, %ymm2
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1],ymm15[2],ymm14[3,4],ymm15[5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2],xmm5[3],xmm4[4],xmm5[5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-SLOW-NEXT: vmovdqa 880(%rdi), %xmm5
+; AVX2-SLOW-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 864(%rdi), %xmm6
+; AVX2-SLOW-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2,3,4,5],xmm5[6],xmm6[7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,0,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,7,6]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm2[0],ymm4[1,2,3,4,5,6,7],ymm2[8],ymm4[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm8[2],mem[3,4],ymm8[5],mem[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm4
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1],xmm2[2],xmm4[3],xmm2[4],xmm4[5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm2, %xmm2
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vmovdqa 208(%rdi), %xmm4
+; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %xmm5
+; AVX2-SLOW-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5],xmm4[6],xmm5[7]
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,0,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,7,6]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7],ymm2[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm5[0,1,2],mem[3],ymm5[4,5],mem[6],ymm5[7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm6 = ymm5[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm6[0],ymm5[1,2,3,4,5,6],ymm6[7,8],ymm5[9,10,11,12,13,14],ymm6[15]
-; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm5, %ymm5
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1],ymm10[2,3],mem[4,5],ymm10[6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm10, %ymm12
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm6, %xmm11
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm11 = xmm11[2,1,2,3]
-; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm11 = xmm11[0,2,2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %ymm12, %ymm5, %ymm5
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = ymm13[0,1],mem[2,3],ymm13[4,5],mem[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm6, %xmm12
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm12 = xmm12[2,1,2,3]
+; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm12 = xmm12[0,2,2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[2,1,2,3]
; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm11[0],xmm6[1],xmm11[1],xmm6[2],xmm11[2],xmm6[3],xmm11[3]
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3,4,5,6,7],ymm5[8],ymm4[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm10 = ymm5[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3],mem[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4,5,6,7,8],ymm5[9],ymm4[10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3,4,5,6,7],ymm5[8],ymm2[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3],mem[4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm2[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4,5,6,7,8],ymm5[9],ymm2[10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-SLOW-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,2],ymm5[3],mem[4,5],ymm5[6],mem[7]
@@ -10975,109 +10972,107 @@
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,7,6,7]
; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm11 = <u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31>
-; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm5[0,1],mem[2,3],ymm5[4,5],mem[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3],xmm6[4],xmm5[5],xmm6[6,7]
-; AVX2-SLOW-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm2 = mem[0],xmm2[1],mem[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm7 = [8,9,10,11,8,9,6,7,4,5,4,5,4,5,4,5]
-; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm5, %xmm5
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm5[0,1,2,3,4],ymm2[5,6,7],ymm5[8,9,10,11,12],ymm2[13,14,15]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm2[1,2,3,4,5,6,7],ymm4[8],ymm2[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0],ymm13[1],ymm14[2,3],ymm13[4],ymm14[5,6,7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4,5,6,7,8],ymm5[9],ymm4[10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm9[0,1,2],mem[3],ymm9[4,5],mem[6],ymm9[7]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm6 = xmm5[0,1,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
-; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1],ymm3[2,3],ymm8[4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3],xmm6[4],xmm5[5],xmm6[6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm12 = <u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31>
+; AVX2-SLOW-NEXT: vpshufb %ymm12, %ymm2, %ymm2
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm11[2,3],mem[4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm6
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2],xmm2[3],xmm6[4],xmm2[5],xmm6[6,7]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm5, %xmm1
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm2 = [8,9,10,11,8,9,6,7,4,5,4,5,4,5,4,5]
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm6, %xmm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2,3,4,5,6,7],ymm4[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6,7]
-; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm4 = ymm0[2,3,0,1]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4,5,6,7,8],ymm4[9],ymm0[10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,2],ymm3[3],mem[4,5],ymm3[6],mem[7]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm4[0,1,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
-; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3],xmm5[4],xmm4[5],xmm5[6,7]
-; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-SLOW-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm5 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm5 = mem[0],xmm3[1],mem[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm4, %xmm4
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,0,3]
-; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2,3,4,5,6,7],ymm5[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3],mem[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm1[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4,5,6,7,8],ymm5[9],ymm1[10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,2],ymm5[3],mem[4,5],ymm5[6],mem[7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm6 = xmm5[0,1,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm12, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3],xmm6[4],xmm5[5],xmm6[6,7]
+; AVX2-SLOW-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm3 = mem[0],xmm3[1],mem[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm5
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm0[0],ymm4[1,2,3,4,5,6,7],ymm0[8],ymm4[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0,1],ymm15[2,3],mem[4,5],ymm15[6,7]
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3],xmm5[4],xmm4[5],xmm5[6,7]
-; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm4, %xmm3
-; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm3[5,6,7],ymm5[8,9,10,11,12],ymm3[13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm1[0],ymm3[1,2,3,4,5,6,7],ymm1[8],ymm3[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = mem[0],ymm10[1],mem[2,3],ymm10[4],mem[5,6,7]
+; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm3[2,3,0,1]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2,3,4,5,6,7,8],ymm5[9],ymm3[10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm5[0,1,2],mem[3],ymm5[4,5],mem[6],ymm5[7]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm6 = xmm5[0,1,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm12, %ymm3, %ymm3
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm5 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3],xmm6[4],xmm5[5],xmm6[6,7]
+; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX2-SLOW-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm6 = mem[0],xmm6[1],mem[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm5
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
+; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,4,7]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7],ymm5[8,9,10,11,12],ymm6[13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm5 = ymm3[0],ymm5[1,2,3,4,5,6,7],ymm3[8],ymm5[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[0,1],ymm8[2,3],mem[4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3],xmm6[4],xmm5[5],xmm6[6,7]
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm2
; AVX2-SLOW-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7],ymm3[8,9,10,11,12],ymm4[13,14,15]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7],ymm2[8,9,10,11,12],ymm4[13,14,15]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[0],ymm4[1],mem[2,3],ymm4[4],mem[5,6,7]
+; AVX2-SLOW-NEXT: vpblendd $18, (%rsp), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3],mem[4],ymm4[5,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4,5,6,7,8],ymm5[9],ymm4[10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpshufb %ymm11, %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,2],ymm12[3],mem[4,5],ymm12[6],mem[7]
+; AVX2-SLOW-NEXT: vpshufb %ymm12, %ymm4, %ymm4
+; AVX2-SLOW-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm13[0,1,2],mem[3],ymm13[4,5],mem[6],ymm13[7]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm6 = xmm5[0,1,2,3,6,4,6,7]
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm5
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
; AVX2-SLOW-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,7,6,7]
; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3,4,5,6,7],ymm4[8],ymm3[9,10,11,12,13,14,15]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm2[1,2,3,4,5,6,7],ymm4[8],ymm2[9,10,11,12,13,14,15]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%rsi)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
@@ -11115,9 +11110,9 @@
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%r9)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm4, (%r9)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm4, 64(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%rax)
@@ -11125,13 +11120,14 @@
; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%rax)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm4, 64(%rax)
-; AVX2-SLOW-NEXT: vmovdqa %ymm10, (%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%rax)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovdqa %ymm0, 32(%rax)
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, (%rax)
-; AVX2-SLOW-NEXT: vmovdqa %ymm1, 96(%rax)
-; AVX2-SLOW-NEXT: vmovdqa %ymm2, 64(%rax)
-; AVX2-SLOW-NEXT: addq $1464, %rsp # imm = 0x5B8
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, 96(%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, 32(%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, (%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, 64(%rax)
+; AVX2-SLOW-NEXT: addq $1480, %rsp # imm = 0x5C8
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
@@ -11139,181 +11135,183 @@
; AVX2-FAST: # %bb.0:
; AVX2-FAST-NEXT: subq $1544, %rsp # imm = 0x608
; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm5
-; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm6
-; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm9
-; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %ymm7
-; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %ymm11
-; AVX2-FAST-NEXT: vmovdqa 448(%rdi), %ymm12
-; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm10
-; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm13
-; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm5, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm15
+; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm11
+; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %ymm12
+; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %ymm13
+; AVX2-FAST-NEXT: vmovdqa 448(%rdi), %ymm14
+; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm3
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0],ymm1[1],ymm3[2,3,4],ymm1[5],ymm3[6,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm7
+; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm8
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [0,1,14,15,12,13,14,15,2,3,6,7,12,13,2,3,16,17,30,31,28,29,30,31,18,19,22,23,28,29,18,19]
; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1],ymm13[2],ymm10[3,4,5],ymm13[6],ymm10[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm4[2],ymm2[3,4,5],ymm4[6],ymm2[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm9
+; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm10
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4],xmm0[5],xmm2[6],xmm0[7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,14,15,12,13,10,11,8,9,8,9,8,9,8,9]
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm4
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm4, %ymm3, %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0,1],ymm11[2],ymm12[3,4,5],ymm11[6],ymm12[7]
-; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0],ymm9[1],ymm7[2,3,4],ymm9[5],ymm7[6,7]
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
-; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm3, %ymm4, %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1],ymm5[2],ymm6[3,4,5],ymm5[6],ymm6[7]
-; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm9
-; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm7
-; AVX2-FAST-NEXT: vmovdqu %ymm5, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm6
-; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm14
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0],ymm6[1],ymm14[2,3,4],ymm6[5],ymm14[6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
-; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm3, %ymm4, %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm4, %ymm3, %ymm3
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 672(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0,1],ymm3[2],ymm0[3,4,5],ymm3[6],ymm0[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1],ymm13[2],ymm14[3,4,5],ymm13[6],ymm14[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0],ymm11[1],ymm12[2,3,4],ymm11[5],ymm12[6,7]
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
+; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm4
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm5[2],ymm15[3,4,5],ymm5[6],ymm15[7]
+; AVX2-FAST-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
+; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm12
+; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0],ymm12[1],ymm4[2,3,4],ymm12[5],ymm4[6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
+; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm4
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 672(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4,5],ymm4[6],ymm3[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm2
-; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 768(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0],ymm0[1],ymm3[2,3,4],ymm0[5],ymm3[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4],ymm4[5],ymm3[6,7]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm1
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm2, %ymm1, %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm13, %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm13[3],ymm10[4,5],ymm13[6],ymm10[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2],ymm9[3],ymm10[4,5],ymm9[6],ymm10[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm2[0],xmm1[1],xmm2[2,3,4,5],xmm1[6],xmm2[7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1],ymm8[2],ymm13[3,4],ymm8[5],ymm13[6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1],ymm8[2],ymm7[3,4],ymm8[5],ymm7[6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <2,5,1,u,4,u,u,u>
; AVX2-FAST-NEXT: vpermd %ymm2, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [2,3,2,3,2,3,2,3,8,9,0,1,6,7,8,9,18,19,18,19,18,19,18,19,24,25,16,17,22,23,24,25]
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm5
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [2,3,0,1,14,15,12,13,10,11,10,11,10,11,10,11]
-; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm4
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm4, %ymm5, %ymm4
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [2,3,2,3,2,3,2,3,8,9,0,1,6,7,8,9,18,19,18,19,18,19,18,19,24,25,16,17,22,23,24,25]
+; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm5
+; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm3
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = [2,3,0,1,14,15,12,13,10,11,10,11,10,11,10,11]
+; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm4, %xmm4
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm4, %ymm5, %ymm4
+; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0,1,2],ymm11[3],ymm12[4,5],ymm11[6],ymm12[7]
+; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1,2],ymm13[3],ymm14[4,5],ymm13[6],ymm14[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
-; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm4
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm4, %xmm4
+; AVX2-FAST-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,1],ymm11[2],mem[3,4],ymm11[5],mem[6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <2,5,1,u,4,u,u,u>
+; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
+; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm4, %ymm5, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd $72, (%rsp), %ymm15, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm15[0,1,2],mem[3],ymm15[4,5],mem[6],ymm15[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
+; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm4, %xmm4
+; AVX2-FAST-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,1],ymm12[2],mem[3,4],ymm12[5],mem[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
+; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vpblendvb %ymm2, %ymm4, %ymm5, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm11[0,1],ymm12[2],ymm11[3,4],ymm12[5],ymm11[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm4, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1,2],ymm7[3],ymm9[4,5],ymm7[6],ymm9[7]
+; AVX2-FAST-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1,2],ymm12[3],mem[4,5],ymm12[6],mem[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
-; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm14[0,1],ymm6[2],ymm14[3,4],ymm6[5],ymm14[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm4, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vmovdqa %ymm15, %ymm14
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm4, %xmm3
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0,1,2],ymm15[3],ymm7[4,5],ymm15[6],ymm7[7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
-; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm3
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1],ymm9[2],ymm6[3,4],ymm9[5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm15[0,1],ymm6[2],ymm15[3,4],ymm6[5],ymm15[6,7]
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm3, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0],ymm10[1],ymm0[2,3],ymm10[4],ymm0[5,6,7]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[2,3,2,3,2,3,2,3,8,9,0,1,6,7,8,9,18,19,18,19,18,19,18,19,24,25,16,17,22,23,24,25]
+; AVX2-FAST-NEXT: vpblendvb %ymm2, %ymm3, %ymm1, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0],ymm10[1],ymm9[2,3],ymm10[4],ymm9[5,6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm1[0],xmm2[1],xmm1[2,3,4,5],xmm2[6],xmm1[7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1],ymm8[2,3],ymm13[4,5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5],ymm8[6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <2,6,1,u,5,u,u,u>
; AVX2-FAST-NEXT: vpermd %ymm2, %ymm1, %ymm3
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [0,1,2,3,4,5,2,3,8,9,2,3,4,5,10,11,16,17,18,19,20,21,18,19,24,25,18,19,20,21,26,27]
; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm5
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [4,5,2,3,0,1,14,15,12,13,12,13,12,13,12,13]
; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm4
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm4, %ymm5, %ymm0
-; AVX2-FAST-NEXT: vmovdqa %ymm14, %ymm8
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0],ymm13[1],ymm14[2,3],ymm13[4],ymm14[5,6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm4, %ymm5, %ymm4
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0],ymm14[1],ymm13[2,3],ymm14[4],ymm13[5,6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm5[1],xmm4[2,3,4,5],xmm5[6],xmm4[7]
; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm11[0,1],ymm12[2,3],ymm11[4,5],ymm12[6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm11, %ymm0
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vpblendvb %ymm8, %ymm4, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vmovdqa %ymm8, %ymm10
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm15, %ymm8
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm15[0],ymm7[1],ymm15[2,3],ymm7[4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm5[1],xmm4[2,3,4,5],xmm5[6],xmm4[7]
-; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1],ymm9[2,3],ymm6[4,5],ymm9[6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm15
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm4, %ymm5, %ymm4
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm11[0],ymm10[1],ymm11[2,3],ymm10[4],ymm11[5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm10[0,1],ymm11[2,3],ymm10[4,5],ymm11[6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
+; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm4, %ymm5, %ymm4
+; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm9
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm13
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0],ymm0[1],ymm12[2,3],ymm0[4],ymm12[5,6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm5[1],xmm4[2,3,4,5],xmm5[6],xmm4[7]
+; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm4
+; AVX2-FAST-NEXT: vmovdqa %ymm15, %ymm8
+; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm7
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm15[0,1],ymm6[2,3],ymm15[4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
+; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vpblendvb %ymm9, %ymm4, %ymm5, %ymm4
+; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm12
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0],ymm6[1],ymm9[2,3],ymm6[4],ymm9[5,6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm5[1],xmm4[2,3,4,5],xmm5[6],xmm4[7]
; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm3
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1],ymm9[2,3],ymm6[4,5],ymm9[6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1],ymm15[2,3],ymm14[4,5],ymm15[6,7]
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
-; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm3, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm12, %ymm3, %ymm1, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3,4],ymm1[5],mem[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -11324,39 +11322,40 @@
; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm2, %ymm4
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [6,7,4,5,2,3,0,1,14,15,14,15,14,15,14,15]
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-NEXT: vpblendvb %ymm12, %ymm3, %ymm4, %ymm3
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0],ymm13[1],ymm14[2,3,4],ymm13[5],ymm14[6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0],ymm3[1],mem[2,3,4],ymm3[5],mem[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2],xmm3[3],xmm4[4,5,6,7]
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0,1,2],ymm12[3],ymm0[4,5],ymm12[6],ymm0[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0,1,2],ymm11[3],ymm10[4,5],ymm11[6],ymm10[7]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm4 = ymm4[1,3,2,3]
; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0],ymm10[1],ymm11[2,3,4],ymm10[5],ymm11[6,7]
+; AVX2-FAST-NEXT: vpblendvb %ymm12, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0],ymm6[1],ymm9[2,3,4],ymm6[5],ymm9[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2],xmm3[3],xmm4[4,5,6,7]
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2],ymm9[3],ymm6[4,5],ymm9[6],ymm6[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1,2],ymm15[3],ymm14[4,5],ymm15[6],ymm14[7]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm4 = ymm4[1,3,2,3]
; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0],ymm7[1],ymm8[2,3,4],ymm7[5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpblendvb %ymm12, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0],ymm0[1],ymm13[2,3,4],ymm0[5],ymm13[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2],xmm3[3],xmm4[4,5,6,7]
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm2
-; AVX2-FAST-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1,2],ymm15[3],mem[4,5],ymm15[6],mem[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1,2],ymm7[3],ymm8[4,5],ymm7[6],ymm8[7]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,3,2,3]
; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm1
-; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm2, %ymm1, %ymm0
+; AVX2-FAST-NEXT: vpblendvb %ymm12, %ymm2, %ymm1, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm8
-; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm11
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2],ymm8[3],ymm11[4,5],ymm8[6],ymm11[7]
+; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm11
+; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm10
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm11[3],ymm10[4,5],ymm11[6],ymm10[7]
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [3,6,2,5,3,6,2,5]
; AVX2-FAST-NEXT: # ymm1 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm0
@@ -11365,124 +11364,112 @@
; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm5 = ymm2[0,1,0,2]
-; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm13 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
-; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm5, %ymm4
+; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm9 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
+; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm5, %ymm4
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm4[7]
; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm0 = mem[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %ymm15
-; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %ymm14
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm15[3],ymm14[4,5],ymm15[6],ymm14[7]
+; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %ymm12
+; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm12[3],ymm0[4,5],ymm12[6],ymm0[7]
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm0
; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm0, %ymm0
; AVX2-FAST-NEXT: vmovdqa 640(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm2[0,1,0,2]
-; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm3, %ymm7
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm3, %ymm13
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm13[7]
; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm0 = mem[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm2[3],ymm4[4,5],ymm2[6],ymm4[7]
-; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm10
-; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm9
+; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm14
+; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm8
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm14[3],ymm8[4,5],ymm14[6],ymm8[7]
+; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm0
; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm0, %ymm0
; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,1,0,2]
-; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm2, %ymm7
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm2, %ymm15
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm15[7]
; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm0 = mem[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %ymm12
-; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %ymm7
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2],ymm12[3],ymm7[4,5],ymm12[6],ymm7[7]
+; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %ymm13
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm4[3],ymm13[4,5],ymm4[6],ymm13[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm7
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm0
; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm0, %ymm0
; AVX2-FAST-NEXT: vmovdqa 864(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,2]
-; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm1, %ymm6
+; AVX2-FAST-NEXT: vpshufb %ymm9, %ymm1, %ymm6
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm6[7]
; AVX2-FAST-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm0 = mem[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1],ymm11[2],ymm8[3,4,5],ymm11[6],ymm8[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1],ymm10[2],ymm11[3,4,5],ymm10[6],ymm11[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm6
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm6[4],xmm0[5],xmm6[6],xmm0[7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = [0,1,0,1,14,15,12,13,10,11,8,9,12,13,14,15]
; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm0, %xmm0
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm4 = [22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29]
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm15 = [22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29]
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm5, %ymm5
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm5[7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2,3,4,5,6,7],ymm5[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2,3,4,5,6,7],ymm4[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm14, %ymm13
-; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm14[2],ymm15[3,4,5],ymm14[6],ymm15[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm9
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1],ymm12[2],ymm9[3,4,5],ymm12[6],ymm9[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm5
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm5[4],xmm0[5],xmm5[6],xmm0[7]
; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm0, %xmm0
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm3, %ymm3
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4,5,6,7],ymm3[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1],ymm10[2],ymm9[3,4,5],ymm10[6],ymm9[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1],ymm8[2],ymm14[3,4,5],ymm8[6],ymm14[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm3[4],xmm0[5],xmm3[6],xmm0[7]
; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm0, %xmm0
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1],ymm7[2],ymm12[3,4,5],ymm7[6],ymm12[7]
+; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1],ymm13[2],ymm7[3,4,5],ymm13[6],ymm7[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4],xmm0[5],xmm2[6],xmm0[7]
; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm1
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm11[3],ymm8[4,5],ymm11[6],ymm8[7]
-; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm8, %ymm4
-; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm4
+; AVX2-FAST-NEXT: vmovdqa %ymm11, %ymm10
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2],ymm4[3],ymm11[4,5],ymm4[6],ymm11[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3,4,5],xmm0[6],xmm1[7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,0,1,14,15,12,13,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm2 = [2,5,2,5,2,5,2,5]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermd %ymm14, %ymm2, %ymm3
-; AVX2-FAST-NEXT: vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4,5,6,7],ymm3[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2],ymm13[3],ymm15[4,5],ymm13[6],ymm15[7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm3
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0],xmm0[1],xmm3[2,3,4,5],xmm0[6],xmm3[7]
-; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FAST-NEXT: vpermd %ymm5, %ymm2, %ymm3
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
@@ -11491,388 +11478,404 @@
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4,5,6,7],ymm3[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1,2],ymm7[3],ymm12[4,5],ymm7[6],ymm12[7]
-; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm13
-; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm7, %ymm8
-; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2],ymm12[3],ymm9[4,5],ymm12[6],ymm9[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm15
+; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm12
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0],xmm0[1],xmm3[2,3,4,5],xmm0[6],xmm3[7]
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermd %ymm12, %ymm2, %ymm3
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermd %ymm6, %ymm2, %ymm3
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4,5,6,7],ymm3[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2],ymm10[3],ymm9[4,5],ymm10[6],ymm9[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2],ymm13[3],ymm7[4,5],ymm13[6],ymm7[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm13, %ymm8
+; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0],xmm0[1],xmm3[2,3,4,5],xmm0[6],xmm3[7]
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermd %ymm6, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermd %ymm9, %ymm2, %ymm3
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1,2,3,4,5,6,7],ymm3[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm14, %ymm11
+; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2],ymm13[3],ymm14[4,5],ymm13[6],ymm14[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0],xmm0[1],xmm3[2,3,4,5],xmm0[6],xmm3[7]
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermd %ymm14, %ymm2, %ymm1
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,4,7,8,9,10,11,12,13,12,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0],ymm4[1],ymm11[2,3],ymm4[4],ymm11[5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0],ymm10[1],ymm4[2,3],ymm10[4],ymm4[5,6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,4,5,2,3,0,1,14,15,12,13,12,13,14,15]
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm14[0,1,1,3]
-; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm2 = [18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25]
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm4[7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2,3,4,5,6,7],ymm4[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0],ymm15[1],ymm7[2,3],ymm15[4],ymm7[5,6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm4
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1],xmm0[2,3,4,5],xmm4[6],xmm0[7]
-; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm4 = ymm5[0,1,1,3]
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm5
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm2 = ymm5[0,1,1,3]
+; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm3 = [18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25]
+; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm2, %ymm5
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm5[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2,3,4,5,6,7],ymm5[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm9[1],ymm10[2,3],ymm9[4],ymm10[5,6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm14
-; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm11
-; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm7
+; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0],ymm12[1],ymm15[2,3],ymm12[4],ymm15[5,6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm5
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm5[1],xmm0[2,3,4,5],xmm5[6],xmm0[7]
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm5
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm6[0,1,1,3]
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm6
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm5 = ymm6[0,1,1,3]
+; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm5, %ymm6
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm6[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm6[0],ymm5[1,2,3,4,5,6,7],ymm6[8],ymm5[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0],ymm13[1],ymm8[2,3],ymm13[4],ymm8[5,6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0],xmm6[1],xmm5[2,3,4,5],xmm6[6],xmm5[7]
-; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm5, %xmm1
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm0[1,2,3,4,5,6,7],ymm6[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0],ymm11[1],ymm13[2,3],ymm11[4],ymm13[5,6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm6
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm6[1],xmm0[2,3,4,5],xmm6[6],xmm0[7]
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm0
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm6
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm12 = ymm14[0,1,1,3]
+; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm12, %ymm15
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm11[0],ymm6[1,2,3,4,5,6,7],ymm11[8],ymm6[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm11[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0],ymm15[1],ymm8[2,3],ymm15[4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm11
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0],xmm11[1],xmm6[2,3,4,5],xmm11[6],xmm6[7]
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm6, %xmm1
; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm13 = ymm12[0,1,1,3]
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm13, %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3,4,5,6,7],ymm2[8],ymm1[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm11 = ymm9[0,1,1,3]
+; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm11, %ymm3
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm3[0],ymm1[1,2,3,4,5,6,7],ymm3[8],ymm1[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0],ymm15[1],ymm7[2,3,4],ymm15[5],ymm7[6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,6,7,4,5,2,3,0,1,14,15,12,13,14,15]
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm1, %xmm1
-; AVX2-FAST-NEXT: vmovdqa %xmm2, %xmm8
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0],ymm10[1],ymm4[2,3,4],ymm10[5],ymm4[6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm3[0],xmm1[1],xmm3[2],xmm1[3],xmm3[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,6,7,4,5,2,3,0,1,14,15,12,13,14,15]
+; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm1, %xmm1
; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm6 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
-; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm8 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
+; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1],mem[2],ymm1[3,4],mem[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm1, %xmm4
-; AVX2-FAST-NEXT: vmovdqa %xmm5, %xmm12
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm9[2],ymm10[3,4],ymm9[5],ymm10[6,7]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm1[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm1[0,1],mem[2],ymm1[3,4,5],mem[6],ymm1[7]
+; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm14[0,1],ymm13[2],ymm14[3,4,5],ymm13[6],ymm14[7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <0,3,7,2,6,u,u,u>
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm2[1,2,3,4,5,6,7],ymm4[8],ymm2[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm5[1],ymm10[2,3,4],ymm5[5],ymm10[6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm4
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0],xmm2[1],xmm4[2],xmm2[3],xmm4[4,5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm2, %xmm2
-; AVX2-FAST-NEXT: vmovdqa %xmm8, %xmm15
-; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm3, %ymm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm3, %xmm4
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1],ymm8[2],ymm9[3,4,5],ymm8[6],ymm9[7]
-; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermd %ymm6, %ymm1, %ymm6
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3,4,5,6,7],ymm3[8],ymm2[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0],ymm11[1],ymm14[2,3,4],ymm11[5],ymm14[6,7]
+; AVX2-FAST-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0],ymm7[1],mem[2,3,4],ymm7[5],mem[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3],xmm3[4,5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm2, %xmm2
+; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm2, %xmm2
; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27]
+; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm5, %ymm3
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1],ymm7[2],ymm8[3,4],ymm7[5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm3[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
+; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = ymm4[0,1],mem[2],ymm4[3,4,5],mem[6],ymm4[7]
+; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31>
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3,4,5,6,7],ymm3[8],ymm2[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0],ymm2[1],mem[2,3,4],ymm2[5],mem[6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0],xmm2[1],xmm3[2],xmm2[3],xmm3[4,5,6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm2, %xmm2
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm12[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $219, (%rsp), %ymm14, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm14[2],mem[3,4],ymm14[5],mem[6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm2, %xmm3
-; AVX2-FAST-NEXT: vmovdqa %xmm12, %xmm15
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $219, (%rsp), %ymm12, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm12[2],mem[3,4],ymm12[5],mem[6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm2, %xmm3
; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm2
; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm12[2],ymm11[3,4,5],ymm12[6],ymm11[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm3[0,1],mem[2],ymm3[3,4,5],mem[6],ymm3[7]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31>
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm3, %ymm3
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = mem[0],ymm15[1],mem[2,3,4],ymm15[5],mem[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2],xmm0[3],xmm2[4,5,6,7]
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,6,7,4,5,2,3,0,1,14,15,12,13,14,15]
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27]
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,20,21,26,27]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm2[0,1],mem[2],ymm2[3,4],mem[5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm2, %xmm3
-; AVX2-FAST-NEXT: vmovdqa %xmm15, %xmm13
+; AVX2-FAST-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm2[2],mem[3,4],ymm2[5],mem[6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm2, %xmm3
; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm2
; AVX2-FAST-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $187, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm3[2],mem[3,4,5],ymm3[6],mem[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $187, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm15[2],mem[3,4,5],ymm15[6],mem[7]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [0,3,3,3,0,3,7,7]
-; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm2 = [0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25]
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1],ymm5[2],ymm10[3,4],ymm5[5],ymm10[6,7]
-; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [2,5,1,4,2,5,1,4]
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermd %ymm0, %ymm3, %ymm5
-; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm0 = [30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25]
-; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm15
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7],ymm5[8,9,10,11,12],ymm4[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1],ymm7[2,3],ymm6[4,5],ymm7[6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm6, %xmm6
-; AVX2-FAST-NEXT: vmovdqa %xmm13, %xmm7
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm9[0,1,2],ymm8[3],ymm9[4,5],ymm8[6],ymm9[7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = <0,4,7,3,6,u,u,u>
-; AVX2-FAST-NEXT: vpermd %ymm10, %ymm9, %ymm13
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29>
-; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm13, %ymm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm6[0],ymm4[1,2,3,4,5,6,7],ymm6[8],ymm4[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermd %ymm8, %ymm1, %ymm4
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = ymm5[0,1],mem[2],ymm5[3,4],mem[5],ymm5[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm6, %ymm3, %ymm6
-; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm6, %ymm6
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm6[0,1,2,3,4],ymm4[5,6,7],ymm6[8,9,10,11,12],ymm4[13,14,15]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm13
-; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm13, %xmm13
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [0,3,3,3,0,3,7,7]
+; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm1
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm2[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm0[2],mem[3,4],ymm0[5],mem[6,7]
+; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm11 = [2,5,1,4,2,5,1,4]
+; AVX2-FAST-NEXT: # ymm11 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermd %ymm2, %ymm11, %ymm5
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31,u,u,u,u,u,u]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm3[5,6,7],ymm5[8,9,10,11,12],ymm3[13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm10[0,1],ymm9[2,3],ymm10[4,5],ymm9[6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm3
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm9 = xmm3[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1],xmm6[2],xmm13[2],xmm6[3],xmm13[3]
+; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm14[0,1,2],ymm13[3],ymm14[4,5],ymm13[6],ymm14[7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <0,4,7,3,6,u,u,u>
+; AVX2-FAST-NEXT: vpermd %ymm10, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm0[0],ymm5[1,2,3,4,5,6,7],ymm0[8],ymm5[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm3
+; AVX2-FAST-NEXT: vpermd %ymm13, %ymm1, %ymm0
+; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm2 = [0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25,0,0,16,17,22,23,24,25]
+; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm13 = ymm10[0,1,2],mem[3],ymm10[4,5],mem[6],ymm10[7]
-; AVX2-FAST-NEXT: vpermd %ymm13, %ymm9, %ymm13
-; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm13, %ymm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm6[0],ymm4[1,2,3,4,5,6,7],ymm6[8],ymm4[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm4
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1],ymm6[2],mem[3,4],ymm6[5],mem[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm6, %ymm3, %ymm6
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31,u,u,u,u,u,u]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm6[0,1,2,3,4],ymm4[5,6,7],ymm6[8,9,10,11,12],ymm4[13,14,15]
-; AVX2-FAST-NEXT: vpblendd $51, (%rsp), %ymm14, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1],ymm14[2,3],mem[4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm13
-; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm13, %xmm13
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1],ymm10[2],ymm6[3,4],ymm10[5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm5, %ymm11, %ymm5
+; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm1 = [30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25,30,31,16,17,22,23,24,25]
+; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0,1,2,3,4],ymm0[5,6,7],ymm5[8,9,10,11,12],ymm0[13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm9
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm9, %xmm9
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm7 = [10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm6, %xmm6
-; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1],xmm6[2],xmm13[2],xmm6[3],xmm13[3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm11[0,1,2],ymm12[3],ymm11[4,5],ymm12[6],ymm11[7]
-; AVX2-FAST-NEXT: vpermd %ymm13, %ymm9, %ymm13
-; AVX2-FAST-NEXT: vpshufb %ymm0, %ymm13, %ymm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm6[0],ymm4[1,2,3,4,5,6,7],ymm6[8],ymm4[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermd %ymm7, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm4[2],ymm6[3,4],ymm4[5],ymm6[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm2, %ymm3, %ymm0
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31,u,u,u,u,u,u]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm5, %xmm5
+; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm14[2,3],ymm12[4,5],ymm14[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm4[0,1,2],ymm14[3],ymm4[4,5],ymm14[6],ymm4[7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <0,4,7,3,6,u,u,u>
+; AVX2-FAST-NEXT: vpermd %ymm9, %ymm4, %ymm9
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1],ymm9[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2,3,4,5,6,7],ymm5[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = ymm5[0,1],mem[2],ymm5[3,4],mem[5],ymm5[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm5, %ymm11, %ymm5
+; AVX2-FAST-NEXT: vpshufb %ymm1, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0,1,2,3,4],ymm0[5,6,7],ymm5[8,9,10,11,12],ymm0[13,14,15]
+; AVX2-FAST-NEXT: vpblendd $51, (%rsp), %ymm12, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,1],ymm12[2,3],mem[4,5],ymm12[6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm9
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm9, %xmm9
+; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm5, %xmm5
+; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = ymm8[0,1,2],mem[3],ymm8[4,5],mem[6],ymm8[7]
+; AVX2-FAST-NEXT: vpermd %ymm9, %ymm4, %ymm9
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29>
+; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm9, %ymm9
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1],ymm9[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2,3,4,5,6,7],ymm5[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [0,3,3,3,0,3,7,7]
+; AVX2-FAST-NEXT: vpermd %ymm7, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpshufb %ymm2, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm12[2],ymm8[3,4],ymm12[5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm1, %ymm11, %ymm1
+; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,16,17,22,23,24,25,30,31,u,u,u,u,u,u]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7],ymm1[8,9,10,11,12],ymm0[13,14,15]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2],ymm11[3],ymm13[4,5],ymm11[6],ymm13[7]
-; AVX2-FAST-NEXT: vpermd %ymm2, %ymm9, %ymm2
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,2,3,4,5,10,11,12,13,18,19,18,19,18,19,18,19,18,19,20,21,26,27,28,29]
+; AVX2-FAST-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2],ymm15[3],mem[4,5],ymm15[6],mem[7]
+; AVX2-FAST-NEXT: vpermd %ymm2, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm5[0,1],mem[2,3],ymm5[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm10[2,3],ymm6[4,5],ymm10[6,7]
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [0,4,7,0,0,4,7,0]
; AVX2-FAST-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermd %ymm8, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm8 = [0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27]
-; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vpermd %ymm13, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm10 = [0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27,0,0,18,19,20,21,26,27]
+; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm2, %ymm2
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm9 = [2,6,1,5,2,6,1,5]
; AVX2-FAST-NEXT: # ymm9 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm9, %ymm1
-; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,26,27,28,29,u,u,u,u,u,u]
+; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm15 = [28,29,18,19,20,21,26,27,28,29,18,19,20,21,26,27,28,29,18,19,20,21,26,27,28,29,18,19,20,21,26,27]
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1,2,3,4],ymm2[5,6,7],ymm1[8,9,10,11,12],ymm2[13,14,15]
-; AVX2-FAST-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm15[0,1,2],mem[3],ymm15[4,5],mem[6],ymm15[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1,2],mem[3],ymm1[4,5],mem[6],ymm1[7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm3
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = [0,1,2,3,4,5,6,7,10,11,6,7,4,5,6,7]
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm3, %xmm3
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm11 = [0,1,2,3,4,5,6,7,10,11,6,7,4,5,6,7]
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm3, %xmm3
; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,4,6,7]
; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; AVX2-FAST-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0],ymm10[1],mem[2,3],ymm10[4],mem[5,6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <1,4,0,3,7,u,u,u>
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31>
-; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm14[0],mem[1],ymm14[2,3],mem[4],ymm14[5,6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = <1,4,0,3,7,u,u,u>
+; AVX2-FAST-NEXT: vpermd %ymm4, %ymm13, %ymm4
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,u,u,u,u,u,u,0,1,6,7,8,9,14,15,16,17,22,23,20,21,22,23,16,17,22,23,24,25,30,31>
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm4, %ymm4
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3,4,5,6,7],ymm3[8],ymm2[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm4[2,3],ymm6[4,5],ymm4[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm7, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm3, %ymm3
; AVX2-FAST-NEXT: vpermd %ymm2, %ymm9, %ymm2
-; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm4 = [28,29,18,19,20,21,26,27,28,29,18,19,20,21,26,27,28,29,18,19,20,21,26,27,28,29,18,19,20,21,26,27]
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7],ymm2[8,9,10,11,12],ymm3[13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0,1,2],ymm14[3],ymm12[4,5],ymm14[6],ymm12[7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm5
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm5, %xmm5
-; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,4,6,7]
-; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm11[0],ymm13[1],ymm11[2,3],ymm13[4],ymm11[5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
-; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0],ymm3[1],mem[2,3],ymm3[4],mem[5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5],mem[6],ymm4[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm5, %xmm5
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,4,6,7]
+; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; AVX2-FAST-NEXT: vpermd %ymm3, %ymm13, %ymm3
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3,4,5,6,7],ymm3[8],ymm2[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
-; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm1[0,1],mem[2,3],ymm1[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm4, %ymm4
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm9, %ymm3
-; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm3, %ymm3
-; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm11
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7],ymm3[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7],ymm3[8,9,10,11,12],ymm4[13,14,15]
+; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm1[0,1,2],mem[3],ymm1[4,5],mem[6],ymm1[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm5, %xmm5
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,4,6,7]
+; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0],ymm1[1],mem[2,3],ymm1[4],mem[5,6,7]
+; AVX2-FAST-NEXT: vpermd %ymm5, %ymm13, %ymm5
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0],ymm3[1,2,3,4,5,6,7],ymm4[8],ymm3[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vpermd %ymm7, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1],ymm12[2,3],ymm8[4,5],ymm12[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm4, %ymm9, %ymm4
+; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm4, %ymm4
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm0[5,6,7],ymm4[8,9,10,11,12],ymm0[13,14,15]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0],ymm4[1],mem[2,3],ymm4[4],mem[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2],ymm4[3],mem[4,5],ymm4[6],mem[7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm7
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm7, %xmm7
-; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,4,6,7]
-; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7]
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm5
-; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm5[0],ymm3[1,2,3,4,5,6,7],ymm5[8],ymm3[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm9, %ymm5
-; AVX2-FAST-NEXT: vpshufb %ymm11, %ymm5, %ymm5
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0,1,2,3,4],ymm0[5,6,7],ymm5[8,9,10,11,12],ymm0[13,14,15]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $183, (%rsp), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,1,2],ymm4[3],mem[4,5],ymm4[6],mem[7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm6, %xmm4
-; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,4,6,7]
-; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3],mem[4],ymm5[5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm5, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpshufb %ymm10, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1,2],ymm4[3],mem[4,5],ymm4[6],mem[7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm5, %xmm5
+; AVX2-FAST-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,4,6,7]
+; AVX2-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = ymm1[0],mem[1],ymm1[2,3],mem[4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vpermd %ymm5, %ymm13, %ymm1
+; AVX2-FAST-NEXT: vpshufb %ymm6, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -11913,9 +11916,9 @@
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%r9)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm1, (%r9)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%r9)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm1, (%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%rax)
@@ -11926,9 +11929,9 @@
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, (%rax)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovdqa %ymm0, 32(%rax)
-; AVX2-FAST-NEXT: vmovdqa %ymm3, (%rax)
-; AVX2-FAST-NEXT: vmovdqa %ymm2, 96(%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm0, 96(%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm3, 32(%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm2, (%rax)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 64(%rax)
; AVX2-FAST-NEXT: addq $1544, %rsp # imm = 0x608
@@ -11938,12 +11941,13 @@
; AVX2-FAST-PERLANE-LABEL: load_i16_stride7_vf64:
; AVX2-FAST-PERLANE: # %bb.0:
; AVX2-FAST-PERLANE-NEXT: subq $1528, %rsp # imm = 0x5F8
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm5
; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %ymm11
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 448(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 448(%rdi), %ymm9
; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm4
@@ -11951,7 +11955,7 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3,4],ymm0[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm1 = [0,1,14,15,12,13,14,15,2,3,6,7,12,13,2,3,16,17,30,31,28,29,30,31,18,19,22,23,28,29,18,19]
@@ -11961,206 +11965,204 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4],xmm0[5],xmm2[6],xmm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,14,15,12,13,10,11,8,9,8,9,8,9,8,9]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm0, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm14 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm4, %ymm3, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm13 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm4, %ymm3, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1],ymm11[2],ymm10[3,4,5],ymm11[6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm11[2],ymm9[3,4,5],ymm11[6],ymm9[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm7[1],ymm5[2,3,4],ymm7[5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm6[1],ymm5[2,3,4],ymm6[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm3, %ymm4, %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm15[2],ymm9[3,4,5],ymm15[6],ymm9[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm15, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm8[2],ymm15[3,4,5],ymm8[6],ymm15[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm13
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0],ymm8[1],ymm13[2,3,4],ymm8[5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm13, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm13, %ymm10
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 704(%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 672(%rdi), %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm3[2],ymm15[3,4,5],ymm3[6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 704(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 672(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0,1],ymm4[2],ymm12[3,4,5],ymm4[6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm13
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm4[4],xmm3[5],xmm4[6],xmm3[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 736(%rdi), %ymm13
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 768(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0],ymm13[1],ymm9[2,3,4],ymm13[5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 736(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 768(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0],ymm14[1],ymm4[2,3,4],ymm14[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm8
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm3, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm2, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm7[2],ymm5[3,4],ymm7[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm0[2],ymm5[3,4],ymm0[5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm1[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6,7,8,9,10],ymm2[11],ymm1[12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1,2],ymm11[3],ymm10[4,5],ymm11[6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1,2],ymm11[3],ymm9[4,5],ymm11[6],ymm9[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm2[1],xmm3[2,3,4,5],xmm2[6],xmm3[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = [2,3,2,3,2,3,2,3,8,9,8,9,6,7,4,5,18,19,18,19,18,19,18,19,24,25,24,25,22,23,20,21]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm4
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = [2,3,0,1,14,15,12,13,10,11,10,11,10,11,10,11]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm3, %ymm4, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1],ymm8[2],ymm6[3,4],ymm8[5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm3, %ymm4, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1],ymm7[2],ymm6[3,4],ymm7[5],ymm6[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3],ymm3[4,5,6,7,8,9,10],ymm4[11],ymm3[12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0,1,2],ymm12[3],ymm0[4,5],ymm12[6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $72, (%rsp), %ymm15, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm15[0,1,2],mem[3],ymm15[4,5],mem[6],ymm15[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm3, %ymm3
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm4, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm4, %ymm3, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm4, %ymm3, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm13[2],ymm9[3,4],ymm13[5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1],ymm14[2],ymm8[3,4],ymm14[5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm8, %ymm9
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3],ymm3[4,5,6,7,8,9,10],ymm4[11],ymm3[12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0,1,2],ymm13[3],ymm12[4,5],ymm13[6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm3, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm4, %ymm3, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm15[0,1,2],ymm11[3],ymm15[4,5],ymm11[6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm15, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0],xmm4[1],xmm5[2,3,4,5],xmm4[6],xmm5[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm3, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm4, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm4, %ymm3, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm6[2],ymm9[3,4],ymm6[5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm11[2],ymm5[3,4],ymm11[5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3],ymm3[4,5,6,7,8,9,10],ymm4[11],ymm3[12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm3, %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm7[3],ymm5[4,5],ymm7[6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1,2],ymm7[3],ymm8[4,5],ymm7[6],ymm8[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2,3,4,5],xmm3[6],xmm4[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm3, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm1, %ymm2, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm1, %ymm2, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm6[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1],ymm6[2,3],ymm9[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm11[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm11[2,3],ymm5[4,5],ymm11[6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4],ymm2[5,6,7,8,9,10,11],ymm1[12],ymm2[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0],ymm5[1],ymm7[2,3],ymm5[4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0],ymm8[1],ymm7[2,3],ymm8[4],ymm7[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0],xmm3[1],xmm1[2,3,4,5],xmm3[6],xmm1[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm1 = [4,5,4,5,4,5,4,5,8,9,10,11,8,9,6,7,20,21,20,21,20,21,20,21,24,25,26,27,24,25,22,23]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm2, %ymm4
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [4,5,2,3,0,1,14,15,12,13,12,13,12,13,12,13]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm14, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm3, %ymm4, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm3, %ymm4, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0],ymm14[1],ymm0[2,3],ymm14[4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0],ymm15[1],ymm0[2,3],ymm15[4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5],xmm4[6],xmm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1],ymm5[2,3],ymm8[4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm5[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1],ymm11[2,3],ymm8[4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm11[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7,8,9,10,11],ymm5[12],ymm4[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm3, %ymm4, %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0],ymm10[1],ymm15[2,3],ymm10[4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0],ymm12[1],ymm13[2,3],ymm12[4],ymm13[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5],xmm4[6],xmm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0,1],ymm13[2,3],ymm7[4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm13[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm9[0,1],ymm14[2,3],ymm9[4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm14[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4],ymm4[5,6,7,8,9,10,11],ymm5[12],ymm4[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm3, %ymm4, %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0],ymm10[1],ymm5[2,3],ymm10[4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0],ymm10[1],ymm9[2,3],ymm10[4],ymm9[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm4[1],xmm3[2,3,4,5],xmm4[6],xmm3[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1],ymm12[2,3],ymm13[4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm12[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm5[2,3],ymm7[4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm5[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4],ymm3[5,6,7,8,9,10,11],ymm4[12],ymm3[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm3, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm2, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $183, (%rsp), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2],ymm1[3],mem[4,5],ymm1[6],mem[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0,1,2],mem[3],ymm1[4,5],mem[6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm1[1,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm1 = [6,7,6,7,6,7,6,7,8,9,4,5,10,11,0,1,22,23,22,23,22,23,22,23,24,25,20,21,26,27,16,17]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm2, %ymm4
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [6,7,4,5,2,3,0,1,14,15,14,15,14,15,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm3, %ymm4, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm3, %ymm4, %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0],ymm14[1],ymm0[2,3,4],ymm14[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm0[0],ymm15[1],ymm0[2,3,4],ymm15[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2],xmm3[3],xmm4[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm8[0,1,2],mem[3],ymm8[4,5],mem[6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2],ymm11[3],ymm8[4,5],ymm11[6],ymm8[7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[1,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm3, %ymm4, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm3, %ymm4, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0],ymm10[1],ymm5[2,3,4],ymm10[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0],ymm10[1],ymm9[2,3,4],ymm10[5],ymm9[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2],xmm3[3],xmm4[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0,1,2],ymm12[3],ymm13[4,5],ymm12[6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0,1,2],ymm5[3],ymm7[4,5],ymm5[6],ymm7[7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm4[1,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm3, %ymm4, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0],ymm11[1],ymm15[2,3,4],ymm11[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm3, %ymm4, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0],ymm12[1],ymm13[2,3,4],ymm12[5],ymm13[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2],xmm3[3],xmm4[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm3, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm7[0,1,2],mem[3],ymm7[4,5],mem[6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm6[0,1,2],mem[3],ymm6[4,5],mem[6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[1,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm1, %ymm3, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm2, %ymm1, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm2, %ymm1, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm13
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm9[3],ymm13[4,5],ymm9[6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm9[3],ymm10[4,5],ymm9[6],ymm10[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = [0,1,2,3,4,5,6,7,8,9,10,11,6,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm1, %xmm1
@@ -12170,33 +12172,33 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm1[0,1,0,2]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm15 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm12, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm14 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm12, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2],ymm6[3],ymm7[4,5],ymm6[6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3],ymm0[4,5],ymm8[6],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 640(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm1[0,1,0,2]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm2, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 640(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm15[0,1,0,2]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm2, %ymm3
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm10
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm10[3],ymm5[4,5],ymm10[6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2],ymm6[3],ymm7[4,5],ymm6[6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
@@ -12205,33 +12207,34 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,2]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm1, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm1, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm13[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 800(%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 832(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3],ymm0[4,5],ymm3[6],ymm0[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 832(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm3[3],ymm5[4,5],ymm3[6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm13
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm11
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm11, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 864(%rdi), %ymm14
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm11 = ymm14[0,1,0,2]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm11, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 864(%rdi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm11 = ymm0[0,1,0,2]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm11, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm14[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd $31, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1],ymm7[2],ymm6[3,4,5],ymm7[6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm15
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm15[4],xmm4[5],xmm15[6],xmm4[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm15 = [0,1,0,1,14,15,12,13,10,11,8,9,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1],ymm3[2],ymm8[3,4,5],ymm3[6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm14
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm14[4],xmm4[5],xmm14[6],xmm4[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm14 = [0,1,0,1,14,15,12,13,10,11,8,9,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm4, %xmm4
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm0 = [22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29,22,23,28,29]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm2, %ymm2
@@ -12240,10 +12243,10 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm4[0],ymm2[1,2,3,4,5,6,7],ymm4[8],ymm2[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1],ymm5[2],ymm10[3,4,5],ymm5[6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm7[2],ymm6[3,4,5],ymm7[6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm4[4],xmm2[5],xmm4[6],xmm2[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm2, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5,6],ymm1[7]
@@ -12251,11 +12254,11 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3,4,5,6,7],ymm2[8],ymm1[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm3[2],ymm8[3,4,5],ymm3[6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1],ymm5[2],ymm13[3,4,5],ymm5[6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm8
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4],xmm1[5],xmm2[6],xmm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm11, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
@@ -12263,11 +12266,10 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3,4,5,6,7],ymm2[8],ymm1[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm13[2],ymm9[3,4,5],ymm13[6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm10[2],ymm9[3,4,5],ymm10[6],ymm9[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4],xmm1[5],xmm2[6],xmm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm14, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm12, %ymm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
@@ -12275,17 +12277,17 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2],ymm13[3],ymm9[4,5],ymm13[6],ymm9[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm13, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2],ymm10[3],ymm9[4,5],ymm10[6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3,4,5],xmm0[6],xmm1[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,0,1,14,15,12,13,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm5[0,1,1,2]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm6[0,1,1,2]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm4 = [16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31,16,17,30,31]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm2, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
@@ -12293,43 +12295,43 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2],ymm15[3],ymm6[4,5],ymm15[6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1,2],ymm3[3],ymm12[4,5],ymm3[6],ymm12[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm6[0,1,1,2]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm15[0,1,1,2]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm2, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm3[3],ymm8[4,5],ymm3[6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2],ymm8[3],ymm13[4,5],ymm8[6],ymm13[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm14[0,1,1,2]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm3[0,1,1,2]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm2, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm13[3],ymm10[4,5],ymm13[6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm7[3],ymm10[4,5],ymm7[6],ymm10[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3,4,5],xmm0[6],xmm2[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm3[0,1,1,2]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm7[0,1,1,2]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm4, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
@@ -12337,13 +12339,13 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0],ymm7[1],ymm9[2,3],ymm7[4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0],ymm9[1],ymm5[2,3],ymm9[4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3,4,5],xmm1[6],xmm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,4,5,2,3,0,1,14,15,12,13,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm5[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm6[0,1,1,3]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm2 = [18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25,18,19,24,25]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm4, %ymm4
@@ -12352,120 +12354,121 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0],ymm0[1,2,3,4,5,6,7],ymm4[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0],ymm12[1],ymm15[2,3],ymm12[4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm12, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0],ymm12[1],ymm14[2,3],ymm12[4],ymm14[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm4[1],xmm0[2,3,4,5],xmm4[6],xmm0[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm6[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm15[0,1,1,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm4, %ymm6
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm0[1,2,3,4,5,6,7],ymm6[8],ymm0[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0],ymm10[1],ymm13[2,3],ymm10[4],ymm13[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm6[1],xmm0[2,3,4,5],xmm6[6],xmm0[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm13 = ymm3[0,1,1,3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm13, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2,3,4,5,6,7],ymm5[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0],ymm8[1],ymm11[2,3],ymm8[4],ymm11[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0],ymm10[1],ymm11[2,3],ymm10[4],ymm11[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, %ymm11
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm6[1],xmm0[2,3,4,5],xmm6[6],xmm0[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm14[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm7[0,1,1,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm7, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm5[0],ymm0[1,2,3,4,5,6,7],ymm5[8],ymm0[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0],ymm13[1],ymm8[2,3],ymm13[4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm6[1],xmm0[2,3,4,5],xmm6[6],xmm0[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm3[0,1,1,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1,2,3,4,5,6,7],ymm2[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm8[2],ymm5[3,4,5],ymm8[6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm9[0,1],mem[2],ymm9[3,4,5],mem[6],ymm9[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm0 = <u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27>
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm2, %ymm3
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7,8,9,10,11,12,13],ymm2[14],ymm3[15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0,1],ymm7[2],ymm12[3,4],ymm7[5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm8[2],ymm5[3,4],ymm8[5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm6 = xmm3[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0],ymm9[1],ymm15[2,3,4],ymm9[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0],ymm12[1],ymm14[2,3,4],ymm12[5],ymm14[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm6[0],xmm3[1],xmm6[2],xmm3[3],xmm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm6 = [0,1,6,7,4,5,2,3,0,1,14,15,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm14 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd {{.*#+}} ymm13 = [20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27,20,21,26,27]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm13, %ymm4, %ymm4
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm2[0],ymm3[1,2,3,4,5,6,7],ymm2[8],ymm3[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1],ymm11[2],ymm10[3,4,5],ymm11[6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm10[2],ymm2[3,4,5],ymm10[6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm0, %ymm2, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, %ymm12
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7,8,9,10,11,12,13],ymm2[14],ymm3[15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm9[2],ymm15[3,4],ymm9[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm14[0,1],mem[2],ymm14[3,4],mem[5],ymm14[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0],ymm3[1],mem[2,3,4],ymm3[5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0],ymm11[1],ymm15[2,3,4],ymm11[5],ymm15[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2],xmm3[3],xmm4[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm13, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm13, %ymm7, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm13, %ymm15
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm2[0],ymm3[1,2,3,4,5,6,7],ymm2[8],ymm3[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm13[0,1],mem[2],ymm13[3,4,5],mem[6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm3 = ymm2[u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm2[0,1],mem[2],ymm2[3,4,5],mem[6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm12, %ymm2, %ymm3
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7,8,9,10,11,12,13],ymm2[14],ymm3[15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1],ymm3[2],mem[3,4],ymm3[5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm7[2],ymm11[3,4],ymm7[5],ymm11[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0],ymm3[1],mem[2,3,4],ymm3[5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm13[0],mem[1],ymm13[2,3,4],mem[5],ymm13[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm3, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1],xmm4[2],xmm3[3],xmm4[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1,2,3,4,5,6,7],ymm2[8],ymm1[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
@@ -12477,18 +12480,18 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3],xmm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm2, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm2, %ymm2
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $187, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm2[2],mem[3,4,5],ymm2[6],mem[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm0 = ymm2[u,u,u,u,u,u,u,u,0,1,14,15,u,u,10,11,24,25,24,25,24,25,24,25,16,17,30,31,u,u,26,27]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm15[0,1],mem[2],ymm15[3,4,5],mem[6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm12, %ymm2, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6],ymm0[7,8,9,10,11,12,13],ymm2[14],ymm0[15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0,1],ymm4[2],ymm14[3,4],ymm4[5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm4[0,1],mem[2],ymm4[3,4],mem[5],ymm4[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm6 = [8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm2, %xmm3
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm2
@@ -12498,238 +12501,242 @@
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3,4,5,6,7],ymm0[8],ymm1[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm8[3],ymm5[4,5],ymm8[6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm9[0,1,2],mem[3],ymm9[4,5],mem[6],ymm9[7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7,8],ymm0[9,10,11,12,13,14],ymm1[15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm7[2,3],ymm12[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm8[2,3],ymm5[4,5],ymm8[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm2, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm6, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1],ymm12[2],mem[3,4],ymm12[5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm2, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm6, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0,1],mem[2],ymm0[3,4],mem[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3],xmm0[4],xmm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = [8,9,8,9,6,7,4,5,2,3,2,3,2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm1, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT: vmovdqa 656(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 640(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm6 = xmm3[0,1,2,3,4,5],xmm1[6],xmm3[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm9 = xmm3[0,1,2,3,4,5],xmm1[6],xmm3[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,4,5,6,7,0,1,0,1,14,15,12,13]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm6, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7],ymm5[8,9,10,11,12],ymm6[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm2[0],ymm5[1,2,3,4,5,6,7],ymm2[8],ymm5[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1,2],ymm11[3],ymm10[4,5],ymm11[6],ymm10[7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm2[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3,4,5,6],ymm5[7,8],ymm2[9,10,11,12,13,14],ymm5[15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm15[0,1],ymm9[2,3],ymm15[4,5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm6, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm7, %xmm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm11 = [10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm8, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $219, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1],ymm3[2],mem[3,4],ymm3[5],mem[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1],xmm5[2],xmm6[3],xmm5[4],xmm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 432(%rdi), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5],xmm3[6],xmm6[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm6, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7],ymm5[8,9,10,11,12],ymm6[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm2[0],ymm5[1,2,3,4,5,6,7],ymm2[8],ymm5[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2],ymm9[3],ymm13[4,5],ymm9[6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm2[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm5[0],ymm2[1,2,3,4,5,6],ymm5[7,8],ymm2[9,10,11,12,13,14],ymm5[15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0,1],ymm10[2,3],ymm7[4,5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm6, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm8, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1],ymm8[2],ymm3[3,4],ymm8[5],ymm3[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm6[0,1],xmm2[2],xmm6[3],xmm2[4],xmm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm2, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm13
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 880(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 864(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm15 = xmm6[0,1,2,3,4,5],xmm2[6],xmm6[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm15, %xmm15
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm15, %ymm0, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm15[5,6,7],ymm13[8,9,10,11,12],ymm15[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm13 = ymm5[0],ymm13[1,2,3,4,5,6,7],ymm5[8],ymm13[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm9, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm9[5,6,7],ymm6[8,9,10,11,12],ymm9[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm6 = ymm5[0],ymm6[1,2,3,4,5,6,7],ymm5[8],ymm6[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,2],ymm10[3],mem[4,5],ymm10[6],mem[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm5[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm6[0],ymm5[1,2,3,4,5,6],ymm6[7,8],ymm5[9,10,11,12,13,14],ymm6[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm14[0,1],mem[2,3],ymm14[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm9, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm6, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm8, %ymm5, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $36, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm10[0,1],mem[2],ymm10[3,4],mem[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm9[0,1],xmm5[2],xmm9[3],xmm5[4],xmm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 432(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm14 = xmm5[0,1,2,3,4,5],xmm8[6],xmm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm14, %xmm14
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm14[5,6,7],ymm9[8,9,10,11,12],ymm14[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm9 = ymm6[0],ymm9[1,2,3,4,5,6,7],ymm6[8],ymm9[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm8[0,1,2],mem[3],ymm8[4,5],mem[6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm9 = ymm6[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm6 = ymm9[0],ymm6[1,2,3,4,5,6],ymm9[7,8],ymm6[9,10,11,12,13,14],ymm9[15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0,1],ymm7[2,3],ymm11[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm9, %xmm14
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm14, %xmm14
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm9, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm9[0],xmm14[0],xmm9[1],xmm14[1],xmm9[2],xmm14[2],xmm9[3],xmm14[3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm6, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0,1],ymm6[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm11[0,1],ymm15[2],ymm11[3,4],ymm15[5],ymm11[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm13
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm13[0,1],xmm5[2],xmm13[3],xmm5[4],xmm13[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm5, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 208(%rdi), %xmm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm0[0,1,2,3,4,5],xmm5[6],xmm0[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm5, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm9 = ymm13[0,1],ymm11[2],ymm13[3,4],ymm11[5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm9, %xmm14
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm9 = xmm14[0,1],xmm9[2],xmm14[3],xmm9[4],xmm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm9, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 880(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 864(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm14 = xmm12[0,1,2,3,4,5],xmm7[6],xmm12[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm14, %xmm14
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm14[5,6,7],ymm9[8,9,10,11,12],ymm14[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm9 = ymm6[0],ymm9[1,2,3,4,5,6,7],ymm6[8],ymm9[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm6 = ymm12[0,1],ymm13[2],ymm12[3,4],ymm13[5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm6, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm6 = xmm9[0,1],xmm6[2],xmm9[3],xmm6[4],xmm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm6, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 208(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm6 = xmm7[0,1,2,3,4,5],xmm6[6],xmm7[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm6, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7],ymm0[8,9,10,11,12],ymm1[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1],ymm4[2,3],ymm14[4,5],ymm4[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[10,11,6,7,4,5,6,7,8,9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,2],ymm2[3],mem[4,5],ymm2[6],mem[7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm5[0],ymm4[1,2,3,4,5,6],ymm5[7,8],ymm4[9,10,11,12,13,14],ymm5[15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm4 = ymm4[u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm4[0,1],mem[2,3],ymm4[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm6 = xmm6[8,9,4,5,4,5,6,7,8,9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1,2],ymm14[3],ymm15[4,5],ymm14[6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm2[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm6[0],ymm2[1,2,3,4,5,6],ymm6[7,8],ymm2[9,10,11,12,13,14],ymm6[15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[u,u,u,u,u,u,u,u,2,3,0,1,14,15,12,13,26,27,26,27,26,27,26,27,18,19,16,17,30,31,28,29]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6,7],ymm1[8],ymm0[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0],ymm0[1],mem[2,3],ymm0[4],mem[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7,8],ymm1[9],ymm0[10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2],ymm1[3],mem[4,5],ymm1[6],mem[7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,7,10,11,6,7,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm12[2,3],mem[4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm4[0,1,2],xmm1[3],xmm4[4],xmm1[5],xmm4[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm5 = xmm2[0],mem[1],xmm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = [8,9,10,11,8,9,6,7,4,5,4,5,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm1, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm13 = [0,1,2,3,4,5,6,7,0,1,2,3,0,1,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm5[5,6,7],ymm1[8,9,10,11,12],ymm5[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3,4,5,6,7],ymm0[8],ymm1[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm9[0],mem[1],ymm9[2,3],mem[4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7,8],ymm1[9],ymm0[10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1,2],ymm10[3],ymm7[4,5],ymm10[6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm9 = [0,1,2,3,4,5,6,7,10,11,6,7,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm7, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm2, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,4,5,2,3,0,1,14,15,28,29,28,29,28,29,28,29,20,21,18,19,16,17,30,31>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1],ymm8[2,3],ymm3[4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm7[0,1,2],xmm1[3],xmm7[4],xmm1[5],xmm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm2 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm2 = mem[0],xmm6[1],mem[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0,1],mem[2,3],ymm1[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3],xmm2[4],xmm1[5],xmm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm4 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm4 = mem[0],xmm2[1],mem[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm7 = [8,9,10,11,8,9,6,7,4,5,4,5,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm2, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7],ymm1[8,9,10,11,12],ymm2[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,6,7,0,1,2,3,0,1,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5,6,7],ymm1[8,9,10,11,12],ymm4[13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3,4,5,6,7],ymm0[8],ymm1[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3],mem[4],ymm0[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4,5,6,7,8],ymm2[9],ymm0[10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2],ymm2[3],mem[4,5],ymm2[6],mem[7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm6, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm6[0,1,2],xmm2[3],xmm6[4],xmm2[5],xmm6[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm6 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm6 = mem[0],xmm3[1],mem[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm2, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm6, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm6[5,6,7],ymm2[8,9,10,11,12],ymm6[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm0[0],ymm2[1,2,3,4,5,6,7],ymm0[8],ymm2[9,10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1],ymm15[2,3],ymm11[4,5],ymm15[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm2 = xmm6[0,1,2],xmm2[3],xmm6[4],xmm2[5],xmm6[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm4, %xmm2, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm4 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm4 = mem[0],xmm3[1],mem[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm4, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7],ymm2[8,9,10,11,12],ymm3[13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3],mem[4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm3[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2,3,4,5,6,7,8],ymm4[9],ymm3[10,11,12,13,14,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm14, %ymm3, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3],mem[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4,5,6,7,8],ymm4[9],ymm1[10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm4[0,1,2],mem[3],ymm4[4,5],mem[6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $183, (%rsp), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,2],ymm4[3],mem[4,5],ymm4[6],mem[7]
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm6, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm4[4],xmm6[4],xmm4[5],xmm6[5],xmm4[6],xmm6[6],xmm4[7],xmm6[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm10[0,1],mem[2,3],ymm10[4,5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm4 = xmm6[0,1,2],xmm4[3],xmm6[4],xmm4[5],xmm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm5 = mem[0],xmm5[1],mem[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7],ymm4[8,9,10,11,12],ymm5[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm1[0],ymm4[1,2,3,4,5,6,7],ymm1[8],ymm4[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0],ymm8[1],mem[2,3],ymm8[4],mem[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm4[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3,4,5,6,7,8],ymm5[9],ymm4[10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $72, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm3[0,1,2],mem[3],ymm3[4,5],mem[6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm6, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1],ymm11[2,3],mem[4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3],xmm6[4],xmm5[5],xmm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm6 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm6 = xmm3[0],mem[1],xmm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm6, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm0, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5,6,7],ymm5[8,9,10,11,12],ymm6[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm5 = ymm4[0],ymm5[1,2,3,4,5,6,7],ymm4[8],ymm5[9,10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm5 = ymm12[0,1],ymm13[2,3],ymm12[4,5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1,2],xmm5[3],xmm6[4],xmm5[5],xmm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm5, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendw $253, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm5 = mem[0],xmm5[1],mem[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm5, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7],ymm3[8,9,10,11,12],ymm2[13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm14[0],mem[1],ymm14[2,3],mem[4],ymm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm5 = ymm3[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2,3,4,5,6,7,8],ymm5[9],ymm3[10,11,12,13,14,15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm15, %ymm3, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd $183, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,2],ymm5[3],mem[4,5],ymm5[6],mem[7]
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm6, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3,4,5,6,7],ymm3[8],ymm2[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -12769,9 +12776,9 @@
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%r9)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 96(%rax)
@@ -12782,10 +12789,10 @@
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rax)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, 96(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, 32(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, (%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, 96(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, 64(%rax)
; AVX2-FAST-PERLANE-NEXT: addq $1528, %rsp # imm = 0x5F8
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
index e88f9e1..7032733 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
@@ -1706,7 +1706,7 @@
; SSE-NEXT: movdqa 496(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 480(%rdi), %xmm14
-; SSE-NEXT: movdqa %xmm14, (%rsp) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 208(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 192(%rdi), %xmm3
@@ -1720,7 +1720,7 @@
; SSE-NEXT: movdqa 128(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 176(%rdi), %xmm7
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm7, (%rsp) # 16-byte Spill
; SSE-NEXT: movdqa 160(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
@@ -1918,7 +1918,7 @@
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm11[2,3]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: punpckhwd (%rsp), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
@@ -1927,151 +1927,154 @@
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT: # xmm9 = xmm9[4],mem[4],xmm9[5],mem[5],xmm9[6],mem[6],xmm9[7],mem[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm14, %xmm0
-; SSE-NEXT: movdqa %xmm14, %xmm11
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: movdqa %xmm3, %xmm10
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = xmm8[4],mem[4],xmm8[5],mem[5],xmm8[6],mem[6],xmm8[7],mem[7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT: # xmm14 = xmm14[4],mem[4],xmm14[5],mem[5],xmm14[6],mem[6],xmm14[7],mem[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm14[0,0,0,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[0,0,0,0]
; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm3, %xmm12
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; SSE-NEXT: movdqa %xmm9, %xmm0
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT: # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm6[0,0,0,0]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = xmm8[4],mem[4],xmm8[5],mem[5],xmm8[6],mem[6],xmm8[7],mem[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[0,0,0,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,0,0]
+; SSE-NEXT: movdqa %xmm4, %xmm7
; SSE-NEXT: punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm3, %xmm2
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movdqa %xmm6, %xmm0
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: # xmm7 = xmm7[4],mem[4],xmm7[5],mem[5],xmm7[6],mem[6],xmm7[7],mem[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,0,0,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = xmm9[4],mem[4],xmm9[5],mem[5],xmm9[6],mem[6],xmm9[7],mem[7]
-; SSE-NEXT: movdqa %xmm9, %xmm4
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT: # xmm12 = xmm12[4],mem[4],xmm12[5],mem[5],xmm12[6],mem[6],xmm12[7],mem[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; SSE-NEXT: movdqa %xmm0, %xmm4
+; SSE-NEXT: movdqa %xmm0, %xmm13
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1]
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[2,3]
; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[1,1,1,1]
-; SSE-NEXT: movdqa %xmm10, %xmm13
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT: movdqa %xmm14, %xmm11
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[1,1,1,1]
+; SSE-NEXT: movdqa %xmm3, %xmm15
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: movaps %xmm15, %xmm0
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movaps %xmm14, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm14, %xmm0
-; SSE-NEXT: movdqa (%rsp), %xmm12 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT: movdqa %xmm7, %xmm5
+; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[1,1,1,1]
-; SSE-NEXT: movdqa %xmm2, %xmm5
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT: movdqa %xmm6, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[2,3]
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm9, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm9[1,1,1,1]
; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm9[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT: movdqa %xmm7, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,1],xmm0[2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, %xmm0
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
+; SSE-NEXT: movaps %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm13[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movaps %xmm2, %xmm0
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm0[2,3]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
-; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm15[2,2,2,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm14[2,2,2,2]
; SSE-NEXT: punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm0[2],xmm10[3],xmm0[3]
; SSE-NEXT: movdqa %xmm11, %xmm0
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm13[2],xmm0[3],xmm13[3]
-; SSE-NEXT: movdqa %xmm13, %xmm15
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,2,2,2]
-; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm6[2,2,2,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm5[2,2,2,2]
+; SSE-NEXT: movdqa %xmm5, %xmm8
; SSE-NEXT: punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm0[2],xmm11[3],xmm0[3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm5[0],xmm11[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[2,2,2,2]
-; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm14[2,2,2,2]
+; SSE-NEXT: movdqa %xmm6, %xmm5
+; SSE-NEXT: movdqa %xmm6, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm6[2],xmm0[3],xmm6[3]
+; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,2,2,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm3[2,2,2,2]
; SSE-NEXT: punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm0[2],xmm12[3],xmm0[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm5, %xmm0
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
-; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[2,2,2,2]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm4[2],xmm9[3],xmm4[3]
+; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm9[0],xmm12[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,2,2,2]
+; SSE-NEXT: movaps %xmm13, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[2,2,2,2]
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm4, %xmm13
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm3[2],xmm13[3],xmm3[3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, %xmm13
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: unpckhps {{.*#+}} xmm13 = xmm13[2],xmm3[2],xmm13[3],xmm3[3]
; SSE-NEXT: shufps {{.*#+}} xmm13 = xmm13[0,1],xmm1[2,3]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm14, %xmm1
; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
@@ -2079,28 +2082,27 @@
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,1],xmm1[2,3]
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[3,3,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT: # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,3,3,3]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,3,3,3]
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm6[2,3]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: unpckhps (%rsp), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT: # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[3,3,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3]
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,3,3,3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: unpckhps {{.*#+}} xmm6 = xmm6[2],xmm9[2],xmm6[3],xmm9[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm3[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT: movaps %xmm6, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
@@ -2142,13 +2144,13 @@
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 16(%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT: movaps (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 48(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 32(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 48(%rax)
-; SSE-NEXT: movaps %xmm9, (%rax)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 16(%rax)
+; SSE-NEXT: movaps %xmm7, (%rax)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT: movapd %xmm12, 48(%rax)
; SSE-NEXT: movapd %xmm11, 32(%rax)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll
index b420710..f39e643 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-2.ll
@@ -182,30 +182,30 @@
; SSE-NEXT: movaps 16(%rdi), %xmm1
; SSE-NEXT: movaps 32(%rdi), %xmm2
; SSE-NEXT: movaps 48(%rdi), %xmm3
-; SSE-NEXT: movaps 80(%rdi), %xmm4
-; SSE-NEXT: movaps 64(%rdi), %xmm5
-; SSE-NEXT: movaps 112(%rdi), %xmm6
-; SSE-NEXT: movaps 96(%rdi), %xmm7
+; SSE-NEXT: movaps 112(%rdi), %xmm4
+; SSE-NEXT: movaps 96(%rdi), %xmm5
+; SSE-NEXT: movaps 80(%rdi), %xmm6
+; SSE-NEXT: movaps 64(%rdi), %xmm7
; SSE-NEXT: movaps %xmm7, %xmm8
; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[0,2],xmm6[0,2]
; SSE-NEXT: movaps %xmm5, %xmm9
; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,2],xmm4[0,2]
-; SSE-NEXT: movaps %xmm2, %xmm10
-; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,2],xmm3[0,2]
-; SSE-NEXT: movaps %xmm0, %xmm11
-; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,2],xmm1[0,2]
+; SSE-NEXT: movaps %xmm0, %xmm10
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,2],xmm1[0,2]
+; SSE-NEXT: movaps %xmm2, %xmm11
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,2],xmm3[0,2]
; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[1,3],xmm6[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,3],xmm4[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,3],xmm3[1,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
-; SSE-NEXT: movaps %xmm9, 32(%rsi)
-; SSE-NEXT: movaps %xmm8, 48(%rsi)
-; SSE-NEXT: movaps %xmm11, (%rsi)
-; SSE-NEXT: movaps %xmm10, 16(%rsi)
-; SSE-NEXT: movaps %xmm5, 32(%rdx)
-; SSE-NEXT: movaps %xmm7, 48(%rdx)
-; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,3],xmm3[1,3]
+; SSE-NEXT: movaps %xmm11, 16(%rsi)
+; SSE-NEXT: movaps %xmm10, (%rsi)
+; SSE-NEXT: movaps %xmm9, 48(%rsi)
+; SSE-NEXT: movaps %xmm8, 32(%rsi)
; SSE-NEXT: movaps %xmm2, 16(%rdx)
+; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: movaps %xmm5, 48(%rdx)
+; SSE-NEXT: movaps %xmm7, 32(%rdx)
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i32_stride2_vf16:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
index 2829c15..7d3f515 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
@@ -732,223 +732,219 @@
; SSE-LABEL: load_i32_stride3_vf32:
; SSE: # %bb.0:
; SSE-NEXT: subq $360, %rsp # imm = 0x168
-; SSE-NEXT: movaps 192(%rdi), %xmm3
-; SSE-NEXT: movaps 224(%rdi), %xmm2
-; SSE-NEXT: movaps 208(%rdi), %xmm13
-; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 240(%rdi), %xmm7
-; SSE-NEXT: movaps 272(%rdi), %xmm5
+; SSE-NEXT: movaps 336(%rdi), %xmm13
+; SSE-NEXT: movaps 368(%rdi), %xmm11
+; SSE-NEXT: movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 352(%rdi), %xmm5
; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 256(%rdi), %xmm9
-; SSE-NEXT: movaps (%rdi), %xmm12
-; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 16(%rdi), %xmm10
-; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 32(%rdi), %xmm4
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 48(%rdi), %xmm8
-; SSE-NEXT: movaps 80(%rdi), %xmm1
-; SSE-NEXT: movaps 64(%rdi), %xmm6
-; SSE-NEXT: movaps %xmm6, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT: movaps %xmm1, %xmm15
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm8, %xmm1
-; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm9, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm5[1,0]
-; SSE-NEXT: movaps %xmm7, %xmm1
-; SSE-NEXT: movaps %xmm7, %xmm11
-; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm10, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[1,0]
-; SSE-NEXT: movaps %xmm12, %xmm1
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm13, %xmm0
-; SSE-NEXT: movaps %xmm2, %xmm4
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[1,0]
-; SSE-NEXT: movaps %xmm3, %xmm2
-; SSE-NEXT: movaps %xmm3, %xmm14
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 176(%rdi), %xmm10
+; SSE-NEXT: movaps 144(%rdi), %xmm4
+; SSE-NEXT: movaps 176(%rdi), %xmm2
; SSE-NEXT: movaps 160(%rdi), %xmm12
-; SSE-NEXT: movaps %xmm12, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm10[1,0]
-; SSE-NEXT: movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 144(%rdi), %xmm5
-; SSE-NEXT: movaps %xmm5, %xmm2
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 368(%rdi), %xmm1
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 352(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 288(%rdi), %xmm14
+; SSE-NEXT: movaps 320(%rdi), %xmm3
+; SSE-NEXT: movaps 304(%rdi), %xmm9
+; SSE-NEXT: movaps 96(%rdi), %xmm6
+; SSE-NEXT: movaps 128(%rdi), %xmm1
+; SSE-NEXT: movaps 112(%rdi), %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT: movaps 336(%rdi), %xmm7
-; SSE-NEXT: movaps %xmm7, %xmm2
-; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 128(%rdi), %xmm1
-; SSE-NEXT: movaps %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT: movaps 112(%rdi), %xmm3
-; SSE-NEXT: movaps %xmm3, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT: movaps 96(%rdi), %xmm13
-; SSE-NEXT: movaps %xmm13, %xmm2
-; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 320(%rdi), %xmm2
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 304(%rdi), %xmm1
-; SSE-NEXT: movaps %xmm1, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[1,0]
-; SSE-NEXT: movaps 288(%rdi), %xmm2
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[1,0],xmm6[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[3,1],xmm15[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[0,2],xmm6[0,2]
-; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm11, %xmm2
-; SSE-NEXT: movaps %xmm9, %xmm0
-; SSE-NEXT: movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm9[0,0]
-; SSE-NEXT: shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[3,1],mem[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm0[0,2]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm14, %xmm11
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[1,0],xmm0[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm4[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[0,2],xmm0[0,2]
-; SSE-NEXT: movaps %xmm5, %xmm9
-; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[1,0],xmm12[0,0]
-; SSE-NEXT: movaps %xmm12, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm10[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,2],xmm0[0,2]
-; SSE-NEXT: movaps %xmm7, %xmm14
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[1,0],xmm0[0,0]
-; SSE-NEXT: shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[3,1],mem[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,2],xmm0[0,2]
-; SSE-NEXT: movaps %xmm13, %xmm10
-; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[1,0],xmm3[0,0]
-; SSE-NEXT: movaps %xmm3, %xmm0
-; SSE-NEXT: movaps (%rsp), %xmm15 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm15[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,2],xmm0[0,2]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm1[0,0]
-; SSE-NEXT: shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[3,1],mem[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm1[2,3,2,3]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,0],xmm4[0,0]
-; SSE-NEXT: movaps %xmm4, %xmm7
-; SSE-NEXT: movaps %xmm4, %xmm8
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[3,1],xmm13[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,2],xmm7[0,2]
+; SSE-NEXT: movaps %xmm1, %xmm8
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: # xmm7 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
-; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = xmm4[0,1],mem[0,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm8[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
-; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm13[0,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm12[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
-; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = xmm5[0,1],mem[0,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm3[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm15[0,3]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = mem[1,1,1,1]
+; SSE-NEXT: movaps %xmm6, %xmm1
+; SSE-NEXT: movaps %xmm6, %xmm10
+; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm9, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[1,0]
+; SSE-NEXT: movaps %xmm3, %xmm6
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm14, %xmm1
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm12, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[1,0]
+; SSE-NEXT: movaps %xmm2, %xmm12
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm4, %xmm1
+; SSE-NEXT: movaps %xmm4, %xmm15
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm11[1,0]
+; SSE-NEXT: movaps %xmm13, %xmm1
+; SSE-NEXT: movaps %xmm13, %xmm11
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm5[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 16(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 32(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
+; SSE-NEXT: movaps (%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 224(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 208(%rdi), %xmm7
+; SSE-NEXT: movaps %xmm7, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
+; SSE-NEXT: movaps 192(%rdi), %xmm3
+; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 80(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 64(%rdi), %xmm0
+; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
+; SSE-NEXT: movaps 48(%rdi), %xmm13
+; SSE-NEXT: movaps %xmm13, %xmm1
+; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 272(%rdi), %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps 256(%rdi), %xmm2
+; SSE-NEXT: movaps %xmm2, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
+; SSE-NEXT: movaps 240(%rdi), %xmm4
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,3],xmm0[0,2]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm10, %xmm4
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm0[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm8[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,2]
+; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm9, (%rsp) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[1,0],xmm9[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[3,1],xmm6[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,2],xmm9[0,2]
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm15, %xmm9
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[1,0],xmm0[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm12[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,2],xmm0[0,2]
+; SSE-NEXT: movaps %xmm11, %xmm12
+; SSE-NEXT: movaps %xmm11, %xmm8
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[1,0],xmm0[0,0]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm11[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[0,2],xmm0[0,2]
+; SSE-NEXT: movaps %xmm3, %xmm10
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[1,0],xmm7[0,0]
+; SSE-NEXT: movaps %xmm7, %xmm0
+; SSE-NEXT: shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,2],xmm0[0,2]
+; SSE-NEXT: movaps %xmm13, %xmm15
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[1,0],xmm0[0,0]
+; SSE-NEXT: shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[0,2],xmm0[0,2]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: movaps %xmm13, %xmm5
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,0],xmm2[0,0]
+; SSE-NEXT: movaps %xmm2, %xmm0
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[0,2]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm14[0,0]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[3,1],xmm0[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,2],xmm14[0,2]
+; SSE-NEXT: pshufd $85, (%rsp), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT: # xmm14 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm14[0],xmm6[1],xmm14[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = xmm6[0,1],mem[0,3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT: # xmm14 = mem[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm12[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm14[0],xmm12[1],xmm14[1]
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[0,1],xmm11[0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm7[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm14[0],xmm3[1],xmm14[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = xmm3[0,1],mem[0,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm2[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm13[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
+; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT: # xmm11 = mem[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT: # xmm7 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm11[0],xmm7[1],xmm11[1]
; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT: # xmm7 = xmm7[0,1],mem[0,3]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT: # xmm13 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm8[0],xmm13[1],xmm8[1]
-; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT: # xmm13 = xmm13[0,1],mem[0,3]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT: # xmm14 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT: # xmm11 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm14[0],xmm11[1],xmm14[1]
+; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT: # xmm11 = xmm11[0,1],mem[0,3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT: # xmm14 = mem[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
-; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[0,1],mem[0,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm0[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT: # xmm14 = mem[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
; SSE-NEXT: shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0,1],mem[0,3]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps %xmm8, 96(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps %xmm8, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps %xmm8, 112(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps %xmm8, 48(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps %xmm8, 64(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps %xmm8, (%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps %xmm8, 80(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps %xmm8, 16(%rsi)
-; SSE-NEXT: movaps %xmm2, 96(%rdx)
-; SSE-NEXT: movaps %xmm10, 32(%rdx)
-; SSE-NEXT: movaps %xmm14, 112(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movaps %xmm14, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movaps %xmm14, 16(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movaps %xmm14, 64(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movaps %xmm14, (%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movaps %xmm14, 112(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movaps %xmm14, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movaps %xmm14, 96(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movaps %xmm14, 32(%rsi)
+; SSE-NEXT: movaps %xmm5, 80(%rdx)
+; SSE-NEXT: movaps %xmm15, 16(%rdx)
+; SSE-NEXT: movaps %xmm10, 64(%rdx)
+; SSE-NEXT: movaps %xmm4, (%rdx)
+; SSE-NEXT: movaps %xmm8, 112(%rdx)
; SSE-NEXT: movaps %xmm9, 48(%rdx)
-; SSE-NEXT: movaps %xmm11, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, (%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 80(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 16(%rdx)
-; SSE-NEXT: movaps %xmm0, 96(%rcx)
-; SSE-NEXT: movaps %xmm1, 112(%rcx)
-; SSE-NEXT: movaps %xmm13, 64(%rcx)
-; SSE-NEXT: movaps %xmm7, 80(%rcx)
-; SSE-NEXT: movaps %xmm3, 32(%rcx)
-; SSE-NEXT: movaps %xmm5, 48(%rcx)
-; SSE-NEXT: movaps %xmm6, (%rcx)
-; SSE-NEXT: movaps %xmm4, 16(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 96(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, 32(%rdx)
+; SSE-NEXT: movaps %xmm0, 16(%rcx)
+; SSE-NEXT: movaps %xmm1, (%rcx)
+; SSE-NEXT: movaps %xmm11, 48(%rcx)
+; SSE-NEXT: movaps %xmm7, 32(%rcx)
+; SSE-NEXT: movaps %xmm2, 80(%rcx)
+; SSE-NEXT: movaps %xmm3, 64(%rcx)
+; SSE-NEXT: movaps %xmm12, 112(%rcx)
+; SSE-NEXT: movaps %xmm6, 96(%rcx)
; SSE-NEXT: addq $360, %rsp # imm = 0x168
; SSE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
index 9803126..25254de 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-4.ll
@@ -153,23 +153,23 @@
;
; AVX1-ONLY-LABEL: load_i32_stride4_vf4:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,0]
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,0]
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm5 = xmm0[1],xmm1[1],zero,zero
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm6 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm7 = zero,zero,xmm2[2],xmm3[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm1[3,0],xmm0[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm4, (%rsi)
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm6 = xmm3[1],xmm4[1],zero,zero
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm6 = zero,zero,xmm0[2],xmm1[2]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm7 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm4[3,0],xmm3[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm1[2,0],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, (%rsi)
; AVX1-ONLY-NEXT: vmovaps %xmm5, (%rdx)
; AVX1-ONLY-NEXT: vmovaps %xmm6, (%rcx)
; AVX1-ONLY-NEXT: vmovaps %xmm0, (%r8)
@@ -294,37 +294,37 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm0[2,3,0,1]
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm4[0],ymm0[0],ymm4[1],ymm0[1],ymm4[4],ymm0[4],ymm4[5],ymm0[5]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,0],ymm5[4,5],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm8[0],xmm7[0]
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm9[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm9 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm6[0],xmm5[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm7 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[1,0],ymm4[1,0],ymm0[5,4],ymm4[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm10[2,0],ymm9[2,3],ymm10[6,4],ymm9[6,7]
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm10 = xmm5[1],xmm6[1],zero,zero
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm11 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0,1],xmm11[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm10[2,0],ymm7[2,3],ymm10[6,4],ymm7[6,7]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm11 = xmm8[1],xmm9[1],zero,zero
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm11[0,1],xmm10[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3],ymm7[4,5,6,7]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm10 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm11 = ymm4[2],ymm0[2],ymm4[3],ymm0[3],ymm4[6],ymm0[6],ymm4[7],ymm0[7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm11[0,1],ymm10[2,0],ymm11[4,5],ymm10[6,4]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm11 = xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm12 = zero,zero,xmm7[2],xmm8[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm11[0,1],xmm12[2,3]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm11 = zero,zero,xmm5[2],xmm6[2]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm12 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm12[0,1],xmm11[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm4[3,0],ymm0[7,4],ymm4[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,3],ymm0[6,4],ymm1[6,7]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm7[2],xmm8[2],xmm7[3],xmm8[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm6[3,0],xmm5[3,0]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm9[3,0],xmm8[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm2[2,0],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rsi)
-; AVX1-ONLY-NEXT: vmovaps %ymm9, (%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm7, (%rdx)
; AVX1-ONLY-NEXT: vmovaps %ymm10, (%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r8)
; AVX1-ONLY-NEXT: vzeroupper
@@ -535,7 +535,7 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm3[2,3,0,1]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
; AVX1-ONLY-NEXT: vmovaps %ymm3, %ymm14
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm15
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm7 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
@@ -555,24 +555,22 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm12[0,1],xmm11[2,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm4[2,3,0,1]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm4[0],ymm6[0],ymm4[2],ymm6[2]
; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm8
-; AVX1-ONLY-NEXT: vmovaps %ymm4, %ymm12
-; AVX1-ONLY-NEXT: vmovups %ymm5, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm5[2,3,0,1]
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm13 = ymm11[0],ymm5[0],ymm11[1],ymm5[1],ymm11[4],ymm5[4],ymm11[5],ymm5[5]
+; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm13 = ymm9[0],ymm5[0],ymm9[1],ymm5[1],ymm9[4],ymm5[4],ymm9[5],ymm5[5]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm13[0,1],ymm0[2,0],ymm13[4,5],ymm0[6,4]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm12[0],xmm11[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm5
; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm6[0]
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm5[0],xmm9[1],xmm5[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm6[0,1],xmm0[2,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -587,63 +585,64 @@
; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm6 = xmm5[1],xmm1[1],zero,zero
; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm14
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm10[0],xmm2[0],xmm10[1],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %ymm12, %ymm0
-; AVX1-ONLY-NEXT: vmovaps %ymm8, %ymm12
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, %ymm10
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm3 = ymm8[0],ymm0[0],ymm8[1],ymm0[1],ymm8[4],ymm0[4],ymm8[5],ymm0[5]
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm2[1,0],ymm11[1,0],ymm2[5,4],ymm11[5,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm2[1,0],ymm9[1,0],ymm2[5,4],ymm9[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[2,0],ymm3[2,3],ymm6[6,4],ymm3[6,7]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm11[0],xmm12[0],xmm11[1],xmm12[1]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm6 = xmm9[1],xmm8[1],zero,zero
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[0],xmm8[1],zero,zero
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm6 = ymm13[2],ymm15[2],ymm13[3],ymm15[3],ymm13[6],ymm15[6],ymm13[7],ymm15[7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm6[0,1],ymm3[2,0],ymm6[4,5],ymm3[6,4]
; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm6 = xmm5[2],xmm14[2],xmm5[3],xmm14[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm7 = zero,zero,xmm10[2],xmm3[2]
-; AVX1-ONLY-NEXT: vmovaps %xmm10, %xmm14
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm7 = zero,zero,xmm1[2],xmm3[2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm0[1],ymm12[1],ymm0[3],ymm12[3]
+; AVX1-ONLY-NEXT: vmovaps %ymm10, %ymm14
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm0[1],ymm10[1],ymm0[3],ymm10[3]
; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm7 = ymm11[2],ymm2[2],ymm11[3],ymm2[3],ymm11[6],ymm2[6],ymm11[7],ymm2[7]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm7 = ymm9[2],ymm2[2],ymm9[3],ymm2[3],ymm9[6],ymm2[6],ymm9[7],ymm2[7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,0],ymm7[4,5],ymm6[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm7 = xmm10[2],xmm8[2],xmm10[3],xmm8[3]
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm0 = zero,zero,xmm9[2],xmm1[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm7[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm11, %xmm10
+; AVX1-ONLY-NEXT: vmovaps %xmm12, %xmm11
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm7 = zero,zero,xmm10[2],xmm12[2]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm12[2],xmm8[2],xmm12[3],xmm8[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vunpckhps (%rsp), %ymm2, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm4 = ymm2[2],mem[2],ymm2[3],mem[3],ymm2[6],mem[6],ymm2[7],mem[7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm15[3,0],ymm13[3,0],ymm15[7,4],ymm13[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm4[2,3],ymm2[6,4],ymm4[6,7]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm4 = xmm14[2],xmm3[2],xmm14[3],xmm3[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm3[3,0],mem[3,0]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm4 = xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm1[3,0],mem[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm6[2,0],xmm4[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm12[2],ymm5[2],ymm12[3],ymm5[3],ymm12[6],ymm5[6],ymm12[7],ymm5[7]
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[3,0],ymm11[3,0],ymm4[7,4],ymm11[7,4]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm14[2],ymm5[2],ymm14[3],ymm5[3],ymm14[6],ymm5[6],ymm14[7],ymm5[7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm1[3,0],ymm9[3,0],ymm1[7,4],ymm9[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[2,0],ymm3[2,3],ymm4[6,4],ymm3[6,7]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm9[2],xmm1[2],xmm9[3],xmm1[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm8[3,0],xmm10[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm4[2,0],xmm1[2,3]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm4 = xmm10[2],xmm11[2],xmm10[3],xmm11[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm8[3,0],xmm12[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
@@ -1093,7 +1092,6 @@
; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm2[2,3,0,1]
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm5[0],ymm2[0],ymm5[1],ymm2[1],ymm5[4],ymm2[4],ymm5[5],ymm2[5]
-; AVX1-ONLY-NEXT: vmovaps %ymm5, %ymm11
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1105,7 +1103,7 @@
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm7
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
; AVX1-ONLY-NEXT: vmovaps %xmm7, %xmm9
-; AVX1-ONLY-NEXT: vmovaps %xmm7, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1139,7 +1137,7 @@
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
-; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm15
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm12
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1168,86 +1166,86 @@
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1],ymm8[2,0],ymm10[4,5],ymm8[6,4]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm5[0]
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm15 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm15[0,1],xmm0[2,0]
+; AVX1-ONLY-NEXT: vmovaps %xmm3, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm14 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm14[0,1],xmm0[2,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm13[0],ymm6[0],ymm13[1],ymm6[1],ymm13[4],ymm6[4],ymm13[5],ymm6[5]
-; AVX1-ONLY-NEXT: vmovaps %ymm11, %ymm10
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %ymm5, %ymm10
+; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[1,0],ymm11[1,0],ymm5[5,4],ymm11[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[1,0],ymm10[1,0],ymm5[5,4],ymm10[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = xmm9[1],xmm11[1],zero,zero
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm15 = xmm6[0],xmm9[0],xmm6[1],xmm9[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm14 = xmm6[0],xmm9[0],xmm6[1],xmm9[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[1,0],ymm14[1,0],ymm12[5,4],ymm14[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[1,0],ymm15[1,0],ymm12[5,4],ymm15[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = xmm7[1],xmm4[1],zero,zero
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm2[0],mem[0],xmm2[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm2[0],mem[0],xmm2[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm15 = ymm0[0],ymm2[0],ymm0[1],ymm2[1],ymm0[4],ymm2[4],ymm0[5],ymm2[5]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm14 = ymm0[0],ymm2[0],ymm0[1],ymm2[1],ymm0[4],ymm2[4],ymm0[5],ymm2[5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm3[1,0],ymm1[5,4],ymm3[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm1[2,0],ymm15[2,3],ymm1[6,4],ymm15[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm1[2,0],ymm14[2,3],ymm1[6,4],ymm14[6,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0],xmm1[1],zero,zero
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm4[0],mem[0],xmm4[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm4[0],mem[0],xmm4[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm8 = ymm12[0],ymm7[0],ymm12[1],ymm7[1],ymm12[4],ymm7[4],ymm12[5],ymm7[5]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm14[1,0],ymm1[5,4],ymm14[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm15[1,0],ymm1[5,4],ymm15[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm1[2,0],ymm8[2,3],ymm1[6,4],ymm8[6,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0],xmm1[1],zero,zero
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm15[0],mem[0],xmm15[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = mem[0],xmm14[1],zero,zero
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm1[1],ymm13[1],ymm1[3],ymm13[3]
; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm10[2],ymm5[2],ymm10[3],ymm5[3],ymm10[6],ymm5[6],ymm10[7],ymm5[7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm1[0,1],ymm8[2,0],ymm1[4,5],ymm8[6,4]
-; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm1[2],xmm11[2],xmm1[3],xmm11[3]
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm15 = zero,zero,xmm6[2],xmm9[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,zero,xmm6[2],xmm9[2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
@@ -1257,9 +1255,9 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; AVX1-ONLY-NEXT: vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = zero,zero,xmm4[2],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT: vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = zero,zero,xmm4[2],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
@@ -1274,22 +1272,22 @@
; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm15 = zero,zero,xmm2[2],xmm3[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,zero,xmm2[2],xmm3[2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm12[1],ymm7[3],ymm12[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm14[2],ymm13[2],ymm14[3],ymm13[3],ymm14[6],ymm13[6],ymm14[7],ymm13[7]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm15[2],ymm13[2],ymm15[3],ymm13[3],ymm15[6],ymm13[6],ymm15[7],ymm13[7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm14[2],mem[2],xmm14[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm15 = zero,zero,xmm7[2],xmm12[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = zero,zero,xmm7[2],mem[0]
+; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm14 = xmm12[2],xmm15[2],xmm12[3],xmm15[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
@@ -1300,7 +1298,7 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm1[2],xmm9[2],xmm1[3],xmm9[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $51, (%rsp), %xmm9, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm9 = xmm9[3,0],mem[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm9[2,0],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -1332,9 +1330,9 @@
; AVX1-ONLY-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = ymm13[3,0],mem[3,0],ymm13[7,4],mem[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[2,0],ymm2[2,3],ymm3[6,4],ymm2[6,7]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm3 = xmm7[2],xmm12[2],xmm7[3],xmm12[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,0],xmm14[3,0]
+; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm7[2],mem[2],xmm7[3],mem[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm15[3,0],xmm12[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm5[2,0],xmm3[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
@@ -1359,7 +1357,7 @@
; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovaps %ymm15, (%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm14, (%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%r8)
; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r8)
; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%r8)
@@ -2424,15 +2422,15 @@
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm13 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,1],ymm12[2,0],ymm13[4,5],ymm12[6,4]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0]
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm13 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm13[0,1],xmm0[2,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -2557,12 +2555,12 @@
; AVX1-ONLY-NEXT: # ymm1 = ymm1[1,0],mem[1,0],ymm1[5,4],mem[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0],xmm1[1],zero,zero
+; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm13[0],mem[0],xmm13[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vinsertps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[0],xmm13[1],zero,zero
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm14[1],ymm2[3],ymm14[3]
@@ -2584,9 +2582,9 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm10[1],ymm4[3],ymm10[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm11[1],ymm4[3],ymm11[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm9[2],ymm5[3],ymm9[3],ymm5[6],ymm9[6],ymm5[7],ymm9[7]
@@ -2603,9 +2601,9 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm6[1],ymm8[1],ymm6[3],ymm8[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm7[2],ymm11[2],ymm7[3],ymm11[3],ymm7[6],ymm11[6],ymm7[7],ymm11[7]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm7[2],ymm10[2],ymm7[3],ymm10[3],ymm7[6],ymm10[6],ymm7[7],ymm10[7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
@@ -2672,12 +2670,12 @@
; AVX1-ONLY-NEXT: # ymm1 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT: vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = zero,zero,xmm1[2],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = zero,zero,xmm13[2],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[2],mem[2],xmm13[3],mem[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -2693,8 +2691,7 @@
; AVX1-ONLY-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm13 = xmm13[3,0],mem[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm13[2,0],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
@@ -2710,24 +2707,25 @@
; AVX1-ONLY-NEXT: # xmm13 = xmm13[3,0],mem[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm13[2,0],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm10[2],ymm4[2],ymm10[3],ymm4[3],ymm10[6],ymm4[6],ymm10[7],ymm4[7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm9[3,0],ymm5[3,0],ymm9[7,4],ymm5[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[2,0],ymm1[2,3],ymm13[6,4],ymm1[6,7]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm13 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm12[3,0],xmm15[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm14[2,0],xmm13[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm10 = ymm8[2],ymm6[2],ymm8[3],ymm6[3],ymm8[6],ymm6[6],ymm8[7],ymm6[7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[3,0],ymm7[3,0],ymm11[7,4],ymm7[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm11[2,0],ymm10[2,3],ymm11[6,4],ymm10[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm11[2],ymm4[2],ymm11[3],ymm4[3],ymm11[6],ymm4[6],ymm11[7],ymm4[7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,0],ymm5[3,0],ymm9[7,4],ymm5[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm0[2,3],ymm1[6,4],ymm0[6,7]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm12[3,0],xmm15[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm13[2,0],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm6[2],ymm8[3],ymm6[3],ymm8[6],ymm6[6],ymm8[7],ymm6[7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm10[3,0],ymm7[3,0],ymm10[7,4],ymm7[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm10[2,0],ymm1[2,3],ymm10[6,4],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = xmm2[2],mem[2],xmm2[3],mem[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm2[3,0],mem[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm13[2,0],xmm11[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = xmm2[3,0],mem[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm11[2,0],xmm10[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm6 = ymm2[2],mem[2],ymm2[3],mem[3],ymm2[6],mem[6],ymm2[7],mem[7]
@@ -2739,9 +2737,9 @@
; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm9 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm9 = xmm2[2],mem[2],xmm2[3],mem[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm2[3,0],mem[3,0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm11[2,0],xmm9[2,3]
+; AVX1-ONLY-NEXT: vshufps $51, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = xmm2[3,0],mem[3,0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm10[2,0],xmm9[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
@@ -2840,11 +2838,11 @@
; AVX1-ONLY-NEXT: vmovaps %ymm4, 128(%r8)
; AVX1-ONLY-NEXT: vmovaps %ymm6, 64(%r8)
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 224(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 160(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 160(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm14, 32(%r8)
; AVX1-ONLY-NEXT: addq $2200, %rsp # imm = 0x898
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
index 853ea0f..556e238 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
@@ -84,18 +84,15 @@
; AVX512F-SLOW-NEXT: vmovdqa 16(%rdi), %xmm1
; AVX512F-SLOW-NEXT: vmovdqa 32(%rdi), %xmm2
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm0[0],xmm1[1],xmm0[2,3]
-; AVX512F-SLOW-NEXT: vpextrd $2, %xmm1, %eax
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[1,1,1,1]
-; AVX512F-SLOW-NEXT: vpinsrd $1, %eax, %xmm4, %xmm4
-; AVX512F-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm5
-; AVX512F-SLOW-NEXT: vpextrd $3, %xmm1, %eax
-; AVX512F-SLOW-NEXT: vpinsrd $1, %eax, %xmm5, %xmm1
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[1,2,2,3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX512F-SLOW-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10,11]
; AVX512F-SLOW-NEXT: vpbroadcastd 16(%rdi), %ymm5
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0],xmm2[1],xmm5[2,3]
; AVX512F-SLOW-NEXT: vmovq %xmm3, (%rsi)
; AVX512F-SLOW-NEXT: vmovq %xmm4, (%rdx)
-; AVX512F-SLOW-NEXT: vmovq %xmm1, (%rcx)
+; AVX512F-SLOW-NEXT: vpextrq $1, %xmm1, (%rcx)
; AVX512F-SLOW-NEXT: vmovq %xmm0, (%r8)
; AVX512F-SLOW-NEXT: vmovq %xmm2, (%r9)
; AVX512F-SLOW-NEXT: vzeroupper
@@ -128,18 +125,15 @@
; AVX512BW-SLOW-NEXT: vmovdqa 16(%rdi), %xmm1
; AVX512BW-SLOW-NEXT: vmovdqa 32(%rdi), %xmm2
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm0[0],xmm1[1],xmm0[2,3]
-; AVX512BW-SLOW-NEXT: vpextrd $2, %xmm1, %eax
-; AVX512BW-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[1,1,1,1]
-; AVX512BW-SLOW-NEXT: vpinsrd $1, %eax, %xmm4, %xmm4
-; AVX512BW-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm5
-; AVX512BW-SLOW-NEXT: vpextrd $3, %xmm1, %eax
-; AVX512BW-SLOW-NEXT: vpinsrd $1, %eax, %xmm5, %xmm1
+; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX512BW-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[1,2,2,3]
+; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX512BW-SLOW-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[12,13,14,15],xmm2[0,1,2,3,4,5,6,7,8,9,10,11]
; AVX512BW-SLOW-NEXT: vpbroadcastd 16(%rdi), %ymm5
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm2 = xmm5[0],xmm2[1],xmm5[2,3]
; AVX512BW-SLOW-NEXT: vmovq %xmm3, (%rsi)
; AVX512BW-SLOW-NEXT: vmovq %xmm4, (%rdx)
-; AVX512BW-SLOW-NEXT: vmovq %xmm1, (%rcx)
+; AVX512BW-SLOW-NEXT: vpextrq $1, %xmm1, (%rcx)
; AVX512BW-SLOW-NEXT: vmovq %xmm0, (%r8)
; AVX512BW-SLOW-NEXT: vmovq %xmm2, (%r9)
; AVX512BW-SLOW-NEXT: vzeroupper
@@ -589,26 +583,25 @@
; SSE-LABEL: load_i32_stride5_vf16:
; SSE: # %bb.0:
; SSE-NEXT: subq $296, %rsp # imm = 0x128
-; SSE-NEXT: movdqa 288(%rdi), %xmm3
+; SSE-NEXT: movdqa 208(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 272(%rdi), %xmm4
-; SSE-NEXT: movdqa 240(%rdi), %xmm9
-; SSE-NEXT: movdqa 256(%rdi), %xmm2
-; SSE-NEXT: movdqa (%rdi), %xmm11
-; SSE-NEXT: movdqa 16(%rdi), %xmm15
-; SSE-NEXT: movdqa 32(%rdi), %xmm6
-; SSE-NEXT: movdqa 48(%rdi), %xmm5
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 208(%rdi), %xmm7
+; SSE-NEXT: movdqa 192(%rdi), %xmm4
+; SSE-NEXT: movdqa 160(%rdi), %xmm9
+; SSE-NEXT: movdqa 176(%rdi), %xmm2
+; SSE-NEXT: movdqa 288(%rdi), %xmm10
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 272(%rdi), %xmm6
+; SSE-NEXT: movdqa 240(%rdi), %xmm11
+; SSE-NEXT: movdqa 256(%rdi), %xmm15
+; SSE-NEXT: movdqa (%rdi), %xmm5
+; SSE-NEXT: movdqa 16(%rdi), %xmm12
+; SSE-NEXT: movdqa 32(%rdi), %xmm8
+; SSE-NEXT: movdqa 48(%rdi), %xmm7
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm8
-; SSE-NEXT: movdqa 160(%rdi), %xmm10
-; SSE-NEXT: movdqa 176(%rdi), %xmm12
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
-; SSE-NEXT: movdqa %xmm10, %xmm1
+; SSE-NEXT: movdqa %xmm5, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2]
-; SSE-NEXT: movdqa %xmm8, %xmm13
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm7[2],xmm0[3],xmm7[3]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
@@ -618,14 +611,14 @@
; SSE-NEXT: movdqa %xmm11, %xmm14
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,2,2]
-; SSE-NEXT: movdqa %xmm6, %xmm8
+; SSE-NEXT: movdqa %xmm6, %xmm11
; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm10[2],xmm0[3],xmm10[3]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
-; SSE-NEXT: movdqa %xmm2, %xmm5
-; SSE-NEXT: movdqa %xmm9, %xmm6
+; SSE-NEXT: movdqa %xmm2, %xmm6
+; SSE-NEXT: movdqa %xmm9, %xmm7
; SSE-NEXT: movdqa %xmm9, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa %xmm4, %xmm2
@@ -634,38 +627,23 @@
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 80(%rdi), %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 96(%rdi), %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT: movdqa %xmm1, %xmm11
-; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: movdqa 80(%rdi), %xmm10
+; SSE-NEXT: movdqa 96(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 128(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rdi), %xmm4
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
-; SSE-NEXT: movdqa %xmm4, %xmm9
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 112(%rdi), %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm10, %xmm7
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 224(%rdi), %xmm3
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,0,1,1]
-; SSE-NEXT: movdqa %xmm3, %xmm10
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm13[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
-; SSE-NEXT: movdqa %xmm14, %xmm13
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movdqa 64(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
@@ -673,22 +651,34 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE-NEXT: movdqa %xmm6, %xmm4
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE-NEXT: movdqa %xmm14, %xmm13
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 304(%rdi), %xmm14
+; SSE-NEXT: movdqa 304(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
+; SSE-NEXT: movdqa %xmm7, %xmm4
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movdqa 224(%rdi), %xmm14
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[0,0,1,1]
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE-NEXT: movdqa %xmm11, %xmm3
-; SSE-NEXT: movdqa %xmm11, (%rsp) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
+; SSE-NEXT: movdqa %xmm10, %xmm7
+; SSE-NEXT: movdqa %xmm10, (%rsp) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movdqa 144(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -697,18 +687,17 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: movdqa %xmm5, %xmm9
+; SSE-NEXT: movdqa %xmm5, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[2,2,3,3]
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
-; SSE-NEXT: movdqa %xmm10, %xmm7
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: movdqa %xmm13, %xmm9
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3]
; SSE-NEXT: movdqa %xmm15, %xmm11
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
@@ -718,15 +707,15 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
-; SSE-NEXT: movdqa %xmm5, %xmm10
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[2,2,3,3]
+; SSE-NEXT: movdqa %xmm6, %xmm10
; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,0,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm4[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm6, %xmm0
+; SSE-NEXT: movdqa %xmm7, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[2,2,3,3]
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
@@ -734,34 +723,34 @@
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
; SSE-NEXT: movdqa %xmm8, %xmm7
; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,3,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
-; SSE-NEXT: movdqa %xmm2, %xmm6
+; SSE-NEXT: movdqa %xmm2, %xmm5
; SSE-NEXT: movdqa %xmm2, %xmm15
-; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[3,3,3,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
-; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm0[0],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
-; SSE-NEXT: movdqa %xmm5, %xmm2
+; SSE-NEXT: movdqa %xmm6, %xmm2
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
; SSE-NEXT: movdqa %xmm4, %xmm0
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: pshufd $255, (%rsp), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,3,3,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
@@ -779,50 +768,50 @@
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT: # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm11[0],xmm12[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm5[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm6[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm15[0],xmm10[1],xmm15[1]
; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
; SSE-NEXT: # xmm15 = mem[2,2,2,2]
; SSE-NEXT: punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm14[2],xmm15[3],xmm14[3]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm10[0],xmm15[1]
; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm3[1,1,1,1]
-; SSE-NEXT: movdqa (%rsp), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm14[0],xmm3[1],xmm14[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm14[0],xmm10[1],xmm14[1]
; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm4[2,2,2,2]
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
; SSE-NEXT: # xmm14 = xmm14[2],mem[2],xmm14[3],mem[3]
-; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm3[0],xmm14[1]
+; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm10[0],xmm14[1]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT: movaps %xmm10, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: movaps %xmm11, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT: movaps %xmm11, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT: movaps %xmm11, (%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: movaps %xmm11, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, (%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 32(%rdx)
; SSE-NEXT: movapd %xmm13, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 48(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, (%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 32(%rcx)
; SSE-NEXT: movapd %xmm0, 16(%r8)
-; SSE-NEXT: movapd %xmm2, 48(%r8)
-; SSE-NEXT: movapd %xmm6, (%r8)
-; SSE-NEXT: movapd %xmm7, 32(%r8)
+; SSE-NEXT: movapd %xmm2, 32(%r8)
+; SSE-NEXT: movapd %xmm5, 48(%r8)
+; SSE-NEXT: movapd %xmm7, (%r8)
; SSE-NEXT: movapd %xmm14, 16(%r9)
-; SSE-NEXT: movapd %xmm15, 48(%r9)
-; SSE-NEXT: movapd %xmm12, (%r9)
-; SSE-NEXT: movapd %xmm1, 32(%r9)
+; SSE-NEXT: movapd %xmm15, 32(%r9)
+; SSE-NEXT: movapd %xmm12, 48(%r9)
+; SSE-NEXT: movapd %xmm1, (%r9)
; SSE-NEXT: addq $296, %rsp # imm = 0x128
; SSE-NEXT: retq
;
@@ -1226,378 +1215,386 @@
; SSE-LABEL: load_i32_stride5_vf32:
; SSE: # %bb.0:
; SSE-NEXT: subq $904, %rsp # imm = 0x388
-; SSE-NEXT: movdqa (%rdi), %xmm12
-; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 16(%rdi), %xmm7
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 32(%rdi), %xmm9
-; SSE-NEXT: movdqa 48(%rdi), %xmm6
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 448(%rdi), %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 432(%rdi), %xmm4
-; SSE-NEXT: movdqa 400(%rdi), %xmm11
-; SSE-NEXT: movdqa 416(%rdi), %xmm14
-; SSE-NEXT: movdqa 128(%rdi), %xmm5
+; SSE-NEXT: movdqa 288(%rdi), %xmm5
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rdi), %xmm8
-; SSE-NEXT: movdqa 80(%rdi), %xmm13
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 96(%rdi), %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT: movdqa %xmm1, %xmm10
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: movdqa 272(%rdi), %xmm9
+; SSE-NEXT: movdqa 240(%rdi), %xmm12
+; SSE-NEXT: movdqa 256(%rdi), %xmm10
+; SSE-NEXT: movdqa 528(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 512(%rdi), %xmm3
+; SSE-NEXT: movdqa 480(%rdi), %xmm13
+; SSE-NEXT: movdqa 496(%rdi), %xmm6
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 208(%rdi), %xmm7
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 192(%rdi), %xmm8
+; SSE-NEXT: movdqa 160(%rdi), %xmm11
+; SSE-NEXT: movdqa 176(%rdi), %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm11, %xmm1
+; SSE-NEXT: movdqa %xmm11, %xmm15
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2]
+; SSE-NEXT: movdqa %xmm8, %xmm11
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm7[2],xmm0[3],xmm7[3]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm13, %xmm7
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
+; SSE-NEXT: movdqa %xmm3, %xmm13
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT: movdqa %xmm10, %xmm6
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm1
-; SSE-NEXT: movdqa %xmm11, %xmm5
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT: movdqa %xmm12, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
-; SSE-NEXT: movdqa %xmm9, %xmm12
-; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm6[2],xmm0[3],xmm6[3]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 320(%rdi), %xmm7
-; SSE-NEXT: movdqa 336(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT: movdqa %xmm7, %xmm1
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 368(%rdi), %xmm2
+; SSE-NEXT: movdqa 560(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 352(%rdi), %xmm15
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,2,2]
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 240(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 256(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 288(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 272(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 560(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 576(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT: movdqa 576(%rdi), %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: movdqa 608(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 592(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 160(%rdi), %xmm6
-; SSE-NEXT: movdqa 176(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT: movdqa %xmm6, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 208(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 480(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 496(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 528(%rdi), %xmm3
+; SSE-NEXT: movdqa 592(%rdi), %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
+; SSE-NEXT: movdqa %xmm3, %xmm8
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 512(%rdi), %xmm0
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%rdi), %xmm3
+; SSE-NEXT: movdqa 16(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT: movdqa %xmm3, %xmm2
+; SSE-NEXT: movdqa %xmm3, %xmm10
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movdqa 32(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 48(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 320(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 336(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movdqa 368(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 352(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 144(%rdi), %xmm1
+; SSE-NEXT: movdqa 80(%rdi), %xmm14
+; SSE-NEXT: movdqa 96(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,1,1]
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT: movdqa 128(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 112(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm8[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 464(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm4[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,2,2,2]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm3[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 64(%rdi), %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,0,1,1]
-; SSE-NEXT: movdqa %xmm2, %xmm9
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm12[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 384(%rdi), %xmm1
+; SSE-NEXT: movdqa 400(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 416(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm15[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 304(%rdi), %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT: movdqa 448(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT: movdqa 432(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,2,2,2]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm3[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 624(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT: pshufd $238, (%rsp), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm6, %xmm12
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 224(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa 544(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: # xmm7 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
-; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm1[2],xmm11[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm15[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT: movdqa 224(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm11[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
+; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm7, %xmm1
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[1,1,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm11[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm1[2],xmm8[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm14, %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm12, %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT: movdqa 544(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm13[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
+; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm12[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT: movdqa 304(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm13[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[2,3,2,3]
+; SSE-NEXT: movdqa %xmm5, %xmm9
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT: movdqa 624(%rdi), %xmm15
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm15[0,0,1,1]
+; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm10[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT: movdqa 64(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,0,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT: # xmm4 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm3[0],xmm4[1]
+; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm12[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT: movdqa 384(%rdi), %xmm6
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[0,0,1,1]
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%rsp), %xmm8 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm8[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm14[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT: movdqa 144(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,0,1,1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT: # xmm10 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm2[0],xmm10[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm3[0],xmm10[1]
+; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT: movdqa 464(%rdi), %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[0,0,1,1]
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT: # xmm10 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm3[0],xmm10[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm2[0],xmm10[1]
+; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[0,0,1,1]
+; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm13, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT: # xmm10 = mem[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm6[0],xmm10[1],xmm6[1]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm1[2],xmm13[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm4[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm13[0],xmm10[1]
-; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[2,2,3,3]
-; SSE-NEXT: movdqa %xmm5, %xmm10
-; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT: movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
-; SSE-NEXT: movdqa %xmm12, %xmm13
-; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,3,3,3]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm9, %xmm2
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[0,0,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm6, %xmm2
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[0,0,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,0,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,2,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm14[3,3,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm1[2],xmm14[3],xmm1[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[0,0,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT: movdqa %xmm4, %xmm5
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1]
+; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm3[0],xmm5[1]
+; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3]
+; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
+; SSE-NEXT: movdqa %xmm11, %xmm14
+; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
+; SSE-NEXT: movdqa %xmm15, %xmm13
+; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
-; SSE-NEXT: movdqa %xmm14, %xmm15
-; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,2,3,3]
+; SSE-NEXT: movdqa %xmm7, %xmm12
+; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
-; SSE-NEXT: movdqa %xmm3, %xmm11
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,3,3,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3]
-; SSE-NEXT: movdqa %xmm4, %xmm8
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm0[0],xmm8[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq (%rsp), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm0[0],xmm6[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,2,3,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = mem[2,2,2,2]
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = xmm9[2],mem[2],xmm9[3],mem[3]
-; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm7[0],xmm9[1]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT: # xmm10 = mem[2,2,2,2]
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT: # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm6[0],xmm10[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; SSE-NEXT: movdqa (%rsp), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[2,2,2,2]
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm6[0],xmm8[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: # xmm7 = mem[2,2,2,2]
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT: # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
-; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm3[0],xmm7[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[2,2,2,2]
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm3[0],xmm6[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
@@ -1605,8 +1602,7 @@
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; SSE-NEXT: # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
@@ -1614,7 +1610,8 @@
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
@@ -1630,152 +1627,138 @@
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
-; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,2,2,2]
-; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
-; SSE-NEXT: pshufd $85, (%rsp), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[2,2,2,2]
; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, 96(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, 112(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, 48(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, 64(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, (%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, 80(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, 16(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 96(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 32(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 112(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, (%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, 80(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movaps %xmm14, 16(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 96(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 112(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 64(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 80(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 32(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 48(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, (%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movaps %xmm12, 16(%rcx)
-; SSE-NEXT: movapd %xmm6, 112(%r8)
-; SSE-NEXT: movapd %xmm8, 96(%r8)
-; SSE-NEXT: movapd %xmm10, 80(%r8)
-; SSE-NEXT: movapd %xmm11, 64(%r8)
-; SSE-NEXT: movapd %xmm15, 48(%r8)
-; SSE-NEXT: movapd %xmm13, 32(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: movaps %xmm6, 16(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: movaps %xmm6, (%r8)
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm15[0],xmm0[1]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 16(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 64(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, (%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: movaps %xmm15, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: movaps %xmm15, 96(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: movaps %xmm15, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 80(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 16(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 64(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, (%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 48(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 96(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 16(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, (%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 80(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 64(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 96(%rcx)
+; SSE-NEXT: movapd %xmm7, 112(%r8)
+; SSE-NEXT: movapd %xmm9, 96(%r8)
+; SSE-NEXT: movapd %xmm11, 80(%r8)
+; SSE-NEXT: movapd %xmm12, 64(%r8)
+; SSE-NEXT: movapd %xmm13, 48(%r8)
+; SSE-NEXT: movapd %xmm14, 32(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 16(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, (%r8)
; SSE-NEXT: movapd %xmm0, 112(%r9)
-; SSE-NEXT: movapd %xmm1, 96(%r9)
-; SSE-NEXT: movapd %xmm2, 80(%r9)
-; SSE-NEXT: movapd %xmm3, 64(%r9)
-; SSE-NEXT: movapd %xmm4, 48(%r9)
-; SSE-NEXT: movapd %xmm5, 32(%r9)
-; SSE-NEXT: movapd %xmm7, 16(%r9)
-; SSE-NEXT: movapd %xmm9, (%r9)
+; SSE-NEXT: movapd %xmm2, 96(%r9)
+; SSE-NEXT: movapd %xmm3, 80(%r9)
+; SSE-NEXT: movapd %xmm4, 64(%r9)
+; SSE-NEXT: movapd %xmm5, 48(%r9)
+; SSE-NEXT: movapd %xmm6, 32(%r9)
+; SSE-NEXT: movapd %xmm8, 16(%r9)
+; SSE-NEXT: movapd %xmm10, (%r9)
; SSE-NEXT: addq $904, %rsp # imm = 0x388
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i32_stride5_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1000, %rsp # imm = 0x3E8
+; AVX1-ONLY-NEXT: subq $920, %rsp # imm = 0x398
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm3
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm7
; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm4
; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %ymm5
; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm8
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm12
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm15
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, 288(%rdi), %ymm0, %ymm1
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm11[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4],ymm2[5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm5[2,3],ymm8[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vmovaps %ymm4, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 608(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm4
+; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm15
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm14
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm5
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 128(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
@@ -1787,130 +1770,130 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
-; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%rdi), %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3],ymm2[4,5,6],ymm0[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm8[1,3],ymm0[6,5],ymm8[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[3,0],ymm0[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 304(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm4[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm7[1,3],ymm0[6,5],ymm7[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[3,0],ymm0[6,4],ymm1[7,4]
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 304(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm4[2,3],ymm7[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[3,0],ymm0[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 624(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm6[1,3],ymm0[6,5],ymm6[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm2[2,3],ymm6[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[3,0],ymm0[6,4],ymm1[7,4]
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 624(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm14[1,3],ymm0[6,5],ymm14[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1],ymm5[2,3],ymm14[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[3,0],ymm0[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 144(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm5[1,3],ymm0[6,5],ymm5[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm11[2,3],ymm5[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[3,0],ymm0[6,4],ymm1[7,4]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm3[1,3],ymm0[6,5],ymm3[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm2[3,0],ymm0[6,4],ymm2[7,4]
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 144(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 464(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm2[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm3[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,1],ymm12[1,3],ymm0[6,5],ymm12[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm3[2,3],ymm12[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[3,0],ymm0[6,4],ymm1[7,4]
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 464(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm11[3,0],ymm8[2,0],ymm11[7,4],ymm8[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm8[2,1],ymm0[6,4],ymm8[6,5]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm2 = ymm12[0,1,2,3],mem[4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[1,0],ymm1[0,0],ymm5[5,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm11
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[3,0],ymm7[2,0],ymm13[7,4],ymm7[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm7[2,1],ymm0[6,4],ymm7[6,5]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm15[4,5],ymm10[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm13
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm13[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[1,0],ymm10[0,0],ymm13[5,4],ymm10[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm9
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[3,0],ymm6[2,0],ymm14[7,4],ymm6[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm6[2,1],ymm0[6,4],ymm6[6,5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm14[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm14[1,0],ymm3[0,0],ymm14[5,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm5[2,0],ymm0[7,4],ymm5[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm5[2,1],ymm0[6,4],ymm5[6,5]
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm11[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm11[1,0],ymm6[0,0],ymm11[5,4],ymm6[4,4]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm6[1,0],ymm12[0,0],ymm6[5,4],ymm12[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm10[3,0],ymm14[2,0],ymm10[7,4],ymm14[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm14[2,1],ymm0[6,4],ymm14[6,5]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm4 = ymm15[0,1,2,3],mem[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm4 = xmm4[1,0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm9[1,0],ymm13[0,0],ymm9[5,4],ymm13[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm12[2,0],ymm0[7,4],ymm12[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm12[2,1],ymm0[6,4],ymm12[6,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm3[2,0],ymm0[7,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,1],ymm0[6,4],ymm3[6,5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm4 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
@@ -1918,10 +1901,9 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm4 = xmm4[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm8[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[1,0],ymm5[0,0],ymm8[5,4],ymm5[4,4]
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm4[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[1,0],ymm10[0,0],ymm4[5,4],ymm10[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1929,117 +1911,115 @@
; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm7[3,0],ymm1[4,4],ymm7[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm8[3,0],ymm1[4,4],ymm8[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm8[2,2],ymm15[6,4],ymm8[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm5[2,0],ymm11[1,0],ymm5[6,4],ymm11[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm7[3,0],ymm2[4,4],ymm7[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm7[2,2],ymm15[6,4],ymm7[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm13[2,0],ymm10[1,0],ymm13[6,4],ymm10[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm9[3,0],ymm2[4,4],ymm9[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm9[2,2],ymm15[6,4],ymm9[6,6]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm14[2,0],ymm3[1,0],ymm14[6,4],ymm3[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm6[2,0],ymm12[1,0],ymm6[6,4],ymm12[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm3[0,0],ymm4[3,0],ymm3[4,4],ymm4[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm4[2,2],ymm15[6,4],ymm4[6,6]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm11[0,0],ymm14[3,0],ymm11[4,4],ymm14[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm14[2,2],ymm15[6,4],ymm14[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm11[2,0],ymm6[1,0],ymm11[6,4],ymm6[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm9[2,0],ymm13[1,0],ymm9[6,4],ymm13[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[0,0],ymm12[3,0],ymm4[4,4],ymm12[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm12[2,2],ymm15[6,4],ymm12[6,6]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[2,0],ymm5[1,0],ymm8[6,4],ymm5[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm1[5],ymm7[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm3[3,0],ymm1[4,4],ymm3[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm3[2,2],ymm15[6,4],ymm3[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[2,0],ymm10[1,0],ymm4[6,4],ymm10[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps $32, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm8[0,1,2,3,4],mem[5],ymm8[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm15 = mem[0,1,2,3],ymm15[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm15[0],mem[1],xmm15[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1,2],ymm7[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm13[7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm2[5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],mem[1],xmm10[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1,2],ymm6[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm9 = mem[0,1,2,3,4,5,6],ymm14[7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm3[5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm9 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0],mem[1],xmm9[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm9[0,1,2],ymm5[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2,3,4,5,6],ymm11[7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0],mem[1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2,3,4,5,6],ymm8[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm15[0,1,2],ymm8[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = mem[0,1,2,3,4,5,6],ymm5[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3,4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm2[5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0],mem[1],xmm8[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3,4,5,6],ymm6[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm1[5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5,6],ymm4[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm11[5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0],mem[1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1,2,3,4,5,6],ymm9[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r8)
@@ -2047,87 +2027,91 @@
; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, (%r9)
; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 32(%r9)
-; AVX1-ONLY-NEXT: addq $1000, %rsp # imm = 0x3E8
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%r9)
+; AVX1-ONLY-NEXT: addq $920, %rsp # imm = 0x398
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-ONLY-LABEL: load_i32_stride5_vf32:
; AVX2-ONLY: # %bb.0:
-; AVX2-ONLY-NEXT: subq $1000, %rsp # imm = 0x3E8
-; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm5
+; AVX2-ONLY-NEXT: subq $1032, %rsp # imm = 0x408
+; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm13
+; AVX2-ONLY-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 352(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %ymm7
; AVX2-ONLY-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 576(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovdqa 512(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %ymm10
; AVX2-ONLY-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 480(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovdqa 576(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vmovdqa 512(%rdi), %ymm15
+; AVX2-ONLY-NEXT: vmovdqa 480(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = [0,5,2,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm5 = [0,5,2,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3],ymm1[4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm5, %ymm1
; AVX2-ONLY-NEXT: vinserti128 $1, 288(%rdi), %ymm1, %ymm2
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm14[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm15[4],ymm3[5,6,7]
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm12[0,1,0,3]
+; AVX2-ONLY-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm9[4],ymm3[5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6],ymm2[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm10[2,3],ymm12[4,5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm9[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4],ymm2[5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm8, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm15[2,3],ymm8[4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vmovdqa %ymm15, %ymm9
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm5, %ymm1
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm11[0,1,0,3]
+; AVX2-ONLY-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4],ymm2[5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinserti128 $1, 608(%rdi), %ymm1, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm6[2,3],ymm7[4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vmovdqa %ymm6, %ymm12
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm5[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovdqa %ymm5, %ymm10
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4],ymm2[5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa %ymm11, %ymm7
-; AVX2-ONLY-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm5, %ymm1
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm13[0,1,0,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4],ymm2[5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa %ymm0, %ymm10
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinserti128 $1, 448(%rdi), %ymm1, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm5[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm13[4],ymm1[5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm3[2,3],ymm7[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vmovdqa %ymm3, %ymm13
+; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm5, %ymm0
+; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm2[0,1,0,3]
+; AVX2-ONLY-NEXT: vmovdqa %ymm2, %ymm14
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4],ymm1[5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa %ymm3, %ymm15
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vinserti128 $1, 128(%rdi), %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = <1,6,3,u>
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm6[2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm4[2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1],ymm14[2,3],ymm15[4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm12[2,3],ymm5[4,5],ymm12[6,7]
; AVX2-ONLY-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [5,2,7,0,5,2,7,0]
; AVX2-ONLY-NEXT: # ymm3 = mem[0,1,0,1]
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm3, %ymm2
@@ -2135,174 +2119,173 @@
; AVX2-ONLY-NEXT: vpbroadcastd 304(%rdi), %ymm2
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1],ymm4[2,3],ymm14[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1],ymm9[2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm11[2,3],ymm6[4,5],ymm11[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm3, %ymm2
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastd 624(%rdi), %ymm2
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa %ymm11, %ymm8
-; AVX2-ONLY-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm11[2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm13[2,3],ymm7[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1],ymm5[2,3],ymm13[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1],ymm14[2,3],ymm15[4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vmovdqa %ymm15, %ymm14
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm3, %ymm2
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastd 144(%rdi), %ymm2
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm1 = mem[0,1],ymm12[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = mem[0,1],ymm11[2,3],mem[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm10[2,3],ymm7[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm10[0,1],mem[2,3],ymm10[4,5],mem[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm3, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastd 464(%rdi), %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm9 = <2,7,4,u>
-; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm6[4,5],mem[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm9, %ymm1
+; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm13 = <2,7,4,u>
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm12[4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm13, %ymm1
; AVX2-ONLY-NEXT: vinserti128 $1, 256(%rdi), %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovdqa %ymm15, %ymm7
-; AVX2-ONLY-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT: vmovdqa %ymm5, %ymm7
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm0 = [0,5,0,5,0,5,0,5]
-; AVX2-ONLY-NEXT: vpermd %ymm12, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm4[4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vmovdqa %ymm14, %ymm15
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm9, %ymm1
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm9[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm13, %ymm1
; AVX2-ONLY-NEXT: vinserti128 $1, 576(%rdi), %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovdqu (%rsp), %ymm4 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5,6],ymm2[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vpermd %ymm5, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vpermd %ymm9, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm11[4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm9, %ymm1
-; AVX2-ONLY-NEXT: vinserti128 $1, 96(%rdi), %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovdqa %ymm13, %ymm6
-; AVX2-ONLY-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vpermd %ymm8, %ymm0, %ymm10
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1,2,3],ymm5[4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm13, %ymm1
+; AVX2-ONLY-NEXT: vinserti128 $1, 96(%rdi), %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vmovdqa %ymm14, %ymm2
+; AVX2-ONLY-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vpermd %ymm14, %ymm0, %ymm10
+; AVX2-ONLY-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm10[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm9, %ymm1
-; AVX2-ONLY-NEXT: vinserti128 $1, 416(%rdi), %ymm0, %ymm9
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm14[0,1,2,3,4,5,6],ymm9[7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm9 = ymm9[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm1[0,1,2],ymm9[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm11[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm13, %ymm4
+; AVX2-ONLY-NEXT: vinserti128 $1, 416(%rdi), %ymm0, %ymm8
+; AVX2-ONLY-NEXT: vmovdqu (%rsp), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm1[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm8 = ymm8[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm4[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm0, %ymm0
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm13[4,5],ymm7[6,7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT: vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm8 = mem[12,13,14,15],ymm12[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm12[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm8 = ymm8[0,3,2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2],ymm0[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm8 = [1,6,1,6,1,6,1,6]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm9 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm9 = mem[12,13,14,15],ymm11[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm11[16,17,18,19,20,21,22,23,24,25,26,27]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm9 = [1,6,1,6,1,6,1,6]
-; AVX2-ONLY-NEXT: vpermd %ymm12, %ymm9, %ymm10
+; AVX2-ONLY-NEXT: vpermd %ymm11, %ymm8, %ymm10
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm12[4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,0,2,2,7,4,6,6]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm15[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm15[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,3,2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm5, %ymm9, %ymm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm1[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm12[4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,0,2,2,7,4,6,6]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm5[12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10,11],ymm5[28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm6[12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10,11],ymm6[28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,3,2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm8, %ymm9, %ymm10
-; AVX2-ONLY-NEXT: vmovdqa %ymm8, %ymm15
+; AVX2-ONLY-NEXT: vpermd %ymm9, %ymm8, %ymm10
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm8[4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,0,2,2,7,4,6,6]
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm2[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm15[12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10,11],ymm15[28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,3,2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm14, %ymm8, %ymm10
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm5[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm3[12,13,14,15],ymm15[0,1,2,3,4,5,6,7,8,9,10,11],ymm3[28,29,30,31],ymm15[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vmovdqa %ymm3, %ymm14
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,3,2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermd %ymm7, %ymm9, %ymm9
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm8, %ymm8
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm11[0,1,2,3],mem[4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm9 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm9 = mem[0,1],ymm13[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm13[5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm10 = <4,1,6,u>
-; AVX2-ONLY-NEXT: vpermd %ymm0, %ymm10, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm9 = [2,7,2,7,2,7,2,7]
-; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm4 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
-; AVX2-ONLY-NEXT: vperm2i128 $2, (%rsp), %ymm12, %ymm11 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm11 = mem[0,1],ymm12[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm12[5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm10, %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm11[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm4 = mem[0,1],ymm1[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm1[5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm5, %ymm10, %ymm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm15, %ymm9, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm4 = ymm14[0,1,2,3],mem[4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm10, %ymm4
-; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm5 = mem[0,1],ymm8[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm8[5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm7, %ymm9, %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm7 = ymm7[0,1],ymm13[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm13[5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm8 = <4,1,6,u>
+; AVX2-ONLY-NEXT: vpermd %ymm0, %ymm8, %ymm0
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm7[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm7 = [2,7,2,7,2,7,2,7]
+; AVX2-ONLY-NEXT: vpermd %ymm11, %ymm7, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm9 = mem[0,1],ymm12[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm12[5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm8, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3],ymm14[4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vperm2i128 $2, (%rsp), %ymm5, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm9 = mem[0,1],ymm5[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm5[5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm8, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm7, %ymm1
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = mem[0,1],ymm4[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm5 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm5, %ymm8, %ymm5
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm4, 64(%rsi)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
@@ -2335,11 +2318,11 @@
; AVX2-ONLY-NEXT: vmovaps %ymm4, 96(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%r8)
+; AVX2-ONLY-NEXT: vmovdqa %ymm3, (%r9)
; AVX2-ONLY-NEXT: vmovdqa %ymm1, 64(%r9)
-; AVX2-ONLY-NEXT: vmovdqa %ymm2, (%r9)
-; AVX2-ONLY-NEXT: vmovdqa %ymm3, 96(%r9)
+; AVX2-ONLY-NEXT: vmovdqa %ymm2, 96(%r9)
; AVX2-ONLY-NEXT: vmovdqa %ymm0, 32(%r9)
-; AVX2-ONLY-NEXT: addq $1000, %rsp # imm = 0x3E8
+; AVX2-ONLY-NEXT: addq $1032, %rsp # imm = 0x408
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
;
@@ -3582,7 +3565,7 @@
;
; AVX1-ONLY-LABEL: load_i32_stride5_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $2488, %rsp # imm = 0x9B8
+; AVX1-ONLY-NEXT: subq $2472, %rsp # imm = 0x9A8
; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm4
; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm9
@@ -3606,7 +3589,7 @@
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm5[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
@@ -3615,9 +3598,9 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm10[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm9[4],ymm3[5,6,7]
@@ -3625,280 +3608,280 @@
; AVX1-ONLY-NEXT: vinsertf128 $1, 608(%rdi), %ymm2, %ymm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3],ymm2[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1],ymm4[2,3],ymm0[4,5],ymm4[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %ymm7
; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 928(%rdi), %ymm2, %ymm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
-; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %ymm13
; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm12
; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm9[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm13[4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 1248(%rdi), %ymm2, %ymm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm11[4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 128(%rdi), %ymm2, %ymm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm9[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm14[4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 448(%rdi), %ymm2, %ymm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm15[4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, 768(%rdi), %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm3
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
-; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1088(%rdi), %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, 768(%rdi), %ymm2, %ymm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm2[0],xmm3[1],xmm2[2],xmm3[3]
+; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm1[4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1088(%rdi), %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm5[2,3,0,1]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,1],ymm6[1,3],ymm2[6,5],ymm6[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm3[3,0],ymm2[6,4],ymm3[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm6[1,3],ymm3[6,5],ymm6[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm5
; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 304(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 304(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm6[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,1],ymm5[1,3],ymm2[6,5],ymm5[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm6[2,3],ymm5[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm3[3,0],ymm2[6,4],ymm3[7,4]
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 624(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm13[2,3,0,1]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm0[1,3],ymm3[6,5],ymm0[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0,1],ymm6[2,3],ymm0[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 624(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm7[2,3,0,1]
; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,1],ymm8[1,3],ymm2[6,5],ymm8[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1],ymm13[2,3],ymm8[4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm3[3,0],ymm2[6,4],ymm3[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm8[1,3],ymm3[6,5],ymm8[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm7[2,3],ymm8[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 944(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 944(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,1],ymm9[1,3],ymm2[6,5],ymm9[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm9[0,1],ymm12[2,3],ymm9[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm3[3,0],ymm2[6,4],ymm3[7,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm12[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm13[1,3],ymm3[6,5],ymm13[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm12[2,3],ymm13[4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm13, %ymm6
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 1264(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,1],ymm7[1,3],ymm2[6,5],ymm7[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1],ymm4[2,3],ymm7[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm3[3,0],ymm2[6,4],ymm3[7,4]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 144(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 1264(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm10[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,1],ymm14[1,3],ymm2[6,5],ymm14[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1],ymm10[2,3],ymm14[4,5],ymm10[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm3[3,0],ymm2[6,4],ymm3[7,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm10[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm11[1,3],ymm3[6,5],ymm11[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 144(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm14[1,3],ymm3[6,5],ymm14[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm14[0,1],ymm9[2,3],ymm14[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 464(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm11[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,1],ymm10[1,3],ymm2[6,5],ymm10[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1],ymm11[2,3],ymm10[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm3[3,0],ymm2[6,4],ymm3[7,4]
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 784(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 464(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm15[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,1],ymm11[1,3],ymm2[6,5],ymm11[5,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1],ymm15[2,3],ymm11[4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm11, %ymm13
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm3[3,0],ymm2[6,4],ymm3[7,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm15[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm13[1,3],ymm3[6,5],ymm13[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1],ymm15[2,3],ymm13[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 784(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,1],ymm9[1,3],ymm3[6,5],ymm9[5,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm9[0,1],ymm2[2,3],ymm9[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[3,0],ymm3[6,4],ymm4[7,4]
; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm0[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,2,3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss 1104(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[2,0],ymm0[7,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,1],ymm0[6,4],ymm1[6,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,2,3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss 1104(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[1,0],ymm10[0,0],ymm1[5,4],ymm10[4,4]
+; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[2,0],ymm0[7,4],ymm1[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,1],ymm0[6,4],ymm1[6,5]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[1,0],ymm2[0,0],ymm1[5,4],ymm2[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm5[2,0],ymm0[7,4],ymm5[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm5[2,1],ymm0[6,4],ymm5[6,5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
@@ -3911,13 +3894,13 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm9[2,0],ymm0[7,4],ymm9[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm9[2,1],ymm0[6,4],ymm9[6,5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[2,0],ymm0[7,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,1],ymm0[6,4],ymm1[6,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
@@ -3930,10 +3913,10 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm1[2,0],ymm0[7,4],ymm1[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,1],ymm0[6,4],ymm1[6,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm6[2,0],ymm0[7,4],ymm6[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm6[2,1],ymm0[6,4],ymm6[6,5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
@@ -3942,16 +3925,35 @@
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm2[0,0],ymm1[5,4],ymm2[4,4]
+; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm7[0,0],ymm1[5,4],ymm7[4,4]
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm7[2,0],ymm0[7,4],ymm7[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm7[2,1],ymm0[6,4],ymm7[6,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm9[2,0],ymm0[7,4],ymm9[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm9[2,1],ymm0[6,4],ymm9[6,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[1,0],ymm1[0,0],ymm2[5,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm11[2,0],ymm0[7,4],ymm11[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm11[2,1],ymm0[6,4],ymm11[6,5]
+; AVX1-ONLY-NEXT: vmovaps %ymm11, %ymm4
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
@@ -3959,17 +3961,15 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[1,0],ymm0[0,0],ymm4[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm3[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[1,0],ymm9[0,0],ymm3[5,4],ymm9[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm14[2,0],ymm0[7,4],ymm14[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm14[2,1],ymm1[6,4],ymm14[6,5]
-; AVX1-ONLY-NEXT: vmovaps %ymm14, %ymm6
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm2 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
@@ -3977,45 +3977,26 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm1[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm11[1,0],ymm0[0,0],ymm11[5,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[1,0],ymm12[0,0],ymm6[5,4],ymm12[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm5[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,0],ymm10[2,0],ymm0[7,4],ymm10[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm10[2,1],ymm2[6,4],ymm10[6,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,0],ymm13[2,0],ymm0[7,4],ymm13[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0],ymm13[2,1],ymm2[6,4],ymm13[6,5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm14[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm14[1,0],ymm0[0,0],ymm14[5,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm12[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm0[3,0],ymm13[2,0],ymm0[7,4],ymm13[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm13[2,1],ymm3[6,4],ymm13[6,5]
-; AVX1-ONLY-NEXT: vmovaps %ymm13, %ymm2
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm12 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm12[0,1],mem[2],xmm12[3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm12 = xmm12[1,0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2],ymm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm12[1,0],ymm1[0,0],ymm12[5,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm3
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2],xmm5[3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm5 = xmm5[1,0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm2[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[1,0],ymm1[0,0],ymm8[5,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
@@ -4023,15 +4004,14 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm5[0,0],ymm1[3,0],ymm5[4,4],ymm1[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm11[0,0],ymm1[3,0],ymm11[4,4],ymm1[7,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm13[2,0],mem[1,0],ymm13[6,4],mem[5,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[2,0],ymm10[1,0],ymm1[6,4],ymm10[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4039,9 +4019,9 @@
; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[0,0],ymm1[3,0],ymm8[4,4],ymm1[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm5[3,0],ymm1[4,4],ymm5[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm5[2,2],ymm15[6,4],ymm5[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -4053,9 +4033,10 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,0],ymm9[3,0],ymm15[4,4],ymm9[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm9[2,2],ymm15[6,4],ymm9[6,6]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm10[0,0],ymm1[3,0],ymm10[4,4],ymm1[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -4073,64 +4054,62 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[2,0],ymm7[1,0],ymm1[6,4],ymm7[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm10[0,0],ymm1[3,0],ymm10[4,4],ymm1[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm1[2,2],ymm15[6,4],ymm1[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $18, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm15 = ymm1[2,0],mem[1,0],ymm1[6,4],mem[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm2[3,0],ymm1[4,4],ymm2[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm2[2,2],ymm15[6,4],ymm2[6,6]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm5[0,0],ymm13[3,0],ymm5[4,4],ymm13[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm13[2,2],ymm15[6,4],ymm13[6,6]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm12[2,0],ymm3[1,0],ymm12[6,4],ymm3[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[2,0],ymm2[1,0],ymm8[6,4],ymm2[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm9[0,0],ymm10[3,0],ymm9[4,4],ymm10[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm10[2,2],ymm15[6,4],ymm10[6,6]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm14[2,0],ymm10[1,0],ymm14[6,4],ymm10[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm8
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm3[0,0],ymm6[3,0],ymm3[4,4],ymm6[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm6[2,2],ymm15[6,4],ymm6[6,6]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm11[2,0],ymm6[1,0],ymm11[6,4],ymm6[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm7[3,0],ymm1[4,4],ymm7[7,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm7[2,2],ymm15[6,4],ymm7[6,6]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[2,0],ymm2[1,0],ymm4[6,4],ymm2[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm2[0,0],ymm14[3,0],ymm2[4,4],ymm14[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm14[2,2],ymm15[6,4],ymm14[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm6[2,0],ymm12[1,0],ymm6[6,4],ymm12[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm4[3,0],ymm1[4,4],ymm4[7,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0],ymm4[2,2],ymm15[6,4],ymm4[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,0],mem[1,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm15[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm3[2,0],ymm9[1,0],ymm3[6,4],ymm9[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
@@ -4138,32 +4117,58 @@
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm2[0,1,2,3,4,5,6],ymm4[7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm5 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0],mem[1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm5 = mem[0,1,2,3,4,5,6],ymm13[7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm3[5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm11[5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm4 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0],mem[1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $128, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm4 = ymm1[0,1,2,3,4,5,6],mem[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm2[5],ymm4[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2],ymm5[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5,6],ymm11[7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2],ymm4[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5,6],ymm6[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps $32, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm4 = ymm4[0,1,2,3,4],mem[5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],mem[1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = mem[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm5[5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2,3,4,5,6],ymm8[7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: vblendps $32, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = ymm5[0,1,2,3,4],mem[5],ymm5[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
@@ -4172,48 +4177,23 @@
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],mem[1],xmm6[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = mem[0,1,2,3,4,5,6],ymm2[7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = mem[0,1,2,3,4,5,6],ymm6[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm9[5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3,4,5,6],ymm14[7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vblendps $32, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = ymm6[0,1,2,3,4],mem[5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm10[5],ymm6[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm7 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],mem[1],xmm7[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3,4,5,6],ymm3[7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2,3,4,5,6],ymm7[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vblendps $32, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = ymm7[0,1,2,3,4],mem[5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = mem[0,1,2,3],ymm8[4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0],mem[1],xmm8[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = mem[0,1,2,3,4,5,6],ymm12[7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: vblendps $32, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
@@ -4225,7 +4205,7 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0],mem[1],xmm8[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $128, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vblendps $128, (%rsp), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm8 = ymm8[0,1,2,3,4,5,6],mem[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
@@ -4261,14 +4241,14 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 128(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, (%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 224(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 160(%rcx)
@@ -4292,83 +4272,86 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%r8)
; AVX1-ONLY-NEXT: vmovaps %ymm7, 224(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 192(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 160(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 192(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%r9)
; AVX1-ONLY-NEXT: vmovaps %ymm2, 128(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%r9)
; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%r9)
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r9)
-; AVX1-ONLY-NEXT: addq $2488, %rsp # imm = 0x9B8
+; AVX1-ONLY-NEXT: addq $2472, %rsp # imm = 0x9A8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-ONLY-LABEL: load_i32_stride5_vf64:
; AVX2-ONLY: # %bb.0:
-; AVX2-ONLY-NEXT: subq $2152, %rsp # imm = 0x868
+; AVX2-ONLY-NEXT: subq $2120, %rsp # imm = 0x848
; AVX2-ONLY-NEXT: vmovdqa 864(%rdi), %ymm4
; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 896(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 832(%rdi), %ymm6
; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 800(%rdi), %ymm15
+; AVX2-ONLY-NEXT: vmovdqa 800(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 576(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 512(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 480(%rdi), %ymm11
; AVX2-ONLY-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vmovdqu %ymm12, (%rsp) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %ymm13
+; AVX2-ONLY-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = [0,5,2,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vinserti128 $1, 288(%rdi), %ymm1, %ymm2
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm14[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm13[0,1,0,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4],ymm3[5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5,6],ymm2[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm10[2,3],ymm11[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vmovdqa %ymm10, %ymm14
+; AVX2-ONLY-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm9[0,1,0,3]
+; AVX2-ONLY-NEXT: vmovdqa %ymm9, %ymm13
+; AVX2-ONLY-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4],ymm2[5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa %ymm8, %ymm9
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinserti128 $1, 608(%rdi), %ymm1, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm6[2,3],ymm15[4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm6[2,3],ymm7[4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm5[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovdqa %ymm5, %ymm8
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4],ymm2[5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinserti128 $1, 928(%rdi), %ymm1, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 1152(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovdqa 1120(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovdqa %ymm3, %ymm9
+; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa %ymm2, %ymm8
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1120(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vmovdqa 1184(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa 1216(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm12[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa %ymm3, %ymm11
; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 1216(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm4[0,1,0,3]
+; AVX2-ONLY-NEXT: vmovdqa %ymm4, %ymm11
+; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4],ymm2[5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vinserti128 $1, 1248(%rdi), %ymm1, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
@@ -4380,7 +4363,7 @@
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu %ymm3, (%rsp) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,1,0,3]
@@ -4405,11 +4388,11 @@
; AVX2-ONLY-NEXT: vinserti128 $1, 768(%rdi), %ymm1, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 992(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 960(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovdqa 992(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 960(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vmovdqa 1024(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4421,28 +4404,30 @@
; AVX2-ONLY-NEXT: vinserti128 $1, 1088(%rdi), %ymm1, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm15
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm15[2,3],ymm6[4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm13[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4],ymm1[5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm5[0,1,0,3]
+; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4],ymm1[5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vinserti128 $1, 128(%rdi), %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6],ymm0[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm0 = <1,6,3,u>
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm1 = ymm6[0,1],mem[2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm7[2,3],ymm12[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpblendd $51, (%rsp), %ymm14, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm3 = mem[0,1],ymm14[2,3],mem[4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = ymm10[0,1],mem[2,3],ymm10[4,5],mem[6,7]
; AVX2-ONLY-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [5,2,7,0,5,2,7,0]
; AVX2-ONLY-NEXT: # ymm1 = mem[0,1,0,1]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm1, %ymm3
@@ -4450,53 +4435,50 @@
; AVX2-ONLY-NEXT: vpbroadcastd 304(%rdi), %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0,1],ymm7[2,3],ymm14[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = mem[0,1],ymm14[2,3],mem[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm3 = ymm9[0,1],mem[2,3],ymm9[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1],ymm13[2,3],ymm14[4,5],ymm13[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm1, %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastd 624(%rdi), %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm2 = ymm15[0,1],mem[2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm8[2,3],ymm15[4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm1, %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastd 944(%rdi), %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm2 = mem[0,1],ymm8[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1],ymm8[2,3],ymm9[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm12[2,3],ymm11[4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm11[2,3],ymm9[4,5],ymm11[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm1, %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastd 1264(%rdi), %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1],ymm4[2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1],ymm15[2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovdqa %ymm10, %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1],ymm13[2,3],ymm10[4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm5[2,3],ymm4[4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm1, %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastd 144(%rdi), %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm2 = mem[0,1],ymm5[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = mem[0,1],ymm11[2,3],mem[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovdqu (%rsp), %ymm3 # 32-byte Reload
; AVX2-ONLY-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5],mem[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm1, %ymm3
@@ -4504,375 +4486,377 @@
; AVX2-ONLY-NEXT: vpbroadcastd 464(%rdi), %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1],ymm11[2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm3 = ymm13[0,1],mem[2,3],ymm13[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm1, %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastd 784(%rdi), %ymm3
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1],ymm8[2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm1, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastd 1104(%rdi), %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm3 = <2,7,4,u>
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm0, %ymm3, %ymm0
+; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm2 = <2,7,4,u>
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm7[4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm0, %ymm2, %ymm0
; AVX2-ONLY-NEXT: vinserti128 $1, 256(%rdi), %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vmovdqu (%rsp), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovdqa %ymm10, %ymm4
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3,4,5,6],ymm1[7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,3,0,1,6,7,4,5]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm0 = [0,5,0,5,0,5,0,5]
-; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm7[4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm3, %ymm1
-; AVX2-ONLY-NEXT: vinserti128 $1, 576(%rdi), %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm10[0,1,2,3],mem[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vinserti128 $1, 576(%rdi), %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm14[4,5],mem[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm3, %ymm1
-; AVX2-ONLY-NEXT: vinserti128 $1, 896(%rdi), %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 928(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vinserti128 $1, 896(%rdi), %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 928(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1,2,3],ymm8[4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm3, %ymm1
-; AVX2-ONLY-NEXT: vinserti128 $1, 1216(%rdi), %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 1248(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vinserti128 $1, 1216(%rdi), %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 1248(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3],ymm8[4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vinserti128 $1, 1056(%rdi), %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 1088(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vpermd %ymm8, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm3, %ymm1
-; AVX2-ONLY-NEXT: vinserti128 $1, 96(%rdi), %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm2, %ymm1
+; AVX2-ONLY-NEXT: vinserti128 $1, 96(%rdi), %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = mem[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2],ymm3[3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm2 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm3, %ymm2
-; AVX2-ONLY-NEXT: vinserti128 $1, 416(%rdi), %ymm0, %ymm4
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm4 = mem[0,1,2,3,4,5,6],ymm4[7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm2[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm15
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm12[0,1,2,3],ymm11[4,5],ymm12[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm3, %ymm4
-; AVX2-ONLY-NEXT: vinserti128 $1, 736(%rdi), %ymm0, %ymm15
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm15 = ymm13[0,1,2,3,4,5,6],ymm15[7]
-; AVX2-ONLY-NEXT: vmovdqa %ymm13, %ymm7
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm15 = ymm15[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm15 = ymm4[0,1,2],ymm15[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 768(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm13
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm2[0,1,2,3],ymm5[4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm13, %ymm3, %ymm3
-; AVX2-ONLY-NEXT: vinserti128 $1, 1056(%rdi), %ymm0, %ymm13
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3,4,5,6],ymm13[7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[2,3,0,1,6,7,4,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm3[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 1088(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vmovdqa %ymm3, %ymm4
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = mem[0,1,2,3],ymm11[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm2, %ymm3
+; AVX2-ONLY-NEXT: vinserti128 $1, 416(%rdi), %ymm0, %ymm15
+; AVX2-ONLY-NEXT: vpblendd $127, (%rsp), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = mem[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm15 = ymm15[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm15 = ymm3[0,1,2],ymm15[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm13
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm11[0,1,2,3],ymm9[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm13, %ymm2, %ymm2
+; AVX2-ONLY-NEXT: vinserti128 $1, 736(%rdi), %ymm0, %ymm13
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[2,3,0,1,6,7,4,5]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm2[0,1,2],ymm13[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 768(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm0, %ymm0
+; AVX2-ONLY-NEXT: vmovdqa %ymm2, %ymm3
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm10[0,1,2,3],mem[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,0,2,2,7,4,6,6]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm13 = mem[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm13 = ymm12[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],ymm12[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm13 = ymm13[0,3,2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2],ymm0[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm0 = [1,6,1,6,1,6,1,6]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm13 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm13 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm15 = mem[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm10[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],ymm10[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm13 = ymm3[0,1,2,3],mem[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm13 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
-; AVX2-ONLY-NEXT: vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm15 = mem[12,13,14,15],ymm14[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm14[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = mem[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm13 = ymm6[0,1,2,3],mem[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm13 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm9[12,13,14,15],ymm8[0,1,2,3,4,5,6,7,8,9,10,11],ymm9[28,29,30,31],ymm8[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm7[12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10,11],ymm7[28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm13 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm13 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm2[12,13,14,15],ymm5[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm5[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpalignr $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = mem[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],mem[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm0, %ymm15
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm13 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm8, %ymm0, %ymm15
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm12[4,5],ymm14[6,7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm12[12,13,14,15],ymm11[0,1,2,3,4,5,6,7,8,9,10,11],ymm12[28,29,30,31],ymm11[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm11[12,13,14,15],ymm9[0,1,2,3,4,5,6,7,8,9,10,11],ymm11[28,29,30,31],ymm9[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vmovdqa %ymm11, %ymm14
+; AVX2-ONLY-NEXT: vmovdqa %ymm9, %ymm11
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermd %ymm14, %ymm0, %ymm15
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm10[0,1,2,3],ymm9[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm0, %ymm15
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm9[0,1,2,3],ymm8[4,5],ymm9[6,7]
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm3[12,13,14,15],ymm2[0,1,2,3,4,5,6,7,8,9,10,11],ymm3[28,29,30,31],ymm2[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm0, %ymm15
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm7[0,1,2,3],ymm1[4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpermd %ymm5, %ymm0, %ymm15
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm10[0,1,2,3],ymm7[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[3,0,2,2,7,4,6,6]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm5[12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10,11],ymm5[28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm4[12,13,14,15],ymm6[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm6[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[0,3,2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpermd %ymm15, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm7[0,1],ymm1[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm8 = <4,1,6,u>
-; AVX2-ONLY-NEXT: vpermd %ymm5, %ymm8, %ymm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm7 = [2,7,2,7,2,7,2,7]
-; AVX2-ONLY-NEXT: vpermd %ymm15, %ymm7, %ymm1
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm0, %ymm0
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm15 = ymm13[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm10[0,1],ymm7[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0,1,2,3],ymm4[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vmovdqa {{.*#+}} xmm7 = <4,1,6,u>
+; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm7, %ymm4
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq {{.*#+}} ymm6 = [2,7,2,7,2,7,2,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm6, %ymm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vperm2i128 $2, (%rsp), %ymm6, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm5 = mem[0,1],ymm6[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm6[5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm8, %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = mem[0,1],ymm10[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm10[5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm7, %ymm1
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm5 = ymm10[0,1],ymm9[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm9[5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm8, %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm7, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm4 = ymm9[0,1],ymm8[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm8[5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm1, %ymm7, %ymm1
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm5, %ymm6, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm5 = mem[0,1],ymm9[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm9[5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm8, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm2[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1,2,3],ymm12[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = mem[0,1,2,3],ymm2[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = mem[0,1],ymm8[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm8[5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm7, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm14[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm8 = mem[0,1],ymm12[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm12[5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm7, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm9 = mem[0,1],ymm3[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm3[5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm2, %ymm8, %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm9[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm14, %ymm7, %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm4 = ymm3[0,1,2,3],mem[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm8 = mem[0,1],ymm10[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm3, %ymm7, %ymm3
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm8 = ymm5[0,1,2,3],mem[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm10 = mem[0,1],ymm5[0,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm5[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm8, %ymm7, %ymm8
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm9 = mem[0,1,2,3],ymm5[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vpermd %ymm9, %ymm7, %ymm7
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm9 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm9 = mem[0,1],ymm10[0,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm4, %ymm8, %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2],ymm9[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm9 = ymm3[0,1,2,3],mem[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm10 = mem[0,1],ymm11[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm11[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm9, %ymm8, %ymm9
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2],ymm10[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm9 = mem[0,1,2,3],ymm9[4,5],mem[6,7]
-; AVX2-ONLY-NEXT: vpermd %ymm9, %ymm8, %ymm8
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-ONLY-NEXT: vperm2i128 $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm9 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm9 = mem[0,1],ymm10[0,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 192(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 128(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 64(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, (%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 224(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 160(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 96(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 192(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 128(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 64(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, (%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 224(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 160(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 96(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 192(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 128(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 64(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, (%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 224(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 160(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 96(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 32(%rcx)
-; AVX2-ONLY-NEXT: vmovdqa %ymm13, (%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 64(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 128(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 192(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 224(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 160(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 96(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm8, 32(%r8)
-; AVX2-ONLY-NEXT: vmovdqa %ymm7, 224(%r9)
-; AVX2-ONLY-NEXT: vmovdqa %ymm3, 192(%r9)
-; AVX2-ONLY-NEXT: vmovdqa %ymm4, 160(%r9)
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 192(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 128(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm7, 64(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm7, 224(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm7, 160(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm7, 96(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm7, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 192(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 128(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 224(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 160(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 128(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 192(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 224(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 160(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovdqa %ymm15, (%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 128(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 192(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 224(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 160(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%r8)
+; AVX2-ONLY-NEXT: vmovdqa %ymm6, 224(%r9)
+; AVX2-ONLY-NEXT: vmovdqa %ymm8, 192(%r9)
+; AVX2-ONLY-NEXT: vmovdqa %ymm3, 160(%r9)
; AVX2-ONLY-NEXT: vmovdqa %ymm2, 128(%r9)
-; AVX2-ONLY-NEXT: vmovdqa %ymm5, 96(%r9)
+; AVX2-ONLY-NEXT: vmovdqa %ymm4, 96(%r9)
; AVX2-ONLY-NEXT: vmovdqa %ymm1, 64(%r9)
-; AVX2-ONLY-NEXT: vmovdqa %ymm6, 32(%r9)
+; AVX2-ONLY-NEXT: vmovdqa %ymm13, 32(%r9)
; AVX2-ONLY-NEXT: vmovdqa %ymm0, (%r9)
-; AVX2-ONLY-NEXT: addq $2152, %rsp # imm = 0x868
+; AVX2-ONLY-NEXT: addq $2120, %rsp # imm = 0x848
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
index d179de0..bcffe37 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
@@ -106,16 +106,14 @@
; AVX512F-SLOW: # %bb.0:
; AVX512F-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512F-SLOW-NEXT: vmovdqa (%rdi), %xmm0
-; AVX512F-SLOW-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX512F-SLOW-NEXT: vmovdqa 16(%rdi), %xmm1
; AVX512F-SLOW-NEXT: vmovdqa 32(%rdi), %xmm2
-; AVX512F-SLOW-NEXT: vextractps $2, %xmm1, %r10d
+; AVX512F-SLOW-NEXT: vpextrd $2, %xmm1, %r10d
; AVX512F-SLOW-NEXT: vpinsrd $1, %r10d, %xmm0, %xmm3
-; AVX512F-SLOW-NEXT: vextractps $3, %xmm1, %r10d
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512F-SLOW-NEXT: vpinsrd $1, %r10d, %xmm1, %xmm1
-; AVX512F-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm4
-; AVX512F-SLOW-NEXT: vmovd %xmm2, %r10d
-; AVX512F-SLOW-NEXT: vpinsrd $1, %r10d, %xmm4, %xmm4
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3]
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm2[0,1],xmm0[2,3]
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,0,2,3]
; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
; AVX512F-SLOW-NEXT: vmovddup {{.*#+}} xmm2 = [4,2,4,2]
@@ -170,16 +168,14 @@
; AVX512BW-SLOW: # %bb.0:
; AVX512BW-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX512BW-SLOW-NEXT: vmovdqa (%rdi), %xmm0
-; AVX512BW-SLOW-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX512BW-SLOW-NEXT: vmovdqa 16(%rdi), %xmm1
; AVX512BW-SLOW-NEXT: vmovdqa 32(%rdi), %xmm2
-; AVX512BW-SLOW-NEXT: vextractps $2, %xmm1, %r10d
+; AVX512BW-SLOW-NEXT: vpextrd $2, %xmm1, %r10d
; AVX512BW-SLOW-NEXT: vpinsrd $1, %r10d, %xmm0, %xmm3
-; AVX512BW-SLOW-NEXT: vextractps $3, %xmm1, %r10d
-; AVX512BW-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512BW-SLOW-NEXT: vpinsrd $1, %r10d, %xmm1, %xmm1
-; AVX512BW-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm4
-; AVX512BW-SLOW-NEXT: vmovd %xmm2, %r10d
-; AVX512BW-SLOW-NEXT: vpinsrd $1, %r10d, %xmm4, %xmm4
+; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3]
+; AVX512BW-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
+; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm2[0,1],xmm0[2,3]
+; AVX512BW-SLOW-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,0,2,3]
; AVX512BW-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3]
; AVX512BW-SLOW-NEXT: vmovddup {{.*#+}} xmm2 = [4,2,4,2]
@@ -953,21 +949,22 @@
define void @load_i32_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
; SSE-LABEL: load_i32_stride6_vf16:
; SSE: # %bb.0:
-; SSE-NEXT: subq $360, %rsp # imm = 0x168
-; SSE-NEXT: movdqa 240(%rdi), %xmm15
+; SSE-NEXT: subq $376, %rsp # imm = 0x178
+; SSE-NEXT: movdqa 64(%rdi), %xmm4
+; SSE-NEXT: movdqa (%rdi), %xmm15
; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 256(%rdi), %xmm5
-; SSE-NEXT: movdqa 192(%rdi), %xmm13
-; SSE-NEXT: movdqa 208(%rdi), %xmm14
+; SSE-NEXT: movdqa 16(%rdi), %xmm5
+; SSE-NEXT: movdqa 48(%rdi), %xmm14
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 336(%rdi), %xmm11
; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 352(%rdi), %xmm7
-; SSE-NEXT: movdqa 288(%rdi), %xmm12
+; SSE-NEXT: movdqa 288(%rdi), %xmm13
; SSE-NEXT: movdqa 304(%rdi), %xmm8
-; SSE-NEXT: movdqa 64(%rdi), %xmm3
-; SSE-NEXT: movdqa (%rdi), %xmm0
-; SSE-NEXT: movdqa 16(%rdi), %xmm2
-; SSE-NEXT: movdqa 48(%rdi), %xmm4
+; SSE-NEXT: movdqa 240(%rdi), %xmm6
+; SSE-NEXT: movdqa 256(%rdi), %xmm3
+; SSE-NEXT: movdqa 192(%rdi), %xmm0
+; SSE-NEXT: movdqa 208(%rdi), %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
; SSE-NEXT: movdqa %xmm2, %xmm10
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -976,14 +973,13 @@
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[2,2,3,3]
; SSE-NEXT: movdqa %xmm3, %xmm9
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3]
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm12, %xmm2
-; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm13, %xmm2
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[2,2,3,3]
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -991,105 +987,105 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3]
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm15, %xmm2
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm5, %xmm3
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm15[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm2[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 96(%rdi), %xmm5
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm4, %xmm3
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm14[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
+; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 96(%rdi), %xmm14
; SSE-NEXT: movdqa 112(%rdi), %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
; SSE-NEXT: movdqa %xmm2, %xmm11
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm5, %xmm2
+; SSE-NEXT: movdqa %xmm14, %xmm2
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT: movdqa 144(%rdi), %xmm15
-; SSE-NEXT: movdqa 160(%rdi), %xmm6
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[2,2,3,3]
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm15[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm2[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 160(%rdi), %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
+; SSE-NEXT: movdqa %xmm4, %xmm12
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm15[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
+; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm10[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm9[2,3,2,3]
-; SSE-NEXT: movdqa %xmm4, %xmm9
-; SSE-NEXT: movdqa %xmm4, %xmm5
-; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm6, %xmm4
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
+; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm13, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm8[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm7, %xmm5
-; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm8, %xmm4
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
+; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm14[3,3,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm5, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm4, %xmm3
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[1,1,1,1]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm11[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm12[2,3,2,3]
; SSE-NEXT: movdqa %xmm15, %xmm3
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm4[2,2,3,3]
-; SSE-NEXT: movdqa 80(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[2,2,3,3]
+; SSE-NEXT: movdqa 272(%rdi), %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,0,1,1]
+; SSE-NEXT: movdqa %xmm3, %xmm11
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; SSE-NEXT: movdqa 32(%rdi), %xmm10
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
+; SSE-NEXT: movdqa 224(%rdi), %xmm6
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[2,2,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm8[2,2,3,3]
; SSE-NEXT: movdqa 368(%rdi), %xmm3
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,0,1,1]
-; SSE-NEXT: movdqa %xmm3, %xmm12
+; SSE-NEXT: movdqa %xmm3, %xmm10
; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3]
-; SSE-NEXT: movdqa 320(%rdi), %xmm4
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[2,3,2,3]
+; SSE-NEXT: movdqa 320(%rdi), %xmm7
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[2,2,3,3]
-; SSE-NEXT: movdqa %xmm5, %xmm11
-; SSE-NEXT: movdqa 272(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[2,2,3,3]
+; SSE-NEXT: movdqa 80(%rdi), %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,0,1,1]
+; SSE-NEXT: movdqa %xmm2, %xmm12
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[2,3,2,3]
-; SSE-NEXT: movdqa %xmm13, %xmm5
-; SSE-NEXT: movdqa 224(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm13, %xmm4
+; SSE-NEXT: movdqa 32(%rdi), %xmm2
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm15[2,2,3,3]
-; SSE-NEXT: movdqa 176(%rdi), %xmm6
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,0,1,1]
+; SSE-NEXT: movdqa 176(%rdi), %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3]
; SSE-NEXT: movdqa 128(%rdi), %xmm13
@@ -1097,145 +1093,152 @@
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[2,3,2,3]
+; SSE-NEXT: movdqa %xmm11, %xmm3
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[2,3,2,3]
-; SSE-NEXT: movdqa %xmm12, (%rsp) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3]
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
+; SSE-NEXT: movdqa %xmm2, %xmm8
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT: # xmm11 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm12[0],xmm11[1],xmm12[1]
+; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[3,3,3,3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
+; SSE-NEXT: movdqa %xmm13, %xmm2
+; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm15[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1]
-; SSE-NEXT: movdqa %xmm6, %xmm8
+; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1]
+; SSE-NEXT: movdqa %xmm5, (%rsp) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm7, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
-; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT: # xmm15 = mem[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
; SSE-NEXT: movdqa %xmm6, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm9[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm7, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm12[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm5, %xmm0
+; SSE-NEXT: movdqa %xmm8, %xmm13
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[0,0,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm10[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm8[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm10[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm2[0],xmm10[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm1[0],xmm10[1]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT: pshufd $238, (%rsp), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm2[0],xmm12[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm1[0],xmm12[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm2[0],xmm9[1],xmm2[1]
; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
+; SSE-NEXT: movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm2[0],xmm12[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm1[0],xmm12[1]
; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[1,1,1,1]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm13[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm8[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm1[0],xmm11[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm2[0],xmm10[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm1[0],xmm10[1]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT: pshufd $238, (%rsp), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%rsi)
+; SSE-NEXT: movaps %xmm1, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%rsi)
+; SSE-NEXT: movaps %xmm1, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%rdx)
+; SSE-NEXT: movaps %xmm1, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%rdx)
+; SSE-NEXT: movaps %xmm1, 32(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%rcx)
+; SSE-NEXT: movaps %xmm1, (%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 48(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%rcx)
+; SSE-NEXT: movaps %xmm1, 32(%rcx)
; SSE-NEXT: movapd %xmm14, 16(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%r8)
+; SSE-NEXT: movapd %xmm11, (%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 48(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%r8)
+; SSE-NEXT: movaps %xmm1, 32(%r8)
; SSE-NEXT: movapd %xmm0, 16(%r9)
-; SSE-NEXT: movapd %xmm3, 32(%r9)
+; SSE-NEXT: movapd %xmm3, (%r9)
; SSE-NEXT: movapd %xmm4, 48(%r9)
-; SSE-NEXT: movapd %xmm15, (%r9)
+; SSE-NEXT: movapd %xmm15, 32(%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movapd %xmm11, 16(%rax)
-; SSE-NEXT: movapd %xmm9, 32(%rax)
-; SSE-NEXT: movapd %xmm12, 48(%rax)
+; SSE-NEXT: movapd %xmm5, 16(%rax)
; SSE-NEXT: movapd %xmm10, (%rax)
-; SSE-NEXT: addq $360, %rsp # imm = 0x168
+; SSE-NEXT: movapd %xmm12, 48(%rax)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%rax)
+; SSE-NEXT: addq $376, %rsp # imm = 0x178
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i32_stride6_vf16:
@@ -2152,370 +2155,372 @@
define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
; SSE-LABEL: load_i32_stride6_vf32:
; SSE: # %bb.0:
-; SSE-NEXT: subq $1016, %rsp # imm = 0x3F8
-; SSE-NEXT: movdqa 64(%rdi), %xmm10
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa (%rdi), %xmm11
+; SSE-NEXT: subq $1032, %rsp # imm = 0x408
+; SSE-NEXT: movdqa 336(%rdi), %xmm11
; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 16(%rdi), %xmm7
+; SSE-NEXT: movdqa 352(%rdi), %xmm5
+; SSE-NEXT: movdqa 288(%rdi), %xmm12
+; SSE-NEXT: movdqa 304(%rdi), %xmm6
+; SSE-NEXT: movdqa 624(%rdi), %xmm7
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 48(%rdi), %xmm14
-; SSE-NEXT: movdqa 528(%rdi), %xmm6
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 544(%rdi), %xmm4
-; SSE-NEXT: movdqa 480(%rdi), %xmm8
+; SSE-NEXT: movdqa 640(%rdi), %xmm3
+; SSE-NEXT: movdqa 576(%rdi), %xmm8
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 496(%rdi), %xmm15
-; SSE-NEXT: movdqa 144(%rdi), %xmm9
+; SSE-NEXT: movdqa 592(%rdi), %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 240(%rdi), %xmm9
; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 160(%rdi), %xmm12
-; SSE-NEXT: movdqa 96(%rdi), %xmm1
+; SSE-NEXT: movdqa 256(%rdi), %xmm2
+; SSE-NEXT: movdqa 192(%rdi), %xmm10
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 208(%rdi), %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE-NEXT: movdqa %xmm1, %xmm13
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
-; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT: movdqa %xmm2, %xmm14
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm9[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
; SSE-NEXT: movdqa %xmm8, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
-; SSE-NEXT: movdqa %xmm4, %xmm5
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
-; SSE-NEXT: movdqa %xmm11, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
+; SSE-NEXT: movdqa %xmm3, %xmm9
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; SSE-NEXT: movdqa %xmm6, %xmm8
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm14[0,0,1,1]
-; SSE-NEXT: movdqa %xmm14, %xmm7
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 384(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, (%rsp) # 16-byte Spill
-; SSE-NEXT: movdqa 400(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT: movdqa %xmm5, %xmm4
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm11[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 672(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 688(%rdi), %xmm15
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
+; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 432(%rdi), %xmm2
+; SSE-NEXT: movdqa 720(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 736(%rdi), %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT: movdqa %xmm2, %xmm11
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT: movdqa 16(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movdqa 64(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 48(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 384(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 400(%rdi), %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm3, %xmm2
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movdqa 432(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 448(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 288(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 304(%rdi), %xmm6
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 336(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 352(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 672(%rdi), %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 688(%rdi), %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm3, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 720(%rdi), %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 736(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 208(%rdi), %xmm13
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[2,3,2,3]
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: movdqa 240(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 256(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[2,2,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm1[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1]
-; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm3[0],xmm11[1]
-; SSE-NEXT: movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 576(%rdi), %xmm1
-; SSE-NEXT: movdqa 592(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
-; SSE-NEXT: movdqa %xmm1, %xmm4
-; SSE-NEXT: movdqa %xmm1, %xmm11
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT: movdqa 624(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 640(%rdi), %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[2,2,3,3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm1[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm3[0],xmm10[1],xmm3[1]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm4[0],xmm10[1]
-; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[1,1,1,1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm12[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm8, %xmm10
-; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm3[0],xmm10[1]
-; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm15[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm5[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm3[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm14[1,1,1,1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
+; SSE-NEXT: movdqa 96(%rdi), %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 112(%rdi), %xmm7
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[2,3,2,3]
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm7, %xmm5
-; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm3[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $85, (%rsp), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm15, %xmm5
-; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm3[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm12[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm6[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm6
-; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm3[0],xmm6[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT: movdqa 144(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 160(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,2,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm3[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm5[0],xmm6[1]
; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm2[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm6, %xmm2
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm3[0],xmm2[1]
+; SSE-NEXT: movdqa 480(%rdi), %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 496(%rdi), %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT: movdqa 528(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 544(%rdi), %xmm10
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm10[2,2,3,3]
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[1,1,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm13[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm1, %xmm2
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm3[0],xmm2[1]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm13[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm14[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm5[0],xmm14[1]
+; SSE-NEXT: movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm13[1,1,1,1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm9[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm14, %xmm2
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[1,1,1,1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm11, %xmm0
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm12[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm8[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm4[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm8[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm15[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm11[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm4, %xmm2
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $85, (%rsp), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm15, %xmm2
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm3[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm1[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm7[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = mem[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm9, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm7[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm0[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm10[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm10, %xmm0
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
-; SSE-NEXT: movdqa 176(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[2,3,2,3]
-; SSE-NEXT: movdqa 128(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
+; SSE-NEXT: movdqa 656(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm13[2,3,2,3]
+; SSE-NEXT: movdqa 608(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
-; SSE-NEXT: movdqa 80(%rdi), %xmm13
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm13[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm14[2,3,2,3]
-; SSE-NEXT: movdqa 32(%rdi), %xmm2
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT: movdqa %xmm2, %xmm7
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT: movdqa 752(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm8[2,3,2,3]
+; SSE-NEXT: movdqa 704(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT: movdqa 464(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm3[2,3,2,3]
+; SSE-NEXT: movdqa 416(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
-; SSE-NEXT: movdqa 368(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm12[2,3,2,3]
-; SSE-NEXT: movdqa 320(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3]
-; SSE-NEXT: movdqa 272(%rdi), %xmm1
+; SSE-NEXT: movdqa 560(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[2,3,2,3]
-; SSE-NEXT: movdqa %xmm5, %xmm10
-; SSE-NEXT: movdqa 224(%rdi), %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm7[2,3,2,3]
+; SSE-NEXT: movdqa 512(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[2,2,3,3]
-; SSE-NEXT: movdqa 560(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[2,3,2,3]
-; SSE-NEXT: movdqa 512(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
-; SSE-NEXT: movdqa 464(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: movdqa (%rsp), %xmm5 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[2,3,2,3]
-; SSE-NEXT: movdqa 416(%rdi), %xmm4
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
-; SSE-NEXT: movdqa 752(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[2,3,2,3]
-; SSE-NEXT: movdqa 704(%rdi), %xmm15
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm15[0],xmm3[1],xmm15[1]
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
-; SSE-NEXT: movdqa 656(%rdi), %xmm12
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm12[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[2,3,2,3]
-; SSE-NEXT: movdqa 608(%rdi), %xmm8
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm14[3,3,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[1,1,1,1]
-; SSE-NEXT: movdqa %xmm7, %xmm11
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
-; SSE-NEXT: movdqa %xmm1, %xmm14
+; SSE-NEXT: movdqa 272(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm12[2,3,2,3]
+; SSE-NEXT: movdqa 224(%rdi), %xmm11
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm11[0],xmm5[1],xmm11[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
+; SSE-NEXT: movdqa 368(%rdi), %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm3[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm13[2,3,2,3]
+; SSE-NEXT: movdqa 320(%rdi), %xmm10
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm10[0],xmm5[1],xmm10[1]
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm15, %xmm7
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
+; SSE-NEXT: movdqa 80(%rdi), %xmm8
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm8[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: movdqa (%rsp), %xmm4 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm4[2,3,2,3]
+; SSE-NEXT: movdqa 32(%rdi), %xmm2
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
+; SSE-NEXT: movdqa 176(%rdi), %xmm15
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm15[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm6[2,3,2,3]
+; SSE-NEXT: movdqa 128(%rdi), %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[3,3,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm2[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[3,3,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[1,1,1,1]
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
+; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[3,3,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT: movdqa %xmm3, %xmm9
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
@@ -2523,18 +2528,10 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
-; SSE-NEXT: movdqa %xmm12, %xmm8
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
@@ -2542,77 +2539,76 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm15[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
-; SSE-NEXT: movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
+; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
+; SSE-NEXT: movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
+; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT: # xmm14 = mem[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm1[0],xmm14[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
-; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = mem[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,3,2,3]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,2,3,3]
+; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm13 = xmm3[0,0,1,1]
-; SSE-NEXT: movdqa %xmm3, %xmm14
-; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm1[0],xmm13[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm9[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
; SSE-NEXT: # xmm6 = mem[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm10[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
+; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = mem[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm7[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,2,3,3]
-; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = mem[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm15[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
-; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm12[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[1,1,1,1]
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -2620,178 +2616,178 @@
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT: movapd %xmm15, %xmm12
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, %xmm10
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = mem[1,1,1,1]
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
-; SSE-NEXT: movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
-; SSE-NEXT: movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 96(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 112(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 64(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 80(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 96(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 112(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 80(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 96(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 112(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 64(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 80(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 112(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 96(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 80(%r8)
-; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 64(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%r8)
-; SSE-NEXT: movapd %xmm2, 112(%r9)
-; SSE-NEXT: movapd %xmm3, 96(%r9)
-; SSE-NEXT: movapd %xmm4, 80(%r9)
-; SSE-NEXT: movapd %xmm5, 64(%r9)
-; SSE-NEXT: movapd %xmm6, 48(%r9)
-; SSE-NEXT: movapd %xmm13, 32(%r9)
-; SSE-NEXT: movapd %xmm9, 16(%r9)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%r9)
+; SSE-NEXT: punpckldq {{.*#+}} xmm12 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm2[0],xmm12[1]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 16(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 64(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, (%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 48(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 96(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 80(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 16(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 64(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, (%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 48(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 96(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 16(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, (%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 80(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 64(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 96(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 96(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 80(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 64(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 48(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 32(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 16(%r8)
+; SSE-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, (%r8)
+; SSE-NEXT: movapd %xmm3, 112(%r9)
+; SSE-NEXT: movapd %xmm4, 96(%r9)
+; SSE-NEXT: movapd %xmm5, 80(%r9)
+; SSE-NEXT: movapd %xmm6, 64(%r9)
+; SSE-NEXT: movapd %xmm7, 48(%r9)
+; SSE-NEXT: movapd %xmm8, 32(%r9)
+; SSE-NEXT: movapd %xmm14, 16(%r9)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, (%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movapd %xmm14, 112(%rax)
-; SSE-NEXT: movapd %xmm10, 96(%rax)
-; SSE-NEXT: movapd %xmm7, 80(%rax)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 64(%rax)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rax)
+; SSE-NEXT: movapd %xmm12, 112(%rax)
+; SSE-NEXT: movapd %xmm15, 96(%rax)
+; SSE-NEXT: movapd %xmm11, 80(%rax)
+; SSE-NEXT: movapd %xmm13, 64(%rax)
+; SSE-NEXT: movapd %xmm9, 48(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 32(%rax)
-; SSE-NEXT: movapd %xmm15, 16(%rax)
-; SSE-NEXT: movapd %xmm12, (%rax)
-; SSE-NEXT: addq $1016, %rsp # imm = 0x3F8
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%rax)
+; SSE-NEXT: movapd %xmm10, (%rax)
+; SSE-NEXT: addq $1032, %rsp # imm = 0x408
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i32_stride6_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1064, %rsp # imm = 0x428
+; AVX1-ONLY-NEXT: subq $1016, %rsp # imm = 0x3F8
; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm7
; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm6
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm9
@@ -2808,20 +2804,20 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm11[0,1],xmm6[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm6[0,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm2[0,1]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[3],ymm3[2]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm5[2,3],ymm4[0,1]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm12[0],ymm5[0],ymm12[3],ymm5[2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rdi), %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm4[2,0],ymm5[0,0],ymm4[6,4],ymm5[4,4]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%rdi), %ymm3, %ymm5
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[2,0],ymm5[0,0],ymm2[6,4],ymm5[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm5[2,2],ymm0[6,4],ymm5[6,6]
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm7[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovapd 544(%rdi), %ymm1
@@ -2866,41 +2862,41 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, 672(%rdi), %ymm0, %ymm9
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[2,0],ymm9[0,0],ymm1[6,4],ymm9[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm0[2,0],ymm9[2,2],ymm0[6,4],ymm9[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm0[2,0],ymm9[2,2],ymm0[6,4],ymm9[6,6]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm7[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[0,2],xmm1[0,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2],ymm12[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovapd 736(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vmovupd %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm7[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[0,2],xmm1[0,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 736(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm0[2,3],ymm12[0,1]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm12[0],ymm0[0],ymm12[3],ymm0[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm15[3,0],ymm8[1,0],ymm15[7,4],ymm8[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm14[2,0],ymm8[2,3],ymm14[6,4],ymm8[6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm0[2,3],ymm13[0,1]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm15[0],ymm0[0],ymm15[3],ymm0[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm14[3,0],ymm8[1,0],ymm14[7,4],ymm8[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm13[2,0],ymm8[2,3],ymm13[6,4],ymm8[6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm11[1,0],xmm6[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm11[0,2],xmm6[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[3,1],ymm11[1,3],ymm0[7,5],ymm11[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm12[3,1],ymm11[1,3],ymm12[7,5],ymm11[5,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm14[3,0],ymm5[1,0],ymm14[7,4],ymm5[5,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm12[3,0],ymm5[1,0],ymm12[7,4],ymm5[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[2,0],ymm5[2,3],ymm6[6,4],ymm5[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm13[1,0],xmm4[3,0]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm0[1,0],xmm4[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm6[0,2],xmm4[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
@@ -2927,16 +2923,16 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm7[1,0],xmm1[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm2[0,2],xmm1[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[3,1],ymm13[1,3],ymm12[7,5],ymm13[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,1],ymm7[1,3],ymm15[7,5],ymm7[5,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm15[2,1],mem[2,0],ymm15[6,5],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm14[2,1],mem[2,0],ymm14[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2952,8 +2948,8 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm14[2,1],mem[2,0],ymm14[6,5],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm12[2,1],mem[2,0],ymm12[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,0,1]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2961,23 +2957,23 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm10 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm10 = ymm8[0,1,2,3],mem[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm10[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm7[0,0],ymm10[2,0],ymm7[4,4],ymm10[6,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm10[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm0[0,0],ymm10[2,0],ymm0[4,4],ymm10[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm6[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm5[2,1],ymm11[2,0],ymm5[6,5],ymm11[6,4]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm11 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = ymm5[2,1],mem[2,0],ymm5[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm6[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm0[2,0],xmm1[2,3]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm11[2,0],xmm0[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm9[0,1,2],ymm8[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = ymm13[0,1,2,3],mem[4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm8[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm9[0,0],ymm8[2,0],ymm9[4,4],ymm8[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3,4],ymm13[5,6,7]
@@ -3004,207 +3000,206 @@
; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = xmm3[3,1],mem[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[3,1],ymm13[2,1],ymm5[7,5],ymm13[6,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm13[3,1],ymm7[2,1],ymm13[7,5],ymm7[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm5[2,3,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm5[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1],ymm10[3,1],ymm7[4,5],ymm10[7,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,1],ymm10[3,1],ymm3[4,5],ymm10[7,5]
; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = xmm2[3,1],mem[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[3,1],ymm7[2,1],ymm5[7,5],ymm7[6,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[3,1],ymm10[2,1],ymm5[7,5],ymm10[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm4[2,3,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm15[3,1],ymm0[4,5],ymm15[7,5]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm12[3,1],xmm14[3,3]
-; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm12
-; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm15
+; AVX1-ONLY-NEXT: vmovaps %ymm1, %ymm4
+; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm14
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[3,1],ymm6[2,1],ymm1[7,5],ymm6[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm9[0,1],ymm8[3,1],ymm9[4,5],ymm8[7,5]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[3,1],mem[3,3]
+; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm11[3,1],mem[3,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[3,1],ymm11[2,1],ymm2[7,5],ymm11[6,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[3,1],ymm15[2,1],ymm2[7,5],ymm15[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vmovapd 464(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm7[0],ymm1[2],ymm7[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,0],ymm5[4,5],ymm1[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm5[0,0],ymm2[6,4],ymm5[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,2],ymm1[2,0],ymm5[4,6],ymm1[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm14[0],xmm0[1],xmm14[2,3]
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm13[0],ymm1[2],ymm13[2]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,1],ymm1[2,0],ymm10[4,5],ymm1[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[2,0],ymm0[0,0],ymm4[6,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[0,2],ymm2[2,0],ymm0[4,6],ymm2[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,2,3,3]
-; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm7[0],ymm1[2],ymm7[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[0,1],ymm1[2,0],ymm13[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm8[2,0],ymm1[0,0],ymm8[6,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[0,2],ymm2[2,0],ymm1[4,6],ymm2[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vmovapd 272(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vmovapd 464(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1],ymm15[0],ymm2[2],ymm15[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1],ymm2[2,0],ymm12[4,5],ymm2[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm6[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[2,0],ymm1[0,0],ymm6[6,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[0,2],ymm3[2,0],ymm1[4,6],ymm3[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1],ymm10[0],ymm2[2],ymm10[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,0],ymm5[4,5],ymm2[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm6[2,0],ymm3[0,0],ymm6[6,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[0,2],ymm2[2,0],ymm3[4,6],ymm2[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm9 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm13[0],xmm0[1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vmovapd 272(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm12[1],ymm14[0],ymm12[2],ymm14[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,0],ymm4[4,5],ymm2[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm9[2,0],ymm0[0,0],ymm9[6,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,2],ymm5[2,0],ymm0[4,6],ymm5[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm4 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm11[2,2,3,3]
-; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm8[0],xmm3[1],xmm8[2,3]
-; AVX1-ONLY-NEXT: vmovapd 656(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm9[1],ymm3[0],ymm9[2],ymm3[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm11[2,2,3,3]
+; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm10[0],xmm5[1],xmm10[2,3]
+; AVX1-ONLY-NEXT: vmovapd 656(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm5[1],ymm15[0],ymm5[2],ymm15[2]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm7[0,1],ymm15[2,0],ymm7[4,5],ymm15[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm13[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm13[2,0],ymm15[0,0],ymm13[6,4],ymm15[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm15[0,2],ymm12[2,0],ymm15[4,6],ymm12[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm2[0,1,2,3,4],ymm12[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[3,0],ymm0[1,0],ymm4[7,4],ymm0[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm2[2,0],ymm0[4,7],ymm2[6,4]
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm14[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = ymm4[3,1],mem[1,3],ymm4[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm10[1,1],ymm4[2,0],ymm10[5,5],ymm4[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm2[0,1,2,3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm5[1,0],ymm0[7,4],ymm5[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm5[0,3],ymm0[2,0],ymm5[4,7],ymm0[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm7[0,1],ymm14[2,0],ymm7[4,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm4[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[2,0],ymm14[0,0],ymm4[6,4],ymm14[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm14[0,2],ymm15[2,0],ymm14[4,6],ymm15[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm15[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm8[3,0],ymm1[1,0],ymm8[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],ymm2[2,0],ymm1[4,7],ymm2[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm15[1,1],ymm8[2,0],ymm15[5,5],ymm8[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,0],ymm3[1,0],ymm6[7,4],ymm3[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,3],ymm1[2,0],ymm3[4,7],ymm1[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = ymm6[3,1],mem[1,3],ymm6[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[1,1],ymm6[2,0],ymm8[5,5],ymm6[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm9[3,0],ymm0[1,0],ymm9[7,4],ymm0[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm3[2,0],ymm0[4,7],ymm3[6,4]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm13[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = ymm12[3,1],mem[1,3],ymm12[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[1,1],ymm6[2,0],ymm8[5,5],ymm6[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm14[1,0],ymm4[7,4],ymm14[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,3],ymm3[2,0],ymm14[4,7],ymm3[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm10[0,1],xmm11[2,3]
; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm5 = ymm5[3,1],mem[1,3],ymm5[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm10[1,1],ymm5[2,0],ymm10[5,5],ymm5[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[3,0],ymm1[1,0],ymm6[7,4],ymm1[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,3],ymm0[2,0],ymm1[4,7],ymm0[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = ymm2[3,1],mem[1,3],ymm2[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm6[1,1],ymm2[2,0],ymm6[5,5],ymm2[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[3,0],ymm15[1,0],ymm13[7,4],ymm15[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm15[0,3],ymm1[2,0],ymm15[4,7],ymm1[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm8[0,1],xmm11[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm9[3,1],ymm3[1,3],ymm9[7,5],ymm3[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[1,1],ymm3[2,0],ymm7[5,5],ymm3[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm12, 96(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%r9)
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm7[1,1],ymm5[2,0],ymm7[5,5],ymm5[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rax)
-; AVX1-ONLY-NEXT: addq $1064, %rsp # imm = 0x428
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rax)
+; AVX1-ONLY-NEXT: addq $1016, %rsp # imm = 0x3F8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -3608,9 +3603,9 @@
; AVX2-SLOW-NEXT: vmovaps %ymm1, 96(%r9)
; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm1, 32(%r9)
-; AVX2-SLOW-NEXT: vmovaps %ymm5, (%r9)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm1, 64(%r9)
+; AVX2-SLOW-NEXT: vmovaps %ymm5, (%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rax)
@@ -4015,9 +4010,9 @@
; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%r9)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%r9)
-; AVX2-FAST-NEXT: vmovaps %ymm5, (%r9)
; AVX2-FAST-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%r9)
+; AVX2-FAST-NEXT: vmovaps %ymm5, (%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-NEXT: vmovaps %ymm4, 96(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm6, 32(%rax)
@@ -4427,9 +4422,9 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 96(%r9)
; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 32(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%r9)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 96(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rax)
@@ -6050,7 +6045,7 @@
;
; AVX1-ONLY-LABEL: load_i32_stride6_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $2536, %rsp # imm = 0x9E8
+; AVX1-ONLY-NEXT: subq $2488, %rsp # imm = 0x9B8
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm6
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm2
@@ -6300,8 +6295,8 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm14[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[3,0],ymm13[1,0],ymm14[7,4],ymm13[5,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm13[1,0],ymm0[7,4],ymm13[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm13[2,3],ymm0[6,4],ymm13[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm1[1,0],xmm11[3,0]
@@ -6313,16 +6308,16 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm11[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm11[3,0],ymm10[1,0],ymm11[7,4],ymm10[5,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm10[1,0],ymm0[7,4],ymm10[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm10[2,3],ymm0[6,4],ymm10[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm1[1,0],xmm9[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm10[0,2],xmm9[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm1[3,1],ymm10[1,3],ymm1[7,5],ymm10[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm1[3,1],ymm14[1,3],ymm1[7,5],ymm14[5,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6346,22 +6341,22 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm1[1,0],xmm5[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm6[0,2],xmm5[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm1[3,1],ymm7[1,3],ymm1[7,5],ymm7[5,7]
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0],ymm4[1,0],ymm0[7,4],ymm4[5,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[3,0],ymm4[1,0],ymm6[7,4],ymm4[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm4[2,3],ymm0[6,4],ymm4[6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm1[1,0],xmm3[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm4[0,2],xmm3[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[3,1],ymm4[1,3],ymm1[7,5],ymm4[5,7]
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -6373,31 +6368,30 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm1[1,0],xmm12[3,0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm2[0,2],xmm12[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm1[3,1],mem[1,3],ymm1[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,1],ymm2[1,3],ymm1[7,5],ymm2[5,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,2,0,4,5,6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm12 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,3]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm12[2,0],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[2,0],ymm2[4,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,0],ymm1[2,0],ymm4[4,4],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6408,31 +6402,50 @@
; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,3]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm15[0,1,2,3],mem[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm11 = ymm15[0,1,2,3],mem[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm11[2,3,0,1]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[2,0],ymm2[4,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm11[2,0],ymm1[4,4],ymm11[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm14[2,1],mem[2,0],ymm14[6,5],mem[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,3]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = ymm13[0,1,2,3],mem[4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm10 = ymm13[0,1,2,3],mem[4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm10[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm10[2,0],ymm1[4,4],ymm10[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm4
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm14[0,1,2,3],mem[4,5],ymm14[6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm8[2,0],ymm1[4,4],ymm8[6,4]
@@ -6442,18 +6455,18 @@
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm11[2,1],mem[2,0],ymm11[6,5],mem[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[2,1],ymm15[2,0],ymm3[6,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,3]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = ymm10[0,1,2,3],mem[4,5],ymm10[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = ymm2[0,1,2,3],mem[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm7[2,3,0,1]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm6[2,0],ymm1[4,4],ymm6[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm7[2,0],ymm1[4,4],ymm7[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -6478,129 +6491,111 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm1
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm15[2,0],xmm1[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm13[2,0],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm9 = ymm7[0,1,2,3],mem[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[2,3,0,1]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[2,3,0,1]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm9[2,0],ymm1[4,4],ymm9[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0],ymm5[2,0],ymm1[4,4],ymm5[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $38, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[2,1],mem[2,0],ymm0[6,5],mem[6,4]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm4 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[2,1],ymm9[2,0],ymm6[6,5],ymm9[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm7[2,0],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = ymm4[0,1,2,3],mem[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm13[0,0],ymm2[2,0],ymm13[4,4],ymm2[6,4]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm6
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm4[2,0],xmm6[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm3[0,0],ymm1[2,0],ymm3[4,4],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm3[2,1],ymm5[2,0],ymm3[6,5],ymm5[6,4]
-; AVX1-ONLY-NEXT: vmovaps %ymm3, %ymm4
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm14[2,3,0,1]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm3
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm0[2,0],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm14[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vblendps $48, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm14[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,0],ymm14[2,0],ymm1[4,4],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3,4],ymm10[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = ymm10[0,1],mem[3,1],ymm10[4,5],mem[7,5]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm12 = xmm11[3,1],mem[3,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm11 = ymm11[3,1],mem[2,1],ymm11[7,5],mem[6,5]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm11[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2],ymm11[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4],ymm10[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = ymm10[0,1],mem[3,1],ymm10[4,5],mem[7,5]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm11[3,1],mem[3,3]
+; AVX1-ONLY-NEXT: vshufps $116, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[0,1],mem[3,1],ymm0[4,5],mem[7,5]
+; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm12[3,1],mem[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm12 = ymm12[3,1],mem[2,1],ymm12[7,5],mem[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm14[0,1,2],ymm12[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm11[3,1],ymm0[4,5],ymm11[7,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = xmm2[3,1],mem[3,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm12 = ymm14[3,1],mem[2,1],ymm14[7,5],mem[6,5]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4],ymm10[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[0,1],ymm10[3,1],ymm0[4,5],ymm10[7,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = xmm0[3,1],mem[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm10[0,1],ymm8[3,1],ymm10[4,5],ymm8[7,5]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = xmm10[3,1],mem[3,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm11 = ymm11[3,1],mem[2,1],ymm11[7,5],mem[6,5]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm11[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm11[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3,4],ymm8[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[0,1],ymm6[3,1],ymm8[4,5],ymm6[7,5]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm8[3,1],mem[3,3]
+; AVX1-ONLY-NEXT: vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm12 = ymm10[3,1],mem[2,1],ymm10[7,5],mem[6,5]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm12[2,3,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2],ymm12[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm8[3,1],ymm0[4,5],ymm8[7,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = xmm2[3,1],mem[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $103, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm10 = ymm10[3,1],mem[2,1],ymm10[7,5],mem[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm10[2,3,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm10[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm6[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1],ymm14[3,1],ymm1[4,5],ymm14[7,5]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,1],xmm3[3,3]
-; AVX1-ONLY-NEXT: vmovaps %ymm5, %ymm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[3,1],ymm5[2,1],ymm4[7,5],ymm5[6,5]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[0,1],ymm2[3,1],ymm13[4,5],ymm2[7,5]
-; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm7[3,1],mem[3,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm7[3,1],ymm0[4,5],ymm7[7,5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm2[3,1],mem[3,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm2[3,1],ymm15[2,1],ymm2[7,5],ymm15[6,5]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[0,1],ymm1[3,1],ymm3[4,5],ymm1[7,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm4[3,1],xmm6[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[3,1],ymm7[2,1],ymm3[7,5],ymm7[6,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[3,1],ymm9[2,1],ymm3[7,5],ymm9[6,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm9[3,1],ymm0[4,5],ymm9[7,5]
-; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm15[3,1],mem[3,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1],ymm5[3,1],ymm0[4,5],ymm5[7,5]
+; AVX1-ONLY-NEXT: vshufps $247, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm13[3,1],mem[3,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[3,1],ymm5[2,1],ymm4[7,5],ymm5[6,5]
@@ -6658,7 +6653,7 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm6, (%rsp) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm6[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm6[0,0],ymm1[6,4],ymm6[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2],ymm1[2,0],ymm6[4,6],ymm1[6,4]
@@ -6691,14 +6686,13 @@
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,2,3,3]
; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vmovapd 656(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[1],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,1],ymm1[2,0],ymm14[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm12[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[2,0],ymm4[0,0],ymm12[6,4],ymm4[4,4]
@@ -6716,7 +6710,7 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vmovapd 848(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm7[0],ymm1[2],ymm7[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm9[0],ymm1[2],ymm9[2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,0],ymm3[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm11[2,3,0,1]
@@ -6756,7 +6750,7 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vmovapd 1232(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm14[0],ymm1[2],ymm14[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm15[0],ymm1[2],ymm15[2]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,0],ymm7[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
@@ -6800,98 +6794,98 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm13[1,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm13[0,1,2,3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm8[3,0],ymm6[1,0],ymm8[7,4],ymm6[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[0,3],ymm13[2,0],ymm6[4,7],ymm13[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm8[0,1],mem[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[3,0],ymm6[1,0],ymm8[7,4],ymm6[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[0,3],ymm8[2,0],ymm6[4,7],ymm8[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = ymm13[3,1],mem[1,3],ymm13[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm15[1,1],ymm13[2,0],ymm15[5,5],ymm13[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm6[5,6,7]
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm15 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[3,0],ymm5[1,0],ymm8[7,4],ymm5[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[0,3],ymm8[2,0],ymm5[4,7],ymm8[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[1,1],ymm15[2,0],ymm8[5,5],ymm15[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm13[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3,4],ymm6[5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm8[3,0],ymm5[1,0],ymm8[7,4],ymm5[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[0,3],ymm13[2,0],ymm5[4,7],ymm13[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm13 = ymm13[3,1],mem[1,3],ymm13[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm15[1,1],ymm13[2,0],ymm15[5,5],ymm13[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3,4],ymm5[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm12[3,0],ymm4[1,0],ymm12[7,4],ymm4[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[0,3],ymm8[2,0],ymm4[4,7],ymm8[6,4]
-; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm12 = ymm12[3,1],mem[1,3],ymm12[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm13[1,1],ymm12[2,0],ymm13[5,5],ymm12[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm12[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1,2,3,4],ymm4[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm11[3,0],ymm3[1,0],ymm11[7,4],ymm3[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,3],ymm8[2,0],ymm3[4,7],ymm8[6,4]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm15 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[1,1],ymm15[2,0],ymm8[5,5],ymm15[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm13[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm13[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm12[3,0],ymm4[1,0],ymm12[7,4],ymm4[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[0,3],ymm12[2,0],ymm4[4,7],ymm12[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm11 = ymm11[3,1],mem[1,3],ymm11[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm12[1,1],ymm11[2,0],ymm12[5,5],ymm11[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm11[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3,4],ymm3[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm10[3,0],ymm2[1,0],ymm10[7,4],ymm2[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,3],ymm8[2,0],ymm2[4,7],ymm8[6,4]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm8[1,1],ymm13[2,0],ymm8[5,5],ymm13[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm12[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4],ymm4[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[3,0],ymm3[1,0],ymm11[7,4],ymm3[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,3],ymm11[2,0],ymm3[4,7],ymm11[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = ymm10[3,1],mem[1,3],ymm10[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm11[1,1],ymm10[2,0],ymm11[5,5],ymm10[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm10[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm8[0,1,2,3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm9[3,0],ymm1[1,0],ymm9[7,4],ymm1[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],ymm8[2,0],ymm1[4,7],ymm8[6,4]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm12 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm12 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm8[1,1],ymm12[2,0],ymm8[5,5],ymm12[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm11[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm12[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3,4],ymm3[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm10[3,0],ymm2[1,0],ymm10[7,4],ymm2[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,3],ymm10[2,0],ymm2[4,7],ymm10[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm9 = ymm9[3,1],mem[1,3],ymm9[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm10[1,1],ymm9[2,0],ymm10[5,5],ymm9[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm11 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm8[1,1],ymm11[2,0],ymm8[5,5],ymm11[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm10[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3,4],ymm2[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[3,0],ymm1[1,0],ymm9[7,4],ymm1[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],ymm9[2,0],ymm1[4,7],ymm9[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm9 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm10 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm8[1,1],ymm10[2,0],ymm8[5,5],ymm10[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm9[1,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm10[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[3,0],ymm0[1,0],ymm7[7,4],ymm0[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,3],ymm7[2,0],ymm0[4,7],ymm7[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = xmm7[0,1],mem[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm9[1,1],ymm8[2,0],ymm9[5,5],ymm8[6,4]
+; AVX1-ONLY-NEXT: vshufps $215, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm9 = ymm8[3,1],mem[1,3],ymm8[7,5],mem[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm8[1,1],ymm9[2,0],ymm8[5,5],ymm9[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm7[1,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm7, 192(%rsi)
@@ -6926,14 +6920,14 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm7, 32(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 192(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm7, 128(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm7, 64(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm7, (%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 192(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm7, 224(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm7, 160(%rcx)
@@ -6982,7 +6976,7 @@
; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm14, (%rax)
-; AVX1-ONLY-NEXT: addq $2536, %rsp # imm = 0x9E8
+; AVX1-ONLY-NEXT: addq $2488, %rsp # imm = 0x9B8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -6995,7 +6989,7 @@
; AVX2-SLOW-NEXT: vmovaps 608(%rdi), %ymm5
; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %ymm6
-; AVX2-SLOW-NEXT: vmovups %ymm6, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 352(%rdi), %ymm7
; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %ymm8
@@ -7016,8 +7010,8 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm11 = [4,2,4,2,4,2,4,2]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm11, %ymm1
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm9 = [4,2,4,2,4,2,4,2]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7037,7 +7031,7 @@
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm11, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1056(%rdi), %ymm1
@@ -7057,12 +7051,12 @@
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 1088(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1120(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm11, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1440(%rdi), %ymm1
@@ -7087,7 +7081,7 @@
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm11, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %ymm1
@@ -7108,8 +7102,8 @@
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm11, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 480(%rdi), %ymm1
@@ -7117,21 +7111,21 @@
; AVX2-SLOW-NEXT: vmovaps 448(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovaps 416(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm4, %ymm0
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 544(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm11, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 864(%rdi), %ymm1
@@ -7153,7 +7147,7 @@
; AVX2-SLOW-NEXT: vmovaps 928(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm11, %ymm1
+; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 1184(%rdi), %ymm0
@@ -7175,11 +7169,11 @@
; AVX2-SLOW-NEXT: vmovaps 1312(%rdi), %ymm14
; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm11, %ymm11
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm9, %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm11 = <1,7,5,u>
-; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm11, %ymm0
+; AVX2-SLOW-NEXT: vmovaps {{.*#+}} xmm9 = <1,7,5,u>
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm9, %ymm0
; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7]
@@ -7187,53 +7181,53 @@
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm11, %ymm12
+; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm9, %ymm12
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm10
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm11, %ymm7
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm0, %ymm11
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm9, %ymm7
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm10[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm0, %ymm8
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm11, %ymm3
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm9, %ymm3
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm11, %ymm2
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm9, %ymm2
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %xmm4
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; AVX2-SLOW-NEXT: vmovaps 192(%rdi), %xmm6
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm6[2,3,2,3]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
@@ -7244,6 +7238,48 @@
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm4[2,3],ymm13[4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 960(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
; AVX2-SLOW-NEXT: vpermilps $224, (%rsp), %ymm2 # 32-byte Folded Reload
@@ -7252,7 +7288,48 @@
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps 1344(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm14[2,3],ymm11[4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 1152(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2],ymm1[3],ymm7[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
@@ -7267,181 +7344,132 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2],ymm1[3],ymm8[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 960(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2],ymm1[3],ymm5[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 1344(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm13[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm14[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm0
-; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 384(%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm6[1,2,3],ymm0[4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1,2,3],ymm0[4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6],ymm6[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 768(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,3,2,3]
-; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,0,0,0,4,4,4,4]
; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3],ymm9[4,5,6],ymm7[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 1152(%rdi), %xmm12
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm12[2,3,2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm0[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm9[2,3],ymm11[4,5],ymm9[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
+; AVX2-SLOW-NEXT: vmovaps 768(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm8 = xmm0[2,3,2,3]
+; AVX2-SLOW-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = mem[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[3,3,3,3]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm8[0,0,0,0,4,4,4,4]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm5[0,0,2,3,4,4,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2],ymm7[3],ymm10[4,5,6],ymm7[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,3,3,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm5 = mem[3,3,3,3]
; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm6[1,2,3],ymm4[4],ymm6[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4],ymm6[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm13[1],ymm6[2,3,4],ymm13[5],ymm6[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm3[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps $244, (%rsp), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6],ymm6[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6],ymm6[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm4 = mem[3,3,3,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vpermilps $244, (%rsp), %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm3 = mem[3,3,3,3]
; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3],ymm6[4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm11[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm13[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm12[1],ymm3[2,3,4],ymm12[5],ymm3[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
@@ -7453,72 +7481,39 @@
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; AVX2-SLOW-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm15[1],ymm2[2,3,4],ymm15[5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm13[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm12[3,3,3,3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm0[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm11[1],ymm2[2,3,4],ymm11[5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm5[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm0 = mem[3,3,3,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm10[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1],ymm1[2,3,4],ymm13[5],ymm1[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm1[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm14[1],ymm2[2,3,4],ymm14[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,1,0,1,4,5,4,5]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,3,3,3,7,7,7,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,1,0,1,4,5,4,5]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,1,0,1,4,5,4,5]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -7530,8 +7525,8 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -7544,147 +7539,148 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovaps 272(%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm1 = [4,2,4,2]
-; AVX2-SLOW-NEXT: # xmm1 = mem[0,0]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm8 = [4,2,4,2]
+; AVX2-SLOW-NEXT: # xmm8 = mem[0,0]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm8, %ymm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 464(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1],ymm7[2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 464(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 656(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm9[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm14[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 656(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vmovaps 848(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm10[0,1],mem[2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1040(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm14 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm11[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm9[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 848(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm12[0,1],ymm11[2,3],ymm12[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1232(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm10 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1424(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm8, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 1040(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm10 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 1232(%rdi), %xmm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm12[2,3],ymm10[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm11 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm11 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 1424(%rdi), %xmm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm15[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 80(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,1,0,2,4,5,4,6]
-; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm8, %ymm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
@@ -7697,57 +7693,58 @@
; AVX2-SLOW-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm1 = [5,3,5,3]
; AVX2-SLOW-NEXT: # xmm1 = mem[0,0]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm0[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm0, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = ymm15[0,1,2],mem[3],ymm15[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm0, %ymm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, (%rsp), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm0, %ymm13
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm10 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3],ymm10[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm1, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm0, %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm7[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm12[3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm1, %ymm8
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm0, %ymm6
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
@@ -7790,14 +7787,14 @@
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm0, 192(%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 128(%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 64(%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, (%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 192(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 224(%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 160(%rcx)
@@ -7839,9 +7836,9 @@
; AVX2-SLOW-NEXT: vmovaps %ymm0, (%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT: vmovaps %ymm6, 224(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm9, 192(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm13, 160(%rax)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm10, 192(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm14, 160(%rax)
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 128(%rax)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 96(%rax)
@@ -7856,7 +7853,7 @@
;
; AVX2-FAST-LABEL: load_i32_stride6_vf64:
; AVX2-FAST: # %bb.0:
-; AVX2-FAST-NEXT: subq $2472, %rsp # imm = 0x9A8
+; AVX2-FAST-NEXT: subq $2504, %rsp # imm = 0x9C8
; AVX2-FAST-NEXT: vmovaps 672(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovaps 640(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8038,11 +8035,11 @@
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,2,2,2,4,6,6,6]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1280(%rdi), %ymm14
-; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1312(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovaps 1280(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1312(%rdi), %ymm14
+; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm1, %ymm10, %ymm10
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8100,38 +8097,18 @@
; AVX2-FAST-NEXT: vpermps %ymm1, %ymm0, %ymm0
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 192(%rdi), %xmm4
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; AVX2-FAST-NEXT: vmovaps 192(%rdi), %xmm6
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm6[2,3,2,3]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm0 = [2,0,6,4,2,0,6,7]
; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm5[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 576(%rdi), %xmm1
-; AVX2-FAST-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm2 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
@@ -8140,215 +8117,237 @@
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 960(%rdi), %xmm2
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT: vmovaps 576(%rdi), %xmm1
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm1[2,3,2,3]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm3[1,2,3],ymm1[4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5],ymm3[6,7]
; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm12[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2],ymm3[3],ymm6[4,5,6],ymm3[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 960(%rdi), %xmm2
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm2[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm4, %ymm0, %ymm4
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm13[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 1344(%rdi), %xmm3
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm3[2,3,2,3]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2,3],ymm1[4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm4 = xmm3[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm11[0,1],ymm15[2,3],ymm11[4,5],ymm15[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm6
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm6[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm14[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm11[0,1],ymm14[2,3],ymm11[4,5],ymm14[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm5, %ymm0, %ymm5
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm3[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2],ymm5[3],ymm7[4,5,6],ymm5[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1152(%rdi), %xmm4
+; AVX2-FAST-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm5 = xmm4[2,3,2,3]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6],ymm6[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm1
-; AVX2-FAST-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2,3],ymm1[4],ymm6[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1],ymm6[2,3],mem[4,5],ymm6[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm6
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6],ymm6[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 384(%rdi), %xmm1
-; AVX2-FAST-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm7[1,2,3],ymm1[4],ymm7[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,1],ymm6[2,3],mem[4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm7[1,2,3],ymm5[4],ymm7[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
; AVX2-FAST-NEXT: vpermps %ymm7, %ymm0, %ymm7
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3],ymm9[4,5,6],ymm7[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm12[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2],ymm7[3],ymm8[4,5,6],ymm7[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 768(%rdi), %xmm1
-; AVX2-FAST-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm7 = xmm1[2,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps (%rdi), %xmm4
+; AVX2-FAST-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm5 = xmm4[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm8[1,2,3],ymm5[4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm0, %ymm8
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0],ymm9[1,2,3],ymm7[4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm9 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 384(%rdi), %xmm4
+; AVX2-FAST-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm8 = xmm4[2,3,2,3]
+; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
; AVX2-FAST-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm9 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm10 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,1],ymm6[2,3],mem[4,5],ymm6[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm0, %ymm7
-; AVX2-FAST-NEXT: vmovaps 1152(%rdi), %xmm0
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1],ymm4[2,3],mem[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm0, %ymm8
+; AVX2-FAST-NEXT: vmovaps 768(%rdi), %xmm0
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm9 = xmm0[2,3,2,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm10[1,2,3],ymm9[4],ymm10[5,6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm9 = ymm13[0,0,0,0,4,4,4,4]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm8[0,0,2,3,4,4,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm9 = ymm5[0,0,0,0,4,4,4,4]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm10 = ymm4[0,0,2,3,4,4,6,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,3,3,3]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm7[1,2,3],ymm4[4],ymm7[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm6 = xmm6[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1,0,1,4,5,4,5]
; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm9 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3],ymm9[4,5,6],ymm7[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm4 = mem[3,3,3,3]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $255, (%rsp), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2],ymm5[3],ymm7[4,5,6],ymm5[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1,2,3],ymm2[4],ymm4[5,6,7]
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm12[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm3[3,3,3,3]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm15[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm11[1],ymm3[2,3,4],ymm11[5],ymm3[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3],ymm4[4,5,6],ymm3[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm1[0,1,3,3,4,5,7,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm6[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2,3],ymm1[4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, (%rsp), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6],ymm6[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm2[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[3,3,3,3,7,7,7,7]
; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm13[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vmovaps %ymm8, %ymm13
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm8[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2],ymm2[3],ymm6[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm1 = mem[3,3,3,3]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm11[1],ymm2[2,3,4],ymm11[5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm3 = ymm15[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm1 = mem[3,3,3,3]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm13[1],ymm2[2,3,4],ymm13[5],ymm2[6,7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm12[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vmovaps %ymm7, %ymm12
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[3,3,3,3,7,7,7,7]
@@ -8357,10 +8356,8 @@
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,1,0,1,4,5,4,5]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -8372,13 +8369,13 @@
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,1,0,1,4,5,4,5]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,0,1,4,5,4,5]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
@@ -8392,8 +8389,8 @@
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -8410,148 +8407,150 @@
; AVX2-FAST-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm0[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 272(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm5 = [4,2,4,2]
-; AVX2-FAST-NEXT: # xmm5 = mem[0,0]
-; AVX2-FAST-NEXT: vpermps %ymm1, %ymm5, %ymm1
+; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm8 = [4,2,4,2]
+; AVX2-FAST-NEXT: # xmm8 = mem[0,0]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm8, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm15 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
-; AVX2-FAST-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,2,0,6,0,2,0,6]
+; AVX2-FAST-NEXT: # ymm4 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 464(%rdi), %xmm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 464(%rdi), %xmm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm5, %ymm2
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm8, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm8[0,1],ymm7[2,3],ymm8[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm7[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = mem[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, (%rsp), %ymm1, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, (%rsp), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 656(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm5, %ymm2
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm2[0,1],mem[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm9[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm12[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm14[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 848(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm5, %ymm2
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1],ymm11[2,3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm4, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 1040(%rdi), %xmm1
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm5, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm11 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1040(%rdi), %xmm1
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm14 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm14, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm11[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm11 = ymm13[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 1232(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm2, %ymm5, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = mem[0,1],ymm13[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm10, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm8 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1424(%rdi), %xmm12
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm12[2,3],ymm8[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm13, %ymm5, %ymm2
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm7 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm10, %ymm4, %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm12 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1424(%rdi), %xmm13
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm13[2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-NEXT: vpermps %ymm0, %ymm8, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm4, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = ymm15[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 80(%rdi), %xmm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm2[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm5, %ymm3
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm8, %ymm3
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpermps %ymm3, %ymm4, %ymm0
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
@@ -8560,64 +8559,63 @@
; AVX2-FAST-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vmovddup {{.*#+}} xmm1 = [5,3,5,3]
; AVX2-FAST-NEXT: # xmm1 = mem[0,0]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-FAST-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm0, %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm15[0,1],ymm6[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, (%rsp), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm15[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vpermilps $85, (%rsp), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm15[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = mem[1,1,1,1,5,5,5,5]
; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = ymm15[0,1,2],mem[3],ymm15[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm14, %ymm0, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm6[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm11[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = ymm6[0,1,2],mem[3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm11[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm10, %ymm0, %ymm10
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm6[0,1,2,3,4],ymm10[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm12[3],ymm6[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm13, %ymm1, %ymm8
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1],ymm6[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm7, %ymm0, %ymm7
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm6[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm7[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm7 = ymm12[1,1,1,1,5,5,5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm13[3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpermps %ymm6, %ymm0, %ymm6
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,1,1,1,5,5,5,5]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vpermps %ymm4, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm1, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpermps %ymm3, %ymm0, %ymm0
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm0[5,6,7]
@@ -8654,14 +8652,14 @@
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm0, 192(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 128(%rcx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 64(%rcx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, (%rcx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm0, 192(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 224(%rcx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 160(%rcx)
@@ -8702,19 +8700,18 @@
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, (%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovaps %ymm7, 224(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm6, 224(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm10, 192(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm14, 160(%rax)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 128(%rax)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 96(%rax)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm4, 64(%rax)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 32(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm1, (%rax)
-; AVX2-FAST-NEXT: addq $2472, %rsp # imm = 0x9A8
+; AVX2-FAST-NEXT: addq $2504, %rsp # imm = 0x9C8
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
;
@@ -8727,7 +8724,7 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps 608(%rdi), %ymm5
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 352(%rdi), %ymm7
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %ymm8
@@ -8748,8 +8745,8 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm11 = [4,2,4,2,4,2,4,2]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm11, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm9 = [4,2,4,2,4,2,4,2]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8769,7 +8766,7 @@
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm11, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1056(%rdi), %ymm1
@@ -8789,12 +8786,12 @@
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2,2,2,4,6,6,6]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 1088(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1120(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm11, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1440(%rdi), %ymm1
@@ -8819,7 +8816,7 @@
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm11, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %ymm1
@@ -8840,8 +8837,8 @@
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 160(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm11, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 480(%rdi), %ymm1
@@ -8849,21 +8846,21 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps 448(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[0,1],ymm1[0,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 416(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,2,2,2,4,6,6,6]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2,2,2,4,6,6,6]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 544(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm11, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 864(%rdi), %ymm1
@@ -8885,7 +8882,7 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps 928(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm11, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 1184(%rdi), %ymm0
@@ -8907,11 +8904,11 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps 1312(%rdi), %ymm14
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm11, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm9, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm9[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm11 = <1,7,5,u>
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm11, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} xmm9 = <1,7,5,u>
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm9, %ymm0
; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,3,2,3,5,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2],ymm14[3,4,5,6,7]
@@ -8919,53 +8916,53 @@
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vpermilps $237, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[1,3,2,3,5,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm15[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm11, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm9, %ymm12
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm13[1,3,2,3,5,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2],ymm13[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm11, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm9 = ymm9[1,3,2,3,5,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm0, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm9, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm10[1,3,2,3,5,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm10[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm0, %ymm8
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm11, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm9, %ymm3
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,3,2,3,5,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm6[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm11, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm9, %ymm2
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm4[1,3,2,3,5,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %xmm4
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 192(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm6[2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
@@ -8976,6 +8973,48 @@
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm5[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm4[2,3],ymm13[4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 960(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,0,0,0,4,4,4,4]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, (%rsp), %ymm2 # 32-byte Folded Reload
@@ -8984,7 +9023,48 @@
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 576(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1344(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm14[2,3],ymm11[4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm15[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1152(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2],ymm1[3],ymm7[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
@@ -8999,181 +9079,132 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2],ymm1[3],ymm8[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 960(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm2[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2],ymm1[3],ymm5[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1344(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $204, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm15[0,1],mem[2,3],ymm15[4,5],mem[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm13[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm14[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1,2,3],ymm0[4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 384(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm6[1,2,3],ymm0[4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1],ymm5[2,3],mem[4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1,2,3],ymm0[4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 768(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,0,0,0,4,4,4,4]
; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1,2],ymm7[3],ymm9[4,5,6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4],ymm7[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1152(%rdi), %xmm12
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm12[2,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm0[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm9[2,3],ymm11[4,5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 768(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm8 = xmm0[2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $224, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1,2,3],ymm8[4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2],ymm9[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm7[0,0,2,3,4,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3],ymm10[4,5,6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm8[1,2,3],ymm6[4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm8[0,0,0,0,4,4,4,4]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm5[0,0,2,3,4,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2],ymm7[3],ymm10[4,5,6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm7[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm5[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2],ymm8[3],ymm9[4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm8[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm5 = mem[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm6[1,2,3],ymm4[4],ymm6[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1,2,3],ymm5[4],ymm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm4[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm13[1],ymm6[2,3,4],ymm13[5],ymm6[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm3[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, (%rsp), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2],ymm6[3],ymm7[4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2],ymm6[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3],ymm8[4,5,6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm6[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm4 = mem[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm12[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1,2,3],ymm4[4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm5[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm3[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $244, (%rsp), %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2],ymm5[3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm3 = mem[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3],ymm3[4],ymm4[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3,4],ymm11[5],ymm4[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1,2],ymm4[3],ymm6[4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3],ymm5[4,5,6],ymm4[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm11[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm13[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm12[1],ymm3[2,3,4],ymm12[5],ymm3[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm3[3,4,5,6,7]
@@ -9185,72 +9216,39 @@
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $244, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm15[1],ymm2[2,3,4],ymm15[5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm13[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm12[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm0[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm9[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm11[1],ymm2[2,3,4],ymm11[5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm8[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm5[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm0 = mem[3,3,3,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm10[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm13[1],ymm1[2,3,4],ymm13[5],ymm1[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm1[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm14[1],ymm2[2,3,4],ymm14[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,1,0,1,4,5,4,5]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm11[0,1,3,3,4,5,7,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3],ymm3[4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm0 = mem[3,3,3,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,3,3,3,7,7,7,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm5[1],ymm1[2,3,4],ymm5[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm5[3,3,3,3,7,7,7,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,1,0,1,4,5,4,5]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,1,3,3,4,5,7,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,1,0,1,4,5,4,5]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm6[0,1,3,3,4,5,7,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -9262,8 +9260,8 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[3,3,3,3,7,7,7,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4],ymm15[5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,0,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -9276,147 +9274,148 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm0[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm0[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps 272(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm1 = [4,2,4,2]
-; AVX2-FAST-PERLANE-NEXT: # xmm1 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm8 = [4,2,4,2]
+; AVX2-FAST-PERLANE-NEXT: # xmm8 = mem[0,0]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm8, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,2,0,6,0,2,0,6]
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm3[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm4[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 464(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1],ymm7[2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 464(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm6[2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 656(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm9[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm14[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 656(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vmovaps 848(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm10[0,1],mem[2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm3[0,1],mem[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1040(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm11[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm9[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm13[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 848(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm12[0,1],ymm11[2,3],ymm12[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1232(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3],ymm11[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm13 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1424(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm12[2,3],ymm9[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm8, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1040(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm10 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1232(%rdi), %xmm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1],ymm12[2,3],ymm10[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm11 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm2[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1424(%rdi), %xmm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm15[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm2[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm1[0,1,2,3],mem[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 80(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,1,0,2,4,5,4,6]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm2[2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,1,0,2,4,5,4,6]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm8, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
@@ -9429,57 +9428,58 @@
; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm1 = [5,3,5,3]
; AVX2-FAST-PERLANE-NEXT: # xmm1 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm0 = [0,3,1,7,0,3,1,7]
; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm0, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm15[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0,1,2],mem[3],ymm15[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm0, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, (%rsp), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0,1,2],mem[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm15 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm0, %ymm13
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm10 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3],ymm10[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm9, %ymm0, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3,4],ymm9[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm1, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3,4],ymm14[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0,1,2],mem[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm0, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm7[0,1,2,3,4],ymm10[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm12[3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm1, %ymm8
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4],ymm6[5,6,7]
@@ -9522,14 +9522,14 @@
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%rdx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 192(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 128(%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 64(%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 192(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 224(%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 160(%rcx)
@@ -9571,9 +9571,9 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 224(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 192(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, 160(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, 192(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, 160(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 128(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 96(%rax)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
index 97f4999..a8b737f 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-7.ll
@@ -127,9 +127,8 @@
; AVX512F-SLOW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512F-SLOW-NEXT: vmovdqa 32(%rdi), %xmm1
; AVX512F-SLOW-NEXT: vpinsrd $1, 28(%rdi), %xmm0, %xmm2
-; AVX512F-SLOW-NEXT: vmovd %xmm1, %r11d
-; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
-; AVX512F-SLOW-NEXT: vpinsrd $1, %r11d, %xmm3, %xmm3
+; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX512F-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[1,0,2,3]
; AVX512F-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm4
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0],xmm1[1],xmm4[2,3]
; AVX512F-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3]
@@ -196,9 +195,8 @@
; AVX512BW-SLOW-NEXT: vmovdqa (%rdi), %xmm0
; AVX512BW-SLOW-NEXT: vmovdqa 32(%rdi), %xmm1
; AVX512BW-SLOW-NEXT: vpinsrd $1, 28(%rdi), %xmm0, %xmm2
-; AVX512BW-SLOW-NEXT: vmovd %xmm1, %r11d
-; AVX512BW-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
-; AVX512BW-SLOW-NEXT: vpinsrd $1, %r11d, %xmm3, %xmm3
+; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm1[0],xmm0[1],xmm1[2,3]
+; AVX512BW-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[1,0,2,3]
; AVX512BW-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm4
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0],xmm1[1],xmm4[2,3]
; AVX512BW-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3]
@@ -2745,83 +2743,51 @@
; SSE-LABEL: load_i32_stride7_vf32:
; SSE: # %bb.0:
; SSE-NEXT: subq $1176, %rsp # imm = 0x498
-; SSE-NEXT: movdqa 80(%rdi), %xmm6
-; SSE-NEXT: movdqa (%rdi), %xmm11
+; SSE-NEXT: movdqa 416(%rdi), %xmm6
+; SSE-NEXT: movdqa 384(%rdi), %xmm8
+; SSE-NEXT: movdqa 336(%rdi), %xmm11
; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 16(%rdi), %xmm9
+; SSE-NEXT: movdqa 352(%rdi), %xmm9
; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 48(%rdi), %xmm10
-; SSE-NEXT: movdqa 640(%rdi), %xmm2
-; SSE-NEXT: movdqa 608(%rdi), %xmm3
-; SSE-NEXT: movdqa 560(%rdi), %xmm8
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 576(%rdi), %xmm4
+; SSE-NEXT: movdqa 752(%rdi), %xmm2
+; SSE-NEXT: movdqa 720(%rdi), %xmm14
+; SSE-NEXT: movdqa 672(%rdi), %xmm10
+; SSE-NEXT: movdqa %xmm10, (%rsp) # 16-byte Spill
+; SSE-NEXT: movdqa 688(%rdi), %xmm5
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 304(%rdi), %xmm7
+; SSE-NEXT: movdqa 272(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm5
-; SSE-NEXT: movdqa 160(%rdi), %xmm7
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rdi), %xmm12
-; SSE-NEXT: movdqa 128(%rdi), %xmm0
+; SSE-NEXT: movdqa 224(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 240(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT: movdqa %xmm12, %xmm1
-; SSE-NEXT: movdqa %xmm12, %xmm14
-; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT: movdqa %xmm5, %xmm13
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT: movdqa %xmm7, %xmm13
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
-; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
+; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
-; SSE-NEXT: movdqa %xmm3, %xmm8
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT: movdqa %xmm2, %xmm7
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm2, %xmm4
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
-; SSE-NEXT: movdqa %xmm10, %xmm15
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
+; SSE-NEXT: movdqa %xmm8, %xmm11
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 448(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 464(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 528(%rdi), %xmm2
-; SSE-NEXT: movdqa 496(%rdi), %xmm4
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
-; SSE-NEXT: movdqa %xmm4, %xmm12
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT: movdqa %xmm2, %xmm11
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 336(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 352(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 416(%rdi), %xmm2
-; SSE-NEXT: movdqa 384(%rdi), %xmm5
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: movdqa %xmm6, %xmm15
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 784(%rdi), %xmm1
@@ -2830,390 +2796,447 @@
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movdqa 864(%rdi), %xmm2
-; SSE-NEXT: movdqa 832(%rdi), %xmm4
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
-; SSE-NEXT: movdqa %xmm4, %xmm10
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT: movdqa %xmm2, %xmm9
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 864(%rdi), %xmm9
+; SSE-NEXT: movdqa 832(%rdi), %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT: movdqa %xmm5, %xmm10
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 224(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 240(%rdi), %xmm0
+; SSE-NEXT: movdqa (%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 16(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT: movdqa 304(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 272(%rdi), %xmm0
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: movdqa 80(%rdi), %xmm7
+; SSE-NEXT: movdqa 48(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 672(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 688(%rdi), %xmm0
+; SSE-NEXT: movdqa 448(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 464(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT: movdqa 752(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 720(%rdi), %xmm0
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: movdqa 528(%rdi), %xmm5
+; SSE-NEXT: movdqa 496(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 112(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 128(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: movdqa 192(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 160(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 560(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 576(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: movdqa 640(%rdi), %xmm12
+; SSE-NEXT: movdqa 608(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,2,2,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
-; SSE-NEXT: movdqa 144(%rdi), %xmm14
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2]
-; SSE-NEXT: movdqa %xmm8, %xmm2
-; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: movdqa 592(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,2,2]
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm15, %xmm2
-; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
-; SSE-NEXT: movdqa 32(%rdi), %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm1, %xmm7
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,2,2]
-; SSE-NEXT: movdqa %xmm12, %xmm2
-; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[1,1,1,1]
-; SSE-NEXT: movdqa 480(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT: movdqa 368(%rdi), %xmm4
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
-; SSE-NEXT: movdqa %xmm10, %xmm5
-; SSE-NEXT: punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
-; SSE-NEXT: movdqa 816(%rdi), %xmm2
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: movdqa 256(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
+; SSE-NEXT: movdqa %xmm4, %xmm13
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm0[2],xmm14[3],xmm0[3]
+; SSE-NEXT: movdqa (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
+; SSE-NEXT: movdqa 704(%rdi), %xmm6
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT: movapd %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,2,2]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm0[2],xmm11[3],xmm0[3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: movdqa 368(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
+; SSE-NEXT: movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm10, %xmm1
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT: movdqa 816(%rdi), %xmm4
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm2[1,1,1,1]
-; SSE-NEXT: movdqa 256(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm8[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm15[2,2,2,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm8[2],xmm5[3],xmm8[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm12[1,1,1,1]
-; SSE-NEXT: movdqa 704(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm8[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = mem[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm14[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
-; SSE-NEXT: movdqa 176(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = mem[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm8[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm13[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm7[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
-; SSE-NEXT: movdqa 64(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm8[0],xmm5[1]
-; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm1[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm4[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
-; SSE-NEXT: movdqa 400(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm3[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm2[2,3,2,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: movdqa 32(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; SSE-NEXT: movdqa 480(%rdi), %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm14[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
-; SSE-NEXT: movdqa 288(%rdi), %xmm13
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm10[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm13[0,0,1,1]
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = mem[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm7[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm0[2],xmm7[3],xmm0[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm15[1,1,1,1]
+; SSE-NEXT: movdqa 144(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm2[0],xmm7[1]
+; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm12[2,2,2,2]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm2[2],xmm8[3],xmm2[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm2[1,1,1,1]
+; SSE-NEXT: movdqa 592(%rdi), %xmm0
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm7[0],xmm8[1]
+; SSE-NEXT: movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm3[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm6[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: movdqa 736(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm13[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm10[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm4[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: movdqa 848(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm9[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm11[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm1[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: movdqa 512(%rdi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm5[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm2[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm0[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
; SSE-NEXT: movdqa 624(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = mem[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm12[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm4[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm11[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: movdqa 288(%rdi), %xmm0
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT: # xmm8 = mem[2,3,2,3]
-; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = mem[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
-; SSE-NEXT: movdqa 512(%rdi), %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
+; SSE-NEXT: movdqa %xmm0, %xmm12
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = mem[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm11[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm5[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm10[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: movdqa 400(%rdi), %xmm0
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
+; SSE-NEXT: movdqa %xmm0, %xmm13
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm6[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
-; SSE-NEXT: movdqa 848(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = mem[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm12[2,3,2,3]
-; SSE-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm0[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
-; SSE-NEXT: movdqa 736(%rdi), %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm15[2,3,2,3]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,0,1,1]
-; SSE-NEXT: movdqa %xmm2, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm6[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm9[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: movdqa 64(%rdi), %xmm2
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,1,1]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm15[2,3,2,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm0[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT: movdqa 176(%rdi), %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm14[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1]
+; SSE-NEXT: movdqa %xmm1, %xmm14
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 96(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm1[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = mem[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm12[2],xmm8[3],xmm12[3]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm9[2],xmm6[3],xmm9[3]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm6[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 208(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm1[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm5, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT: # xmm8 = mem[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm10[2],xmm8[3],xmm10[3]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 320(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm1[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm8[0],xmm13[1],xmm8[1]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm14[2],xmm1[3],xmm14[3]
-; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm1[0],xmm13[1]
-; SSE-NEXT: movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 432(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm1[0,0,1,1]
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm15[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm0[2],xmm6[3],xmm0[3]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm6[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 320(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm0[0,0,1,1]
+; SSE-NEXT: movdqa %xmm12, %xmm3
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm4[2,2,3,3]
+; SSE-NEXT: movdqa %xmm11, %xmm15
+; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm11[2],xmm4[3],xmm11[3]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm4[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 432(%rdi), %xmm7
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm7[0,0,1,1]
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm13, %xmm3
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,2,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm5[2,2,3,3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm10[2],xmm4[3],xmm10[3]
+; SSE-NEXT: movdqa %xmm10, %xmm13
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm4[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 544(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,0,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm8[2],xmm1[3],xmm8[3]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 544(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; SSE-NEXT: movdqa %xmm8, %xmm5
+; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,2,3,3]
+; SSE-NEXT: punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
+; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm3[0],xmm5[1]
+; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 656(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm5, %xmm4
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm12[2],xmm3[3],xmm12[3]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm3[0],xmm4[1]
+; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 768(%rdi), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm2, %xmm3
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm11[2],xmm1[3],xmm11[3]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 656(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm15, %xmm3
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 768(%rdi), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT: movdqa %xmm9, %xmm3
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: pshufd $250, (%rsp), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm14[2],xmm3[3],xmm14[3]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm3[0],xmm4[1]
+; SSE-NEXT: movapd %xmm4, (%rsp) # 16-byte Spill
; SSE-NEXT: movdqa 880(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm6, %xmm4
+; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = mem[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm11[2],xmm3[3],xmm11[3]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm3[0],xmm4[1]
+; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm10, %xmm3
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,2,3,3]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm9, %xmm3
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm14, %xmm3
-; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm9, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[2,2,3,3]
-; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm8, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
-; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm13, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm11, %xmm1
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $255, (%rsp), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm7, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[2,2,3,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[3,3,3,3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm15, %xmm1
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[2,2,3,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm7, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
+; SSE-NEXT: movdqa %xmm8, %xmm13
+; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,2,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[3,3,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm5, %xmm1
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,2,2,2]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT: pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,2,2,2]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,2,2,2]
+; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
@@ -3222,81 +3245,48 @@
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,2,2,2]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[2,2,2,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[2,2,2,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, %xmm12
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm12, %xmm14
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[2,2,2,2]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[2,3,2,3]
+; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[2,3,2,3]
+; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT: # xmm10 = mem[0,0,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm1[0],xmm10[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[2,3,2,3]
-; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT: # xmm13 = mem[0,0,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm1[0],xmm13[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = mem[2,3,2,3]
-; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; SSE-NEXT: # xmm8 = mem[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
@@ -3306,7 +3296,7 @@
; SSE-NEXT: # xmm7 = mem[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
@@ -3316,7 +3306,7 @@
; SSE-NEXT: # xmm6 = mem[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
@@ -3326,17 +3316,16 @@
; SSE-NEXT: # xmm5 = mem[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = mem[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = mem[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
@@ -3346,66 +3335,66 @@
; SSE-NEXT: # xmm3 = mem[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
; SSE-NEXT: pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[2,3,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm14[2,3,2,3]
; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = mem[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 96(%rsi)
+; SSE-NEXT: movaps %xmm1, 80(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 112(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 48(%rsi)
+; SSE-NEXT: movaps %xmm1, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 64(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 80(%rsi)
+; SSE-NEXT: movaps %xmm1, 112(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 16(%rsi)
+; SSE-NEXT: movaps %xmm1, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 96(%rdx)
+; SSE-NEXT: movaps %xmm1, 96(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 112(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%rdx)
+; SSE-NEXT: movaps %xmm1, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 80(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 96(%rcx)
+; SSE-NEXT: movaps %xmm1, 64(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 112(%rcx)
+; SSE-NEXT: movaps %xmm1, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 64(%rcx)
+; SSE-NEXT: movaps %xmm1, 112(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 80(%rcx)
+; SSE-NEXT: movaps %xmm1, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%rcx)
+; SSE-NEXT: movaps %xmm1, 96(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 48(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%rcx)
+; SSE-NEXT: movaps %xmm1, 32(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 112(%r8)
+; SSE-NEXT: movaps %xmm1, (%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 80(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 64(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 96(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%r8)
+; SSE-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 96(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 80(%r8)
@@ -3421,7 +3410,7 @@
; SSE-NEXT: movaps %xmm1, (%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 112(%r9)
-; SSE-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 96(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 80(%r9)
@@ -3436,7 +3425,8 @@
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, (%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movapd %xmm12, 112(%rax)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 96(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -3459,13 +3449,13 @@
; SSE-NEXT: movapd %xmm6, 48(%rax)
; SSE-NEXT: movapd %xmm7, 32(%rax)
; SSE-NEXT: movapd %xmm8, 16(%rax)
-; SSE-NEXT: movapd %xmm13, (%rax)
+; SSE-NEXT: movapd %xmm10, (%rax)
; SSE-NEXT: addq $1176, %rsp # imm = 0x498
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i32_stride7_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1496, %rsp # imm = 0x5D8
+; AVX1-ONLY-NEXT: subq $1528, %rsp # imm = 0x5F8
; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm4
; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm3
@@ -3476,14 +3466,14 @@
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm10[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm1
@@ -3491,9 +3481,9 @@
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm6[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3503,8 +3493,8 @@
; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm4[6],ymm3[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm7[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm2
@@ -3512,9 +3502,9 @@
; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm11[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3526,24 +3516,23 @@
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm15
+; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm14
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm12
; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm12[0],ymm1[0],ymm12[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm2
; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm14
+; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm15
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -3554,14 +3543,15 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm10[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 752(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm13
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3574,25 +3564,25 @@
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[1,1],ymm0[2,2],ymm6[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm1[2,2],ymm0[5,5],ymm1[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm8[1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm6[2,3],ymm9[0,1]
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm9[0,1]
; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,0],ymm1[3,3],ymm6[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,0],ymm1[3,3],ymm4[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1,2],mem[0]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm6[2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3602,439 +3592,434 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm7[1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm8[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm13
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm13[2,3],ymm11[0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm13[0,0],ymm2[3,3],ymm13[4,4],ymm2[7,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
-; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = zero,xmm2[1,2],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm7[0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm11[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[1,1],ymm0[2,2],ymm4[5,5],ymm0[6,6]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0],xmm15[1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm1[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm12[1,1],ymm0[2,2],ymm12[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm14[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm1[2,3],ymm0[0,1]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm1[2,3],ymm2[0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm1[0,0],ymm5[3,3],ymm1[4,4],ymm5[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm5 = zero,xmm5[1,2],xmm14[2]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm5 = zero,xmm5[1,2],xmm15[2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm5[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[1,1],ymm6[2,2],ymm4[5,5],ymm6[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[1,1],ymm6[2,2],ymm13[5,5],ymm6[6,6]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm1[0],xmm10[1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm2[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm5[2,3],ymm3[0,1]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm5[0,0],ymm15[3,3],ymm5[4,4],ymm15[7,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm5, %ymm12
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm12
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm15 = zero,xmm15[1,2],xmm2[2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm12[0],xmm2[1],xmm12[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2],ymm0[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm1[2,3],ymm3[0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm1[0,0],ymm15[3,3],ymm1[4,4],ymm15[7,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm15 = zero,xmm15[1,2],xmm0[2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm8[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm0[1],xmm14[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm5[1],xmm14[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm15[3,1],mem[0,3],ymm15[7,5],mem[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm5[2,1],ymm15[2,0],ymm5[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm11[0,3],ymm15[7,5],ymm11[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm10[2,1],ymm15[2,0],ymm10[6,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm5[0],ymm9[0],ymm5[2],ymm9[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm4[0],ymm9[0],ymm4[2],ymm9[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm15 = xmm15[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm7[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm14[0],mem[1],xmm14[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm8[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm1[1],xmm14[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm9[0,3],ymm15[7,5],ymm9[4,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm8[2,1],ymm15[2,0],ymm8[6,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm13[0],ymm11[0],ymm13[2],ymm11[2]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm4[0],ymm7[0],ymm4[2],ymm7[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm15 = xmm15[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm10[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm1[1],xmm14[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm12[1],xmm14[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[3,1],ymm6[0,3],ymm15[7,5],ymm6[4,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm4[2,1],ymm15[2,0],ymm4[6,5],ymm15[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm13[2,1],ymm15[2,0],ymm13[6,5],ymm15[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm12[0],ymm3[0],ymm12[2],ymm3[2]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm2[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm0[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm14[0,1,2,3,4],ymm6[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm15[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm2[1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm0[1],xmm6[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm3[0,3],ymm14[7,5],ymm3[4,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm7[2,1],ymm14[2,0],ymm7[6,5],ymm14[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm11[0],ymm4[0],ymm11[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm12[0],ymm15[0],ymm12[2],ymm15[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm14[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,0],ymm1[0,0],ymm6[5,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm1[3,1],ymm6[0,2],ymm1[7,5],ymm6[4,6]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = xmm0[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm11[1,0],ymm10[0,0],ymm11[5,4],ymm10[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm10[3,1],ymm6[0,2],ymm10[7,5],ymm6[4,6]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = xmm5[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm10[3,2,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[0,1],ymm14[1,3],ymm0[4,5],ymm14[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm5[0,2],ymm10[2,0],ymm5[4,6],ymm10[6,4]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm2[0,1],ymm11[1,3],ymm2[4,5],ymm11[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm2[0,2],ymm10[2,0],ymm2[4,6],ymm10[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm10[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm9[1,0],ymm8[0,0],ymm9[5,4],ymm8[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[3,1],ymm6[0,2],ymm8[7,5],ymm6[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = xmm12[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm9 = xmm1[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm9[3,2,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm1[0,1],ymm0[1,3],ymm1[4,5],ymm0[5,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm13[0,2],ymm10[2,0],ymm13[4,6],ymm10[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3,4],ymm10[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm6[0,1],ymm14[1,3],ymm6[4,5],ymm14[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm4[0,2],ymm10[2,0],ymm4[4,6],ymm10[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4],ymm10[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm3[1,0],ymm7[0,0],ymm3[5,4],ymm7[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[3,1],ymm8[0,2],ymm7[7,5],ymm8[4,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm2[0,1,2],xmm15[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm0[0,1,2],xmm13[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm8[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %ymm4, %ymm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm2[0,1],ymm4[1,3],ymm2[4,5],ymm4[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm11[0,2],ymm9[2,0],ymm11[4,6],ymm9[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3,4],ymm9[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,1],ymm15[1,3],ymm9[4,5],ymm15[5,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm12, %ymm1
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm12[0,2],ymm8[2,0],ymm12[4,6],ymm8[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm3[1,0],ymm4[0,0],ymm3[5,4],ymm4[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[3,1],ymm5[0,2],ymm4[7,5],ymm5[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = xmm7[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm3[1,0],ymm0[0,0],ymm3[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm0[3,1],ymm5[0,2],ymm0[7,5],ymm5[4,6]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2],xmm0[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm8[0,1],ymm6[1,3],ymm8[4,5],ymm6[5,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm8, %ymm15
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm9[0,2],ymm5[2,0],ymm9[4,6],ymm5[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm5[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm13[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm13[3,0],ymm4[0,0],ymm13[7,4],ymm4[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm1[1,0],ymm0[2,0],ymm1[5,4],ymm0[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[2,0],ymm4[2,0],ymm3[6,4],ymm4[6,4]
-; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm1[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm13[0,1,2],xmm5[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm12[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = mem[0],xmm8[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm8[0,1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,1],ymm13[1,3],ymm15[4,5],ymm13[5,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm8[1,0],ymm14[2,0],ymm8[5,4],ymm14[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm0[2,0],ymm3[2,0],ymm0[6,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm4[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm0[0,1,2],xmm5[3]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm12 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = mem[0],xmm12[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm12[0,1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm11[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm11[3,0],ymm3[0,0],ymm11[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm11[1,0],ymm10[2,0],ymm11[5,4],ymm10[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm2[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm5[3]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm12 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = mem[0],xmm12[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm12[0,1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,0],ymm1[0,0],ymm9[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm5
-; AVX1-ONLY-NEXT: vmovaps %ymm15, %ymm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm15[1,0],ymm5[2,0],ymm15[5,4],ymm5[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[2,0],ymm2[2,0],ymm1[6,4],ymm2[6,4]
-; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm1[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm1[0,1,2],xmm12[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm7[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = mem[0],xmm14[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1],xmm12[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = ymm8[2,1],mem[3,3],ymm8[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = mem[0],xmm3[1],mem[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm14[1,0],ymm2[2,0],ymm14[5,4],ymm2[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm4[0,1,2],xmm0[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm3[0,0],mem[1,0],ymm3[4,4],mem[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm14[0,1],xmm0[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm8[0,2],ymm5[2,0],ymm8[4,6],ymm5[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm4[2,1],mem[3,3],ymm4[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm10[0],xmm8[1],xmm10[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[1,0],ymm0[2,0],ymm14[5,4],ymm0[6,4]
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = mem[0,1,2],xmm13[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm12[0,0],ymm15[1,0],ymm12[4,4],ymm15[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm14[0,1],xmm9[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm11[2,1],mem[3,3],ymm11[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, (%rsp), %xmm11, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = mem[0],xmm11[1],mem[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[1,0],ymm0[2,0],ymm14[5,4],ymm0[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,1,2],xmm2[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm14[0,0],mem[1,0],ymm14[4,4],mem[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm14[0,1],xmm13[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[2,1],ymm5[3,3],ymm6[6,5],ymm5[7,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm9
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm3[0],xmm2[1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[1,0],ymm0[2,0],ymm14[5,4],ymm0[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm7[0,1,2],xmm1[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm2[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[3,0],ymm4[0,0],ymm2[7,4],ymm4[4,4]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm5[0,0],ymm6[1,0],ymm5[4,4],ymm6[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm14[0,1],xmm1[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm4[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm4[3,0],ymm0[0,0],ymm4[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm8[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm10[1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
-; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm12[1,0],ymm15[2,0],ymm12[5,4],ymm15[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,0],ymm1[0,0],ymm9[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0],xmm3[1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,0],ymm4[4,5],ymm1[6,4]
-; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1,2],xmm7[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm5[1,0],ymm6[2,0],ymm5[5,4],ymm6[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm5[1,0],ymm11[2,0],ymm5[5,4],ymm11[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,0],ymm4[2,0],ymm3[6,4],ymm4[6,4]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm10[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = mem[0],xmm7[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,0],ymm3[0,0],ymm0[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[1,0],ymm14[2,0],ymm6[5,4],ymm14[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[2,0],ymm3[2,0],ymm2[6,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = mem[0],xmm7[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[3,0],ymm3[0,0],ymm1[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm9[1,0],ymm14[2,0],ymm9[5,4],ymm14[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[2,0],ymm3[2,0],ymm1[6,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm1[0,1,2],xmm4[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = mem[0],xmm7[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm8[3,0],ymm3[0,0],ymm8[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm15[1,0],ymm13[2,0],ymm15[5,4],ymm13[6,4]
+; AVX1-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm3[2,0],ymm0[6,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm11[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = mem[0],xmm13[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm13[0,1],xmm3[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm5[2,1],mem[3,3],ymm5[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[0],xmm2[1],mem[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[1,0],ymm0[2,0],ymm13[5,4],ymm0[6,4]
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = mem[0,1,2],xmm10[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm3[0,0],mem[1,0],ymm3[4,4],mem[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm13[0,1],xmm10[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm10 = ymm6[2,1],mem[3,3],ymm6[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm2[0],xmm0[1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm13[1,0],ymm10[2,0],ymm13[5,4],ymm10[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm5[0,1,2],xmm12[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm3[0,0],ymm4[1,0],ymm3[4,4],ymm4[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm13[0,1],xmm12[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm9[2,1],ymm14[3,3],ymm9[6,5],ymm14[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm9[0],xmm7[1],xmm9[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm13[1,0],ymm12[2,0],ymm13[5,4],ymm12[6,4]
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm10[0,0],ymm14[1,0],ymm10[4,4],ymm14[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm13[0,1],xmm1[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm15[2,1],mem[3,3],ymm15[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm12[0],mem[1],xmm12[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[1,0],ymm1[2,0],ymm13[5,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm11[0,1,2],xmm8[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm15 = ymm11[0,0],mem[1,0],ymm11[4,4],mem[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm15[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm15[0,1],xmm13[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm13[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,0],ymm1[0,0],ymm6[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm2[1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,0],ymm6[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1,2],xmm5[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm3[1,0],ymm4[2,0],ymm3[5,4],ymm4[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm4[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[3,0],ymm1[0,0],ymm2[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,0],ymm4[4,5],ymm1[6,4]
-; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm8 = ymm8[1,0],mem[2,0],ymm8[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm6[0],mem[1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,0],ymm6[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm6[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = ymm0[1,0],mem[2,0],ymm0[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,0],ymm1[0,0],ymm0[7,4],ymm1[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm9[1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,0],ymm6[4,5],ymm1[6,4]
+; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = xmm6[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm10[1,0],ymm14[2,0],ymm10[5,4],ymm14[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1],xmm6[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm0[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm0[3,0],ymm6[0,0],ymm0[7,4],ymm6[4,4]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0],xmm12[1],xmm8[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[0,1],ymm6[2,0],ymm8[4,5],ymm6[6,4]
+; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm8[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm11[1,0],mem[2,0],ymm11[5,4],mem[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm8[0,1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[3,0],ymm1[0,0],ymm2[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm11[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, (%rsp), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm7[0],mem[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,0],ymm7[4,5],ymm1[6,4]
-; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm7[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = ymm2[1,0],mem[2,0],ymm2[5,4],mem[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm7[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rsi)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm14, 96(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm13, 32(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rax)
-; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm15, 96(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rax)
-; AVX1-ONLY-NEXT: addq $1496, %rsp # imm = 0x5D8
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
+; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rax)
+; AVX1-ONLY-NEXT: addq $1528, %rsp # imm = 0x5F8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -4044,15 +4029,14 @@
; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm9
; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm4
; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vmovdqa 544(%rdi), %ymm10
+; AVX2-SLOW-NEXT: vmovdqa 544(%rdi), %ymm11
; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %ymm7
; AVX2-SLOW-NEXT: vmovdqa 448(%rdi), %ymm8
; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm3
; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm6
-; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX2-SLOW-NEXT: vpbroadcastq 80(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm0 = <0,7,6,u>
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm6[6],ymm3[7]
; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm12
@@ -4071,12 +4055,12 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm7[6],ymm8[7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm8, %ymm6
+; AVX2-SLOW-NEXT: vmovdqa %ymm8, %ymm14
; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpbroadcastq 528(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %xmm3
@@ -4088,6 +4072,7 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm6
; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
@@ -4108,7 +4093,7 @@
; AVX2-SLOW-NEXT: vmovdqa 704(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovdqa 672(%rdi), %ymm3
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm14
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm5
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm3
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4135,13 +4120,13 @@
; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-SLOW-NEXT: vmovdqa 512(%rdi), %ymm8
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm8[2,3],ymm10[4,5],ymm8[6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 512(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm2[2,3],ymm11[4,5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm8
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0],ymm6[1],ymm7[2,3,4],ymm6[5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0],ymm14[1],ymm7[2,3,4],ymm14[5],ymm7[6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -4158,11 +4143,11 @@
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm1[2,3],ymm15[4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0],ymm5[1],ymm4[2,3,4],ymm5[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0],ymm6[1],ymm4[2,3,4],ymm6[5],ymm4[6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -4171,21 +4156,22 @@
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa 832(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 800(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqa 832(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT: vmovdqa 800(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10,11],ymm1[28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm15
+; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-SLOW-NEXT: vmovdqa 736(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm2[2,3],ymm9[4,5],ymm2[6,7]
; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm6
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm14[1],ymm3[2,3,4],ymm14[5],ymm3[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm5[1],ymm3[2,3,4],ymm5[5],ymm3[6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -4197,15 +4183,14 @@
; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm2
; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm3
; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm4
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm5
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm5
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm4
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm0
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1],ymm0[2,3],ymm7[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1],ymm0[2,3],ymm10[4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
@@ -4216,60 +4201,61 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 80(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm2 = ymm7[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm2 = ymm10[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %xmm2
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0],xmm2[1],xmm3[2,3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
-; AVX2-SLOW-NEXT: vpbroadcastd 204(%rdi), %ymm15
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
+; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm10
+; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpbroadcastd 204(%rdi), %ymm14
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm14[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 528(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm3 = ymm10[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm3 = ymm11[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vpbroadcastd 456(%rdi), %xmm15
+; AVX2-SLOW-NEXT: vpbroadcastd 456(%rdi), %xmm14
; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0],xmm1[1],xmm14[2,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1],ymm3[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm14 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
; AVX2-SLOW-NEXT: vpbroadcastd 652(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm14[0,1,2,3,4,5,6],ymm12[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm12[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 752(%rdi), %xmm3
; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm12 = ymm9[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2],ymm3[3],ymm12[4,5,6,7]
-; AVX2-SLOW-NEXT: vpbroadcastd 680(%rdi), %xmm15
+; AVX2-SLOW-NEXT: vpbroadcastd 680(%rdi), %xmm14
; AVX2-SLOW-NEXT: vmovdqa 704(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm3[1],xmm15[2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0],xmm3[1],xmm14[2,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm14[0,1],ymm12[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm11[0],ymm9[0],ymm11[2],ymm9[2]
+; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm14 = ymm15[0],ymm9[0],ymm15[2],ymm9[2]
; AVX2-SLOW-NEXT: vpbroadcastd 876(%rdi), %ymm13
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5,6],ymm13[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm13[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 304(%rdi), %xmm12
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm13 = ymm4[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm13 = ymm5[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6,7]
; AVX2-SLOW-NEXT: vpbroadcastd 232(%rdi), %xmm13
-; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %xmm15
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm13 = xmm13[0],xmm15[1],xmm13[2,3]
+; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %xmm14
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm13 = xmm13[0],xmm14[1],xmm13[2,3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm13 = ymm3[0],ymm5[0],ymm3[2],ymm5[2]
-; AVX2-SLOW-NEXT: vpbroadcastd 428(%rdi), %ymm14
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm13 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX2-SLOW-NEXT: vpbroadcastd 428(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm15[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm13[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd $253, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
@@ -4278,135 +4264,135 @@
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm14[0,2],ymm13[1,3],ymm14[4,6],ymm13[5,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,2],ymm12[1,3],ymm10[4,6],ymm12[5,7]
; AVX2-SLOW-NEXT: vbroadcastss 208(%rdi), %ymm10
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm10[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $253, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[0],ymm0[1],mem[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = ymm11[0],mem[1],ymm11[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,2],ymm7[1,3],ymm8[4,6],ymm7[5,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm7, %ymm13
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,2],ymm8[1,3],ymm7[4,6],ymm8[5,7]
; AVX2-SLOW-NEXT: vbroadcastss 656(%rdi), %ymm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0],ymm6[1],ymm4[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm15[0,1,2],mem[3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0],ymm6[1],ymm5[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm14[0,1,2],mem[3]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm15
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm5[1,3],ymm3[4,6],ymm5[5,7]
-; AVX2-SLOW-NEXT: vmovaps %ymm5, %ymm10
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm10
+; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm15
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm4[1,3],ymm3[4,6],ymm4[5,7]
; AVX2-SLOW-NEXT: vbroadcastss 432(%rdi), %ymm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $2, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm5
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm14
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,2],ymm9[1,3],ymm11[4,6],ymm9[5,7]
; AVX2-SLOW-NEXT: vbroadcastss 880(%rdi), %ymm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm1 = [4,3,4,3]
-; AVX2-SLOW-NEXT: # xmm1 = mem[0,0]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm1, %ymm0
-; AVX2-SLOW-NEXT: vbroadcastss 548(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %xmm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1,2],xmm2[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
-; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm4 = [0,7,0,7,0,7,0,7]
-; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm4, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 660(%rdi), %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vbroadcastss 100(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vbroadcastss 100(%rdi), %xmm1
; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0,1,2],xmm2[3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX2-SLOW-NEXT: vmovaps %ymm14, %ymm7
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm4, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 212(%rdi), %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
+; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm3 = [4,3,4,3]
+; AVX2-SLOW-NEXT: # xmm3 = mem[0,0]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm3, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm4 = [0,7,0,7,0,7,0,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm4, %ymm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm12, %ymm5
+; AVX2-SLOW-NEXT: vbroadcastss 212(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm3, %ymm2
+; AVX2-SLOW-NEXT: vbroadcastss 548(%rdi), %xmm7
+; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm7[2,3]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm4, %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm8, %ymm12
+; AVX2-SLOW-NEXT: vbroadcastss 660(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm1, %ymm3
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm3, %ymm7
; AVX2-SLOW-NEXT: vbroadcastss 324(%rdi), %xmm8
; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %xmm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm8 = xmm2[0,1,2],xmm8[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm8[2,3]
-; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm4, %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2,3]
+; AVX2-SLOW-NEXT: vpermps %ymm10, %ymm4, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm15[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 436(%rdi), %ymm9
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm7, %ymm3, %ymm3
+; AVX2-SLOW-NEXT: vbroadcastss 772(%rdi), %xmm8
+; AVX2-SLOW-NEXT: vmovaps 736(%rdi), %xmm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm8 = xmm7[0,1,2],xmm8[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm8[2,3]
+; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm4, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm14, %ymm10
+; AVX2-SLOW-NEXT: vbroadcastss 884(%rdi), %ymm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,2,3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vbroadcastss 772(%rdi), %xmm3
-; AVX2-SLOW-NEXT: vmovaps 736(%rdi), %xmm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1,2],xmm3[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm4, %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vmovaps %ymm5, %ymm12
-; AVX2-SLOW-NEXT: vbroadcastss 884(%rdi), %ymm9
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm9[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0],ymm7[1],ymm13[2,3,4],ymm7[5],ymm13[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0,3,3,5,4,7,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-SLOW-NEXT: vbroadcastss 216(%rdi), %ymm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %xmm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm7[0,1,2],xmm0[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0],ymm6[1],ymm5[2,3,4],ymm6[5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-SLOW-NEXT: vbroadcastss 216(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %xmm14
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm14[0,1,2],xmm0[3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
-; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[1,0,2,3,5,4,6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm3, %xmm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm5[0,1],xmm0[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 544(%rdi), %xmm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7]
; AVX2-SLOW-NEXT: vextractf128 $1, %ymm5, %xmm5
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-SLOW-NEXT: vbroadcastss 664(%rdi), %ymm6
@@ -4419,55 +4405,42 @@
; AVX2-SLOW-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7]
; AVX2-SLOW-NEXT: vextractf128 $1, %ymm5, %xmm5
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm10[0],ymm15[1],ymm10[2,3,4],ymm15[5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-SLOW-NEXT: vbroadcastss 440(%rdi), %ymm6
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm2[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps 768(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm8[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm7[3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm8 = mem[1,0,2,3,5,4,6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm8, %xmm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm8[0,1],xmm5[2,3]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm8 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,0,3,3,5,4,7,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
-; AVX2-SLOW-NEXT: vbroadcastss 888(%rdi), %ymm9
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm5[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm7, %xmm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm7 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm7 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
+; AVX2-SLOW-NEXT: vbroadcastss 888(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastss 584(%rdi), %xmm5
; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vpermps 640(%rdi), %ymm4, %ymm9
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 528(%rdi), %ymm9
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm9[0,1,2],xmm3[3]
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm9, %xmm9
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm9[0,1],xmm3[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm3[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 808(%rdi), %xmm3
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3]
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vpermps 864(%rdi), %ymm4, %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 752(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm14[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm5, %xmm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps 640(%rdi), %ymm4, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 528(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1,2],xmm3[3]
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vbroadcastss 136(%rdi), %xmm3
; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3]
@@ -4475,99 +4448,115 @@
; AVX2-SLOW-NEXT: vpermps 192(%rdi), %ymm4, %ymm5
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 80(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm7[3]
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm7, %xmm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 360(%rdi), %xmm5
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
-; AVX2-SLOW-NEXT: vpermps 416(%rdi), %ymm4, %ymm4
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm14[3]
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm9, %xmm9
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm9[0,1],xmm5[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 360(%rdi), %xmm3
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX2-SLOW-NEXT: vpermps 416(%rdi), %ymm4, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 304(%rdi), %ymm5
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm11[3]
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm7, %xmm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 96(%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 64(%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 96(%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 64(%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rdx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 96(%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 64(%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 96(%r8)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%r8)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 64(%r8)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, (%r8)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 96(%r9)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 32(%r9)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, (%r9)
-; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm5, 64(%r9)
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm10 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm10 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 808(%rdi), %xmm5
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
+; AVX2-SLOW-NEXT: vpermps 864(%rdi), %ymm4, %ymm4
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 752(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 64(%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%r8)
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%r8)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 64(%r8)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%r8)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 96(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 64(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm4, (%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovaps %ymm8, 96(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm7, 96(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm6, 32(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm1, 64(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm0, (%rax)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovaps %ymm4, 32(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm3, (%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm2, 96(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm9, 64(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm3, 32(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm9, (%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm8, 64(%rax)
; AVX2-SLOW-NEXT: addq $1224, %rsp # imm = 0x4C8
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: load_i32_stride7_vf32:
; AVX2-FAST: # %bb.0:
-; AVX2-FAST-NEXT: subq $1192, %rsp # imm = 0x4A8
-; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm8
-; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: subq $1224, %rsp # imm = 0x4C8
+; AVX2-FAST-NEXT: vmovaps 320(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm4
; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm5
; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %ymm13
-; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %ymm7
-; AVX2-FAST-NEXT: vmovdqa 448(%rdi), %ymm9
+; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %ymm9
+; AVX2-FAST-NEXT: vmovdqa 448(%rdi), %ymm8
+; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm6
-; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX2-FAST-NEXT: vpbroadcastq 80(%rdi), %ymm0
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm15
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = <0,7,6,u>
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm6[6],ymm3[7]
; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm11
@@ -4585,15 +4574,11 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1,2,3,4,5],ymm7[6],ymm9[7]
-; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm10
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm9[6],ymm8[7]
; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm7, %ymm6
-; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastq 528(%rdi), %ymm2
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm13, %ymm7
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 608(%rdi), %xmm3
@@ -4605,13 +4590,14 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7]
-; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm14
+; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm15
; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm13
+; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm14
; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastq 304(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %xmm3
@@ -4622,11 +4608,12 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %ymm5
-; AVX2-FAST-NEXT: vmovdqa 672(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6],ymm1[7]
-; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovdqa 672(%rdi), %ymm8
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm2[6],ymm8[7]
+; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm13
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm0
; AVX2-FAST-NEXT: vmovdqa 768(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -4651,11 +4638,8 @@
; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovdqa %ymm15, %ymm8
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm0[2,3],ymm15[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm15
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm6
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm6[2,3],ymm10[4,5],ymm6[6,7]
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm0 = [5,6,5,6,5,6,5,6]
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm3
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [1,0,7,7,5,4,7,7]
@@ -4674,13 +4658,14 @@
; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm4[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm9
-; AVX2-FAST-NEXT: vmovdqa %ymm7, %ymm12
-; AVX2-FAST-NEXT: vmovdqu %ymm7, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0,1],ymm4[2,3],ymm12[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm5
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm6[0],ymm10[1],ymm6[2,3,4],ymm10[5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm1, %ymm4
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
@@ -4695,173 +4680,178 @@
; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm4[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm4[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm4
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm4[2,3],ymm7[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm3
+; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm3[2,3],ymm7[4,5],ymm3[6,7]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd %ymm10, %ymm1, %ymm11
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1,2],ymm3[3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm14[0],ymm15[1],ymm14[2,3,4],ymm15[5],ymm14[6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermd %ymm7, %ymm1, %ymm15
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1,2],ymm3[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %ymm6
-; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %ymm13
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm6[12,13,14,15],ymm13[0,1,2,3,4,5,6,7,8,9,10,11],ymm6[28,29,30,31],ymm13[16,17,18,19,20,21,22,23,24,25,26,27]
-; AVX2-FAST-NEXT: vmovdqa %ymm13, %ymm14
-; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %ymm9
+; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %ymm7
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm9[12,13,14,15],ymm7[0,1,2,3,4,5,6,7,8,9,10,11],ymm9[28,29,30,31],ymm7[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm13[0,1],ymm3[2,3],ymm13[4,5],ymm3[6,7]
+; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm15 = ymm14[0,1],ymm4[2,3],ymm14[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm4, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vpermd %ymm15, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0],ymm8[1],ymm13[2,3,4],ymm8[5],ymm13[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd %ymm11, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpermd %ymm10, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 80(%rdi), %xmm0
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm10[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vpbroadcastd 8(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %xmm1
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0],xmm1[1],xmm2[2,3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm10[0],ymm5[0],ymm10[2],ymm5[2]
-; AVX2-FAST-NEXT: vpbroadcastd 204(%rdi), %ymm11
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm11[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm3[0],ymm8[0],ymm3[2],ymm8[2]
+; AVX2-FAST-NEXT: vpbroadcastd 204(%rdi), %ymm15
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm15[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 528(%rdi), %xmm0
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm2 = ymm12[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm12[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm2 = ymm12[8,9,10,11,12,13,14,15],ymm5[0,1,2,3,4,5,6,7],ymm12[24,25,26,27,28,29,30,31],ymm5[16,17,18,19,20,21,22,23]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2],ymm0[3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 456(%rdi), %xmm11
+; AVX2-FAST-NEXT: vpbroadcastd 456(%rdi), %xmm15
; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %xmm0
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm11 = xmm11[0],xmm0[1],xmm11[2,3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm11 = ymm12[0],ymm9[0],ymm12[2],ymm9[2]
-; AVX2-FAST-NEXT: vpbroadcastd 652(%rdi), %ymm15
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm10[0],ymm5[0],ymm10[2],ymm5[2]
+; AVX2-FAST-NEXT: vpbroadcastd 652(%rdi), %ymm11
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm15[0,1,2,3,4,5,6],ymm11[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm11[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 752(%rdi), %xmm2
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm11 = ymm13[8,9,10,11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7],ymm13[24,25,26,27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm11 = ymm14[8,9,10,11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2],ymm2[3],ymm11[4,5,6,7]
; AVX2-FAST-NEXT: vpbroadcastd 680(%rdi), %xmm15
; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm2[1],xmm15[2,3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm15[0,1],ymm11[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm14[0],ymm6[0],ymm14[2],ymm6[2]
+; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm7[0],ymm9[0],ymm7[2],ymm9[2]
; AVX2-FAST-NEXT: vpbroadcastd 876(%rdi), %ymm14
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm14[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 304(%rdi), %xmm11
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm14 = ymm7[8,9,10,11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm14 = ymm9[8,9,10,11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm14[0,1,2],ymm11[3],ymm14[4,5,6,7]
; AVX2-FAST-NEXT: vpbroadcastd 232(%rdi), %xmm14
-; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %xmm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0],xmm3[1],xmm14[2,3]
+; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %xmm2
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0],xmm2[1],xmm14[2,3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm14[0,1],ymm11[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm14 = ymm2[0],ymm15[0],ymm2[2],ymm15[2]
+; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm14 = ymm15[0],ymm12[0],ymm15[2],ymm12[2]
; AVX2-FAST-NEXT: vpbroadcastd 428(%rdi), %ymm13
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5,6],ymm13[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm13[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = ymm8[0],mem[1],ymm8[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd $253, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0],ymm6[1],mem[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm6[3,1,1,0,7,5,5,4]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1],ymm6[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm10[0,2],ymm5[1,3],ymm10[4,6],ymm5[5,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm8, %ymm14
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm6 = ymm3[0,2],ymm8[1,3],ymm3[4,6],ymm8[5,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm7
; AVX2-FAST-NEXT: vbroadcastss 208(%rdi), %ymm8
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[3,1,1,0,7,5,5,4]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,2],ymm9[1,3],ymm12[4,6],ymm9[5,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm13
+; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm6
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm10[0,2],ymm5[1,3],ymm10[4,6],ymm5[5,7]
; AVX2-FAST-NEXT: vbroadcastss 656(%rdi), %ymm5
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm5[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0],ymm4[1],ymm7[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1,2],mem[3]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0],ymm4[1],ymm9[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],mem[3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm7
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2],ymm15[1,3],ymm2[4,6],ymm15[5,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm15[0,2],ymm12[1,3],ymm15[4,6],ymm12[5,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm12, %ymm10
; AVX2-FAST-NEXT: vbroadcastss 432(%rdi), %ymm4
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm4[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps $2, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,2],ymm13[1,3],ymm14[4,6],ymm13[5,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,2],ymm9[1,3],ymm8[4,6],ymm9[5,7]
; AVX2-FAST-NEXT: vbroadcastss 880(%rdi), %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm0 = [4,3,4,3]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
-; AVX2-FAST-NEXT: vpbroadcastd 548(%rdi), %xmm2
-; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %xmm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1,2],xmm2[3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm11 = [0,7,0,7,0,7,0,7]
-; AVX2-FAST-NEXT: vpermd %ymm12, %ymm11, %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 660(%rdi), %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm4[7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpbroadcastd 100(%rdi), %xmm1
+; AVX2-FAST-NEXT: vpbroadcastd 100(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %xmm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-FAST-NEXT: vpermd %ymm4, %ymm0, %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
-; AVX2-FAST-NEXT: vpermd %ymm10, %ymm11, %ymm4
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 212(%rdi), %ymm5
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1,2],xmm0[3]
+; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm0 = [4,3,4,3]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vpermd %ymm3, %ymm0, %ymm3
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
+; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm11 = [0,7,0,7,0,7,0,7]
+; AVX2-FAST-NEXT: vpermd %ymm7, %ymm11, %ymm3
+; AVX2-FAST-NEXT: vmovdqa %ymm7, %ymm12
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-NEXT: vpbroadcastd 212(%rdi), %ymm4
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
+; AVX2-FAST-NEXT: vpbroadcastd 548(%rdi), %xmm4
+; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %xmm3
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm4 = xmm3[0,1,2],xmm4[3]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; AVX2-FAST-NEXT: vpermd %ymm13, %ymm11, %ymm4
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vpbroadcastd 660(%rdi), %ymm5
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5,6,7]
@@ -4870,38 +4860,40 @@
; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %xmm5
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm5[0,1,2],xmm6[3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3]
-; AVX2-FAST-NEXT: vpermd %ymm7, %ymm11, %ymm6
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm15, %ymm11, %ymm6
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm13
; AVX2-FAST-NEXT: vpbroadcastd 436(%rdi), %ymm7
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm6 = mem[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vpermd %ymm6, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpbroadcastd 772(%rdi), %xmm6
-; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %xmm7
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
-; AVX2-FAST-NEXT: vpermd %ymm14, %ymm11, %ymm6
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm13, %ymm1
+; AVX2-FAST-NEXT: vpbroadcastd 772(%rdi), %xmm7
+; AVX2-FAST-NEXT: vmovdqa 736(%rdi), %xmm6
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm6[0,1,2],xmm7[3]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3]
+; AVX2-FAST-NEXT: vpermd %ymm8, %ymm11, %ymm7
+; AVX2-FAST-NEXT: vmovdqa %ymm8, %ymm4
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm1
; AVX2-FAST-NEXT: vpbroadcastd 884(%rdi), %ymm8
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm9[0],ymm10[1],ymm9[2,3,4],ymm10[5],ymm9[6,7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = [1,0,3,3,1,0,7,7]
-; AVX2-FAST-NEXT: vpermd %ymm6, %ymm9, %ymm6
-; AVX2-FAST-NEXT: vpbroadcastd 216(%rdi), %ymm8
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm6[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %xmm6
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm6[0,1,2],xmm2[3]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm14[0],ymm12[1],ymm14[2,3,4],ymm12[5],ymm14[6,7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = [1,0,3,3,1,0,7,7]
+; AVX2-FAST-NEXT: vpermd %ymm8, %ymm7, %ymm8
+; AVX2-FAST-NEXT: vpbroadcastd 216(%rdi), %ymm9
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
+; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %xmm15
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm15[0,1,2],xmm2[3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,3,2]
-; AVX2-FAST-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm10[0,1],xmm2[2,3]
+; AVX2-FAST-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm9 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm9, %xmm9
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 544(%rdi), %xmm10
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm10[0,1,2],xmm3[3]
@@ -4910,88 +4902,90 @@
; AVX2-FAST-NEXT: # ymm8 = mem[1,0,2,3,5,4,6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm8
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
-; AVX2-FAST-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm8 = mem[0],ymm12[1],mem[2,3,4],ymm12[5],mem[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm8, %ymm9, %ymm8
-; AVX2-FAST-NEXT: vpbroadcastd 664(%rdi), %ymm12
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm8, %ymm7, %ymm8
+; AVX2-FAST-NEXT: vpbroadcastd 664(%rdi), %ymm9
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %xmm8
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm8[0,1,2],xmm5[3]
+; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %xmm9
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm9[0,1,2],xmm5[3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[0,1,3,2]
+; AVX2-FAST-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm8
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm8[0,1],xmm5[2,3]
+; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = ymm13[0],mem[1],ymm13[2,3,4],mem[5],ymm13[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm8, %ymm7, %ymm8
+; AVX2-FAST-NEXT: vpbroadcastd 440(%rdi), %ymm12
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqa 768(%rdi), %xmm8
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
+; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[0,1,3,2]
; AVX2-FAST-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm12 = mem[1,0,2,3,5,4,6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm12, %xmm12
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm5 = xmm12[0,1],xmm5[2,3]
-; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm12 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm12, %ymm9, %ymm12
-; AVX2-FAST-NEXT: vpbroadcastd 440(%rdi), %ymm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5,6],ymm13[7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa 768(%rdi), %xmm12
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm12[0,1,2],xmm7[3]
-; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm7 = xmm7[0,1,3,2]
-; AVX2-FAST-NEXT: vpshufd $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm13 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm13, %xmm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm7 = xmm13[0,1],xmm7[2,3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0],ymm14[1],ymm1[2,3,4],ymm14[5],ymm1[6,7]
-; AVX2-FAST-NEXT: vpermd %ymm13, %ymm9, %ymm9
-; AVX2-FAST-NEXT: vpbroadcastd 888(%rdi), %ymm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6],ymm13[7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 584(%rdi), %xmm9
-; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3]
-; AVX2-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX2-FAST-NEXT: vpermd 640(%rdi), %ymm11, %ymm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 528(%rdi), %ymm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm10 = xmm13[0,1,2],xmm10[3]
-; AVX2-FAST-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm13 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm13 = ymm13[0],mem[1],ymm13[2,3,4],mem[5],ymm13[6,7]
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm13, %xmm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm10 = xmm13[0,1],xmm10[2,3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 808(%rdi), %xmm10
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm12[0,1],xmm6[2,3]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm1[0],ymm4[1],ymm1[2,3,4],ymm4[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermd %ymm12, %ymm7, %ymm7
+; AVX2-FAST-NEXT: vpbroadcastd 888(%rdi), %ymm12
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpbroadcastd 584(%rdi), %xmm7
+; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm7 = xmm7[0],mem[1],xmm7[2,3]
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX2-FAST-NEXT: vpermd 640(%rdi), %ymm11, %ymm12
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-FAST-NEXT: vpbroadcastd 528(%rdi), %ymm12
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm10 = xmm12[0,1,2],xmm10[3]
+; AVX2-FAST-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm12, %xmm12
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-NEXT: vpbroadcastd 136(%rdi), %xmm10
; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm10 = xmm10[0],mem[1],xmm10[2,3]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX2-FAST-NEXT: vpermd 864(%rdi), %ymm11, %ymm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 752(%rdi), %ymm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm13[0,1,2],xmm12[3]
-; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm13 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT: vpermd 192(%rdi), %ymm11, %ymm12
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-FAST-NEXT: vpbroadcastd 80(%rdi), %ymm12
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm12[0,1,2],xmm15[3]
+; AVX2-FAST-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm13 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm13 = ymm13[0],mem[1],ymm13[2,3,4],mem[5],ymm13[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm13, %xmm13
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm12 = xmm13[0,1],xmm12[2,3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 136(%rdi), %xmm12
+; AVX2-FAST-NEXT: vpbroadcastd 360(%rdi), %xmm12
; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
-; AVX2-FAST-NEXT: vpermd 192(%rdi), %ymm11, %ymm13
+; AVX2-FAST-NEXT: vpermd 416(%rdi), %ymm11, %ymm13
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 80(%rdi), %ymm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm13[0,1,2],xmm6[3]
+; AVX2-FAST-NEXT: vpbroadcastd 304(%rdi), %ymm13
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm9 = xmm13[0,1,2],xmm9[3]
; AVX2-FAST-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm13 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm13 = ymm13[0],mem[1],ymm13[2,3,4],mem[5],ymm13[6,7]
; AVX2-FAST-NEXT: vextracti128 $1, %ymm13, %xmm13
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm13[0,1],xmm6[2,3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 360(%rdi), %xmm12
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm9 = xmm13[0,1],xmm9[2,3]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vpbroadcastd 808(%rdi), %xmm12
; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
-; AVX2-FAST-NEXT: vpermd 416(%rdi), %ymm11, %ymm11
+; AVX2-FAST-NEXT: vpermd 864(%rdi), %ymm11, %ymm11
; AVX2-FAST-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm12[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 304(%rdi), %ymm12
+; AVX2-FAST-NEXT: vpbroadcastd 752(%rdi), %ymm12
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm8 = xmm12[0,1,2],xmm8[3]
; AVX2-FAST-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
@@ -5008,10 +5002,10 @@
; AVX2-FAST-NEXT: vmovaps %ymm11, 64(%rsi)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm11, (%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm11, 96(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm11, 32(%rdx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm11, 64(%rdx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
@@ -5022,11 +5016,11 @@
; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%rcx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm1, (%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm11, (%rcx)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 96(%r8)
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 32(%r8)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm1, 64(%r8)
@@ -5035,21 +5029,21 @@
; AVX2-FAST-NEXT: vmovdqa %ymm0, 96(%r9)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, 32(%r9)
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm0, 64(%r9)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm0, (%r9)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm0, 64(%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovdqa %ymm7, 96(%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm6, 96(%rax)
; AVX2-FAST-NEXT: vmovdqa %ymm5, 32(%rax)
; AVX2-FAST-NEXT: vmovdqa %ymm3, 64(%rax)
; AVX2-FAST-NEXT: vmovdqa %ymm2, (%rax)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovdqa %ymm8, 32(%rax)
-; AVX2-FAST-NEXT: vmovdqa %ymm6, (%rax)
-; AVX2-FAST-NEXT: vmovdqa %ymm10, 96(%rax)
-; AVX2-FAST-NEXT: vmovdqa %ymm9, 64(%rax)
-; AVX2-FAST-NEXT: addq $1192, %rsp # imm = 0x4A8
+; AVX2-FAST-NEXT: vmovdqa %ymm8, 96(%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm9, 32(%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm10, (%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm7, 64(%rax)
+; AVX2-FAST-NEXT: addq $1224, %rsp # imm = 0x4C8
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
;
@@ -5059,15 +5053,14 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm9
; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm4
; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm11
; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %ymm7
; AVX2-FAST-PERLANE-NEXT: vmovdqa 448(%rdi), %ymm8
; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm10
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 80(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm0 = <0,7,6,u>
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm6[6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm12
@@ -5086,12 +5079,12 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm7[6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm8, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm8, %ymm14
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 528(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %xmm3
@@ -5103,6 +5096,7 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm6
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
@@ -5123,7 +5117,7 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqa 704(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 672(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm5
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -5150,13 +5144,13 @@
; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm8[2,3],ymm10[4,5],ymm8[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1],ymm2[2,3],ymm11[4,5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0],ymm6[1],ymm7[2,3,4],ymm6[5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0],ymm14[1],ymm7[2,3,4],ymm14[5],ymm7[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -5173,11 +5167,11 @@
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm1[2,3],ymm15[4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0],ymm5[1],ymm4[2,3,4],ymm5[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0],ymm6[1],ymm4[2,3,4],ymm6[5],ymm4[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -5186,21 +5180,22 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 832(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 800(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 832(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 800(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[12,13,14,15],ymm4[0,1,2,3,4,5,6,7,8,9,10,11],ymm1[28,29,30,31],ymm4[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 736(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm2[2,3],ymm9[4,5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm14[1],ymm3[2,3,4],ymm14[5],ymm3[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0],ymm5[1],ymm3[2,3,4],ymm5[5],ymm3[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -5212,15 +5207,14 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm4
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1],ymm0[2,3],ymm7[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1],ymm0[2,3],ymm10[4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
@@ -5231,60 +5225,61 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 80(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm2 = ymm7[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm2 = ymm10[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 8(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0],xmm2[1],xmm3[2,3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 204(%rdi), %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 204(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm14[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 528(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm3 = ymm10[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm3 = ymm11[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2],ymm1[3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 456(%rdi), %xmm15
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 456(%rdi), %xmm14
; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0],xmm1[1],xmm14[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1],ymm3[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm14 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 652(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm14[0,1,2,3,4,5,6],ymm12[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm12[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 752(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm12 = ymm9[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2],ymm3[3],ymm12[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 680(%rdi), %xmm15
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 680(%rdi), %xmm14
; AVX2-FAST-PERLANE-NEXT: vmovdqa 704(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm3[1],xmm15[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0],xmm3[1],xmm14[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm14[0,1],ymm12[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm11[0],ymm9[0],ymm11[2],ymm9[2]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm14 = ymm15[0],ymm9[0],ymm15[2],ymm9[2]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 876(%rdi), %ymm13
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4,5,6],ymm13[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm13[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 304(%rdi), %xmm12
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm13 = ymm4[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm13 = ymm5[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 232(%rdi), %xmm13
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %xmm15
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm13 = xmm13[0],xmm15[1],xmm13[2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %xmm14
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm13 = xmm13[0],xmm14[1],xmm13[2,3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm13 = ymm3[0],ymm5[0],ymm3[2],ymm5[2]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 428(%rdi), %ymm14
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm13 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 428(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm15[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm13[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd $253, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
@@ -5293,135 +5288,135 @@
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm14[0,2],ymm13[1,3],ymm14[4,6],ymm13[5,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm10[0,2],ymm12[1,3],ymm10[4,6],ymm12[5,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 208(%rdi), %ymm10
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm10[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $253, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0],ymm0[1],mem[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm11[0],mem[1],ymm11[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,2],ymm7[1,3],ymm8[4,6],ymm7[5,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,2],ymm8[1,3],ymm7[4,6],ymm8[5,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 656(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0],ymm6[1],ymm4[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm15[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0],ymm6[1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm14[0,1,2],mem[3]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm5[1,3],ymm3[4,6],ymm5[5,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm4[1,3],ymm3[4,6],ymm4[5,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 432(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm14
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,2],ymm9[1,3],ymm11[4,6],ymm9[5,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 880(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm1 = [4,3,4,3]
-; AVX2-FAST-PERLANE-NEXT: # xmm1 = mem[0,0]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm1, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 548(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1,2],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm4 = [0,7,0,7,0,7,0,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm4, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 660(%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 100(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 100(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0,1,2],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm3[0,1,2,3],mem[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm4, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 212(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm3 = [4,3,4,3]
+; AVX2-FAST-PERLANE-NEXT: # xmm3 = mem[0,0]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm3, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm4 = [0,7,0,7,0,7,0,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm4, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm12, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 212(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm3, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 548(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm7[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm4, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 660(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm1, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm3, %ymm7
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 324(%rdi), %xmm8
; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm8 = xmm2[0,1,2],xmm8[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm8[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm4, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm10, %ymm4, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm15[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 436(%rdi), %ymm9
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm7, %ymm3, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 772(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT: vmovaps 736(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm8 = xmm7[0,1,2],xmm8[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm8[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm4, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 884(%rdi), %ymm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2,3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 772(%rdi), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovaps 736(%rdi), %xmm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1,2],xmm3[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm4, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 884(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm9[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0],ymm7[1],ymm13[2,3,4],ymm7[5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0,3,3,5,4,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 216(%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %xmm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm7[0,1,2],xmm0[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0],ymm6[1],ymm5[2,3,4],ymm6[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 216(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %xmm14
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm14[0,1,2],xmm0[3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm3, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm5[0,1],xmm0[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 544(%rdi), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1,2],xmm1[3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm5, %xmm5
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 664(%rdi), %ymm6
@@ -5434,55 +5429,42 @@
; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm5, %xmm5
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm10[0],ymm15[1],ymm10[2,3,4],ymm15[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 440(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm2[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps 768(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm8[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm7[3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm8, %xmm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm8[0,1],xmm5[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm8 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,0,3,3,5,4,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 888(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm5[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm7, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm7 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 888(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm5[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 584(%rdi), %xmm5
; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpermps 640(%rdi), %ymm4, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 528(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm9[0,1,2],xmm3[3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm9, %xmm9
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm9[0,1],xmm3[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm3[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 808(%rdi), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpermps 864(%rdi), %ymm4, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 752(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm14[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps 640(%rdi), %ymm4, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 528(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1,2],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm8 = ymm8[0],mem[1],ymm8[2,3,4],mem[5],ymm8[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 136(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3]
@@ -5490,79 +5472,94 @@
; AVX2-FAST-PERLANE-NEXT: vpermps 192(%rdi), %ymm4, %ymm5
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 80(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm7[3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm7, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 360(%rdi), %xmm5
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps 416(%rdi), %ymm4, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm14[3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm9 = ymm9[0],mem[1],ymm9[2,3,4],mem[5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm9, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm9[0,1],xmm5[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 360(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpermps 416(%rdi), %ymm4, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 304(%rdi), %ymm5
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm11[3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = ymm7[0],mem[1],ymm7[2,3,4],mem[5],ymm7[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm7, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 96(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 32(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 64(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 96(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 32(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 64(%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 32(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 96(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 64(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 96(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 32(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 64(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 96(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 32(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm10 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 808(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps 864(%rdi), %ymm4, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 752(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 96(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 64(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 96(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 96(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 64(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 96(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 64(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, (%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 96(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, 96(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 32(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 64(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, (%rax)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 32(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, (%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 96(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, (%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 64(%rax)
; AVX2-FAST-PERLANE-NEXT: addq $1224, %rsp # imm = 0x4C8
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
@@ -7428,12 +7425,12 @@
;
; AVX1-ONLY-LABEL: load_i32_stride7_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $3144, %rsp # imm = 0xC48
+; AVX1-ONLY-NEXT: subq $3192, %rsp # imm = 0xC78
; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm3
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm5
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %ymm4
; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm1
@@ -7442,7 +7439,7 @@
; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm13
+; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm10
; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6],ymm1[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
@@ -7463,21 +7460,21 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 752(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm0[0],ymm11[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm0[0],ymm5[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps %ymm5, %ymm7
+; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm4
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
@@ -7490,8 +7487,9 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm10[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm5
; AVX1-ONLY-NEXT: vmovaps 1200(%rdi), %xmm1
@@ -7520,11 +7518,11 @@
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %ymm8
; AVX1-ONLY-NEXT: vmovaps 1648(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm8[0],ymm1[0],ymm8[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7547,17 +7545,17 @@
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm3
; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
@@ -7574,28 +7572,29 @@
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm14
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm14[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
+; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6],ymm0[7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7612,9 +7611,9 @@
; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm9[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm13[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7628,52 +7627,54 @@
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 1440(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovaps 1440(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1424(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm7[0],ymm1[0],ymm7[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm15[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm2[1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[1,1],ymm0[2,2],ymm13[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm10[1,1],ymm0[2,2],ymm10[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm13[0],xmm12[1],xmm13[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm15[0],xmm12[1],xmm15[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3],ymm1[0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm8[0,0],ymm1[3,3],ymm8[4,4],ymm1[7,7]
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm8[2]
+; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1,2],mem[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm11[1,1],ymm0[2,2],ymm11[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm7[1,1],ymm0[2,2],ymm7[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm11[0],xmm4[1],xmm11[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm1
@@ -7683,8 +7684,8 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[3,3],ymm2[4,4],ymm1[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm1[1,2],xmm11[2]
+; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = zero,xmm1[1,2],mem[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7692,11 +7693,10 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm5[1,1],ymm0[2,2],ymm5[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0],xmm10[1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm5
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm7[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %ymm1
@@ -7713,13 +7713,12 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[2,2],ymm1[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm8[1,1],ymm0[2,2],ymm8[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm9[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],mem[3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %ymm1
@@ -7736,8 +7735,7 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[2,2],ymm1[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[1,1],ymm0[2,2],ymm3[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7759,7 +7757,7 @@
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[1,1],ymm0[2,2],ymm3[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm14[1,1],ymm0[2,2],ymm14[5,5],ymm0[6,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7774,7 +7772,8 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm0[0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[0,0],ymm3[3,3],ymm1[4,4],ymm3[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm3[1,2],xmm14[2]
+; AVX1-ONLY-NEXT: vinsertps $49, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = zero,xmm3[1,2],mem[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7784,8 +7783,8 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0],xmm6[1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm0[0],mem[1],xmm0[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,0],mem[3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %ymm0
@@ -7795,260 +7794,261 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm0[0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm1[0,0],ymm3[3,3],ymm1[4,4],ymm3[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm3[1,2],xmm9[2]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm3 = zero,xmm3[1,2],xmm13[2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[1,1],ymm0[2,2],ymm7[5,5],ymm0[6,6]
+; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm8[1,1],ymm6[2,2],ymm8[5,5],ymm6[6,6]
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm3[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm0[0],xmm9[1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm7[1,0],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2],ymm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm1[2,3],ymm3[0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm1[0,0],ymm14[3,3],ymm1[4,4],ymm14[7,7]
+; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm4[0],xmm1[1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm13[1,0],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2],ymm3[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm0[2,3],ymm5[0,1]
+; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[0,0],ymm14[3,3],ymm0[4,4],ymm14[7,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm14[1,2],xmm15[2]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm14[1,2],xmm3[2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm12[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],xmm13[1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm12[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0],xmm15[1],xmm13[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[2,1],ymm14[2,0],ymm2[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm12[0,3],ymm14[7,5],ymm12[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[2,1],ymm14[2,0],ymm0[6,5],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm12[0],mem[0],ymm12[2],mem[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0,1,2],xmm8[3]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],xmm8[1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0],xmm11[1],xmm13[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[2,1],ymm14[2,0],ymm0[6,5],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0,1,2],xmm11[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm10[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],xmm5[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[2,1],ymm14[2,0],ymm0[6,5],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[2,1],ymm14[2,0],ymm2[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm15[0],ymm1[0],ymm15[2],ymm1[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm15[0],ymm12[0],ymm15[2],ymm12[2]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],xmm13[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[2,1],ymm14[2,0],ymm0[6,5],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm7[0],mem[1],xmm7[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0],xmm11[1],xmm13[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[2,1],ymm14[2,0],ymm2[6,5],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm7[0],mem[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[2,1],ymm14[2,0],ymm2[6,5],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm6[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm7[0],mem[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[2,1],ymm14[2,0],ymm0[6,5],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm9[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],xmm9[1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm5[0,3],ymm14[7,5],ymm5[4,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[2,1],ymm14[2,0],ymm2[6,5],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm14[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm6[0],ymm3[0],ymm6[2],ymm3[2]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm12[1,0],ymm14[0,0],ymm12[5,4],ymm14[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm14[3,1],ymm7[0,2],ymm14[7,5],ymm7[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = mem[0,1,2],xmm12[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm14[0,1],ymm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm12[0,1],mem[1,3],ymm12[4,5],mem[5,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm12[0,2],ymm14[2,0],ymm12[4,6],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0],ymm14[0,0],ymm7[5,4],ymm14[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm14[3,1],ymm7[0,2],ymm14[7,5],ymm7[4,6]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm8[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm14[0,1],ymm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm10[0],ymm7[0],ymm10[2],ymm7[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm9[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0],xmm9[1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm14[3,1],mem[0,3],ymm14[7,5],mem[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[2,1],ymm14[2,0],ymm2[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm2[0],mem[0],ymm2[2],mem[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm14[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0],xmm4[1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm14[3,1],ymm6[0,3],ymm14[7,5],ymm6[4,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm8[2,1],ymm14[2,0],ymm8[6,5],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm14[0,1,2],xmm3[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm13[0,1,2,3,4],ymm8[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm8[0,1],ymm4[1,3],ymm8[4,5],ymm4[5,7]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = xmm8[0],mem[1],xmm8[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm13[3,1],mem[0,3],ymm13[7,5],mem[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm0[2,1],ymm13[2,0],ymm0[6,5],ymm13[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm13[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = xmm8[0],mem[1],xmm8[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $199, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm13[3,1],mem[0,3],ymm13[7,5],mem[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm0[2,1],ymm13[2,0],ymm0[6,5],ymm13[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm13[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0],xmm1[1],xmm8[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,1],ymm4[0,3],ymm13[7,5],ymm4[4,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm3[2,1],ymm13[2,0],ymm3[6,5],ymm13[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm0[0],ymm6[0],ymm0[2],ymm6[2]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm13[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm13[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,0],ymm0[0,0],ymm8[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[3,1],ymm8[0,2],ymm0[7,5],ymm8[4,6]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm13[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm0[0,1],mem[1,3],ymm0[4,5],mem[5,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[0,2],ymm14[2,0],ymm0[4,6],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0],ymm4[0,0],ymm7[5,4],ymm4[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm4[3,1],ymm7[0,2],ymm4[7,5],ymm7[4,6]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[1,0],ymm13[0,0],ymm0[5,4],ymm13[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm13[3,1],ymm8[0,2],ymm13[7,5],ymm8[4,6]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm14 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm14 = mem[0,1,2],xmm0[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm14[0,1],ymm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1312(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm4[0,1],ymm1[1,3],ymm4[4,5],ymm1[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm15[0,2],ymm14[2,0],ymm15[4,6],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3,4],ymm14[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm4[1,0],ymm1[0,0],ymm4[5,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm1[3,1],ymm7[0,2],ymm1[7,5],ymm7[4,6]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm13[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[3,2,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm14[0,1],ymm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1760(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm1[0,1],ymm10[1,3],ymm1[4,5],ymm10[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm11[0,2],ymm14[2,0],ymm11[4,6],ymm14[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm14[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %ymm2, %ymm0
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[1,0],ymm2[0,0],ymm5[5,4],ymm2[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[0,1],ymm12[1,3],ymm0[4,5],ymm12[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm15[0,2],ymm14[2,0],ymm15[4,6],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[1,0],ymm0[0,0],ymm8[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[3,1],ymm8[0,2],ymm0[7,5],ymm8[4,6]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm11[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm14[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1312(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm0[0,1],ymm7[1,3],ymm0[4,5],ymm7[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm10[0,2],ymm14[2,0],ymm10[4,6],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm5[1,0],ymm0[0,0],ymm5[5,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[3,1],ymm8[0,2],ymm0[7,5],ymm8[4,6]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm14 = xmm9[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[3,2,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm14[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1760(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm14 = ymm0[0,1],mem[1,3],ymm0[4,5],mem[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm2[0,2],ymm14[2,0],ymm2[4,6],ymm14[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4],ymm14[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,0],ymm0[0,0],ymm2[5,4],ymm0[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,1],ymm2[0,2],ymm0[7,5],ymm2[4,6]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = xmm9[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2],xmm0[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,1],ymm3[1,3],ymm0[4,5],ymm3[5,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[0,2],ymm5[2,0],ymm6[4,6],ymm5[6,4]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,1],ymm11[1,3],ymm0[4,5],ymm11[5,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,2],ymm5[2,0],ymm0[4,6],ymm5[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[1,0],ymm0[0,0],ymm1[5,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,1],ymm2[0,2],ymm0[7,5],ymm2[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = xmm10[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[1,0],ymm3[0,0],ymm4[5,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm3[3,1],ymm2[0,2],ymm3[7,5],ymm2[4,6]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm1[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm9[0,1],ymm8[1,3],ymm9[4,5],ymm8[5,7]
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps %ymm6, %ymm14
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,1],ymm6[1,3],ymm0[4,5],ymm6[5,7]
+; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,2],ymm5[2,0],ymm15[4,6],ymm5[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
@@ -8057,18 +8057,18 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm1[1,0],ymm0[0,0],ymm1[5,4],ymm0[4,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm0[3,1],ymm2[0,2],ymm0[7,5],ymm2[4,6]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = xmm7[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm9[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm0[0,1],ymm3[1,3],ymm0[4,5],ymm3[5,7]
-; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vmovaps %ymm0, %ymm12
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm14[0,2],ymm5[2,0],ymm14[4,6],ymm5[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[0,2],ymm5[2,0],ymm6[4,6],ymm5[6,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4],ymm5[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -8080,16 +8080,16 @@
; AVX1-ONLY-NEXT: # xmm5 = xmm4[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm13
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm11[0,1],ymm1[1,3],ymm11[4,5],ymm1[5,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm13[0,1],ymm1[1,3],ymm13[4,5],ymm1[5,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm2[0,2],ymm6[2,0],ymm2[4,6],ymm6[6,4]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4],ymm6[5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm2[0,2],ymm8[2,0],ymm2[4,6],ymm8[6,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4],ymm8[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm2[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,0],ymm5[0,0],ymm2[7,4],ymm5[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm11[1,0],ymm1[2,0],ymm11[5,4],ymm1[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm13[1,0],ymm1[2,0],ymm13[5,4],ymm1[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm5[2,0],ymm0[6,4],ymm5[6,4]
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8097,16 +8097,17 @@
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm1[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm5[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm4[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = mem[0],xmm6[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm4[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = mem[0],xmm7[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm12[3,0],ymm0[0,0],ymm12[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = ymm6[1,0],mem[2,0],ymm6[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm7[1,0],ymm8[2,0],ymm7[5,4],ymm8[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm4[2,0],ymm0[6,4],ymm4[6,4]
; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8120,17 +8121,18 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm14[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[3,0],ymm0[0,0],ymm14[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[1,0],ymm3[2,0],ymm13[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm6[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm6[3,0],ymm0[0,0],ymm6[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[1,0],ymm3[2,0],ymm12[5,4],ymm3[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
-; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm6
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm7[2,3,2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm9[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -8138,9 +8140,9 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm1[1,0],mem[2,0],ymm1[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm5[1,0],mem[2,0],ymm5[5,4],mem[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8156,14 +8158,16 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm15[3,0],ymm0[0,0],ymm15[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm9[1,0],ymm8[2,0],ymm9[5,4],ymm8[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm10[1,0],ymm14[2,0],ymm10[5,4],ymm14[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
-; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0,1,2],xmm1[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = mem[0],xmm4[1],mem[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -8190,9 +8194,8 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[3,0],ymm0[0,0],ymm1[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm1[1,0],mem[2,0],ymm1[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,0],ymm11[2,0],ymm1[5,4],ymm11[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,0],ymm1[2,0],ymm0[6,4],ymm1[6,4]
; AVX1-ONLY-NEXT: vmovaps 1440(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8209,12 +8212,11 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[3,0],ymm0[0,0],ymm2[7,4],ymm0[4,4]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm2[1,0],ymm8[2,0],ymm2[5,4],ymm8[6,4]
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm2[1,0],mem[2,0],ymm2[5,4],mem[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm0[2,0],ymm3[2,0],ymm0[6,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm9[0,1,0,1]
; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm0[0,1,2],xmm10[3]
; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
@@ -8223,53 +8225,69 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = ymm11[2,1],mem[3,3],ymm11[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm13[2,1],mem[3,3],ymm13[6,5],mem[7,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm12[0],xmm9[1],xmm12[2,3]
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = mem[0],xmm12[1],mem[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm10[1,0],ymm3[2,0],ymm10[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = mem[0,1,2],xmm10[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm14[0,0],ymm15[1,0],ymm14[4,4],ymm15[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm11
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm11[0,1],xmm10[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[2,1],ymm8[3,3],ymm7[6,5],ymm8[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = mem[0],xmm11[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm10[1,0],ymm3[2,0],ymm10[5,4],ymm3[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,2],xmm7[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm13[0,0],ymm14[1,0],ymm13[4,4],ymm14[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm10[0,1],xmm7[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = ymm6[2,1],mem[3,3],ymm6[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[0],xmm15[1],mem[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[1,0],ymm3[2,0],ymm7[5,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,2],xmm6[3]
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,2],xmm7[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm10 = ymm10[0,0],mem[1,0],ymm10[4,4],mem[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm10[0,1],xmm7[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm8 = xmm10[0,1],xmm8[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = ymm10[2,1],mem[3,3],ymm10[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[0],xmm7[1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = mem[0],xmm8[1],mem[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm8[1,0],ymm3[2,0],ymm8[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm6[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = ymm8[0,0],mem[1,0],ymm8[4,4],mem[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm8[0,1],xmm7[3,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm5[2,1],mem[3,3],ymm5[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[0],xmm8[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[1,0],ymm3[2,0],ymm7[5,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2],xmm6[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2],xmm5[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm7 = ymm7[0,0],mem[1,0],ymm7[4,4],mem[5,4]
@@ -8286,30 +8304,12 @@
; AVX1-ONLY-NEXT: # xmm6 = mem[0],xmm6[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[1,0],ymm3[2,0],ymm6[5,4],ymm3[6,4]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2],xmm6[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = ymm7[0,0],mem[1,0],ymm7[4,4],mem[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm7[0,1],xmm6[3,2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = ymm3[2,1],mem[3,3],ymm3[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0],xmm6[1],mem[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[1,0],ymm3[2,0],ymm6[5,4],ymm3[6,4]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2],xmm5[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[0,0],ymm7[1,0],ymm6[4,4],ymm7[5,4]
+; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm6 = ymm6[0,0],mem[1,0],ymm6[4,4],mem[5,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm6[0,1],xmm5[3,2]
@@ -8317,9 +8317,9 @@
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = ymm4[2,1],mem[3,3],ymm4[6,5],mem[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[0],xmm5[1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0],xmm7[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm5[1,0],ymm3[2,0],ymm5[5,4],ymm3[6,4]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
@@ -8333,9 +8333,9 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm5[0,1],xmm4[3,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = ymm5[2,1],mem[3,3],ymm5[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm3[2,1],mem[3,3],ymm3[6,5],mem[7,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = mem[0],xmm4[1],mem[2,3]
@@ -8351,14 +8351,14 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm4[0,1],xmm1[3,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[2,1],ymm8[3,3],ymm2[6,5],ymm8[7,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm2[1],mem[2,3]
+; AVX1-ONLY-NEXT: vshufps $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm2[2,1],mem[3,3],ymm2[6,5],mem[7,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $13, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[0],xmm3[1],mem[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[1,0],ymm1[2,0],ymm3[5,4],ymm1[6,4]
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2],xmm0[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm9[0,1,2],xmm0[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $16, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = ymm3[0,0],mem[1,0],ymm3[4,4],mem[5,4]
@@ -8367,17 +8367,18 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm3[0,1],xmm0[3,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm11[3,0],ymm0[0,0],ymm11[7,4],ymm0[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm9[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm12[1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm13[3,0],ymm0[0,0],ymm13[7,4],ymm0[4,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm12[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,0],ymm1[4,5],ymm0[6,4]
; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm13[1,0],ymm14[2,0],ymm13[5,4],ymm14[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm14[1,0],ymm15[2,0],ymm14[5,4],ymm15[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
@@ -8385,7 +8386,7 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[3,0],ymm1[0,0],ymm2[7,4],ymm1[4,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm15[2,3,2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm11[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
@@ -8423,8 +8424,7 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm4[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[3,0],ymm3[0,0],ymm4[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm8[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[1],xmm4[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
@@ -8433,18 +8433,37 @@
; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm4 = xmm4[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = ymm5[1,0],mem[2,0],ymm5[5,4],mem[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[2,0,2,3,6,4,6,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm5[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm5[3,0],ymm3[0,0],ymm5[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,0],ymm5[4,5],ymm3[6,4]
+; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm6 = ymm6[1,0],mem[2,0],ymm6[5,4],mem[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm6[0,1],xmm4[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm6[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[3,0],ymm3[0,0],ymm6[7,4],ymm3[4,4]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm7[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = xmm6[0],mem[1],xmm6[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
@@ -8453,8 +8472,9 @@
; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = xmm6[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm8[1,0],ymm7[2,0],ymm8[5,4],ymm7[6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = ymm7[1,0],mem[2,0],ymm7[5,4],mem[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm7[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
@@ -8478,163 +8498,144 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm8[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm5[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm5[3,0],ymm3[0,0],ymm5[7,4],ymm3[4,4]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm8[2,3,0,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm8[3,0],ymm7[0,0],ymm8[7,4],ymm7[4,4]
; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm8 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm8 = xmm8[0],mem[1],xmm8[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm8[0,1],ymm3[2,0],ymm8[4,5],ymm3[6,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,0],ymm8[4,5],ymm7[6,4]
; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm8 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0,1,2],xmm9[3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm9 = ymm9[1,0],mem[2,0],ymm9[5,4],mem[6,4]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,0,2,3,6,4,6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm9[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm9[3,0],ymm8[0,0],ymm9[7,4],ymm8[4,4]
-; AVX1-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = xmm9[0],mem[1],xmm9[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,0],ymm9[4,5],ymm8[6,4]
-; AVX1-ONLY-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = mem[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = xmm9[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps $33, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = ymm5[1,0],mem[2,0],ymm5[5,4],mem[6,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm10[2,0,2,3,6,4,6,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm9, 224(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm9, 160(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm9, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm9, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, (%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, (%r9)
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 128(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 224(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 160(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 128(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 224(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 160(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 128(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 224(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 160(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, (%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 128(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 224(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 160(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 96(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 224(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 160(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 128(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 96(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rax)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rax)
-; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 224(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 160(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 128(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 96(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, (%rax)
+; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 224(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm3, 192(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 160(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 128(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 160(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
-; AVX1-ONLY-NEXT: addq $3144, %rsp # imm = 0xC48
+; AVX1-ONLY-NEXT: addq $3192, %rsp # imm = 0xC78
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-SLOW-LABEL: load_i32_stride7_vf64:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $2664, %rsp # imm = 0xA68
+; AVX2-SLOW-NEXT: subq $2680, %rsp # imm = 0xA78
; AVX2-SLOW-NEXT: vmovdqa 1216(%rdi), %ymm12
+; AVX2-SLOW-NEXT: vmovdqu %ymm12, (%rsp) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 1152(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 1120(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 768(%rdi), %ymm13
; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 704(%rdi), %ymm6
@@ -8681,12 +8682,10 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm6
-; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vpbroadcastq 1200(%rdi), %ymm2
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm12, %ymm7
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 1248(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 1280(%rdi), %xmm3
@@ -8697,11 +8696,11 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 1600(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 1568(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqa 1600(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
+; AVX2-SLOW-NEXT: vmovdqa 1568(%rdi), %ymm12
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5],ymm1[6],ymm12[7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vmovdqa 1664(%rdi), %ymm3
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8738,16 +8737,16 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovdqa 448(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm2[6],ymm5[7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 448(%rdi), %ymm13
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm2[6],ymm13[7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm15
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa 544(%rdi), %ymm3
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 544(%rdi), %ymm6
; AVX2-SLOW-NEXT: vpbroadcastq 528(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 608(%rdi), %xmm3
@@ -8758,19 +8757,16 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 928(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovdqa 896(%rdi), %ymm3
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm14
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm13
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 928(%rdi), %ymm14
+; AVX2-SLOW-NEXT: vmovdqa 896(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm14[6],ymm1[7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa 992(%rdi), %ymm3
+; AVX2-SLOW-NEXT: vmovdqa 992(%rdi), %ymm7
; AVX2-SLOW-NEXT: vpbroadcastq 976(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm12
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqa 1024(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa 1056(%rdi), %xmm3
@@ -8788,7 +8784,7 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
; AVX2-SLOW-NEXT: vpermd %ymm1, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vmovdqa 1440(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpbroadcastq 1424(%rdi), %ymm1
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -8847,20 +8843,21 @@
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa 1280(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vmovdqa 1280(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 1248(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm8[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm8[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-SLOW-NEXT: vmovdqa 1184(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3],ymm7[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3],ymm9[4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0],ymm5[1],mem[2,3,4],ymm5[5],mem[6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -8878,13 +8875,12 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-SLOW-NEXT: vmovdqa 1632(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3],ymm6[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3],ymm5[4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[0],ymm12[1],mem[2,3,4],ymm12[5],mem[6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -8902,11 +8898,10 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-SLOW-NEXT: vmovdqa 512(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3],ymm6[4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0],ymm5[1],ymm15[2,3,4],ymm5[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0],ymm13[1],ymm15[2,3,4],ymm13[5],ymm15[6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -8924,10 +8919,11 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-SLOW-NEXT: vmovdqa 960(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm1[2,3],ymm12[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3],ymm7[4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -8937,16 +8933,17 @@
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vmovdqa 1504(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 1472(%rdi), %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm12
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-SLOW-NEXT: vmovdqa 1408(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm13 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1],ymm1[2,3],ymm13[4,5],ymm1[6,7]
+; AVX2-SLOW-NEXT: vmovdqa 1408(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm7[2,3],ymm15[4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -8960,20 +8957,18 @@
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm13
; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm3
-; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm12
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm13[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm13[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm14
; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm14
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm2[2,3],ymm10[4,5],ymm2[6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm15
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm6[2,3],ymm8[4,5],ymm6[6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -8994,9 +8989,9 @@
; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm9[0],ymm5[0],ymm9[2],ymm5[2]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm10[0],ymm4[0],ymm10[2],ymm4[2]
; AVX2-SLOW-NEXT: vpbroadcastd 428(%rdi), %ymm2
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -9011,54 +9006,68 @@
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
+; AVX2-SLOW-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm3[0],mem[0],ymm3[2],mem[2]
; AVX2-SLOW-NEXT: vpbroadcastd 876(%rdi), %ymm2
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 1200(%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm7[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm9[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vpbroadcastd 1128(%rdi), %xmm1
; AVX2-SLOW-NEXT: vmovdqa 1152(%rdi), %xmm2
; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm8[0],ymm1[2],ymm8[2]
+; AVX2-SLOW-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
; AVX2-SLOW-NEXT: vpbroadcastd 1324(%rdi), %ymm2
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 1648(%rdi), %xmm0
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm6[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm6[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm5[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vpbroadcastd 1576(%rdi), %xmm1
; AVX2-SLOW-NEXT: vmovdqa 1600(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-SLOW-NEXT: vpbroadcastd 1772(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpbroadcastd 1772(%rdi), %ymm2
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa 1424(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm15[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm15[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
+; AVX2-SLOW-NEXT: vpbroadcastd 1352(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovdqa 1376(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm12[0],ymm1[2],ymm12[2]
+; AVX2-SLOW-NEXT: vpbroadcastd 1548(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm7[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 80(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm10[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vpbroadcastd 8(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %xmm4
-; AVX2-SLOW-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %xmm5
+; AVX2-SLOW-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm12[0],ymm14[0],ymm12[2],ymm14[2]
-; AVX2-SLOW-NEXT: vpbroadcastd 204(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm14[0],ymm13[0],ymm14[2],ymm13[2]
+; AVX2-SLOW-NEXT: vpbroadcastd 204(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa 528(%rdi), %xmm0
@@ -9066,89 +9075,90 @@
; AVX2-SLOW-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpbroadcastd 456(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %xmm4
-; AVX2-SLOW-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-SLOW-NEXT: vpbroadcastd 652(%rdi), %ymm15
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 976(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpbroadcastd 904(%rdi), %xmm15
-; AVX2-SLOW-NEXT: vmovdqa 928(%rdi), %xmm12
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm12[1],xmm15[2,3]
+; AVX2-SLOW-NEXT: vpbroadcastd 456(%rdi), %xmm15
+; AVX2-SLOW-NEXT: vmovdqa 480(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
-; AVX2-SLOW-NEXT: vpbroadcastd 1100(%rdi), %ymm14
+; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
+; AVX2-SLOW-NEXT: vpbroadcastd 652(%rdi), %ymm14
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 1424(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm14 = ymm13[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm13[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-SLOW-NEXT: vmovdqa 976(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpalignr {{.*#+}} ymm14 = ymm9[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0,1,2],ymm0[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vpbroadcastd 1352(%rdi), %xmm15
-; AVX2-SLOW-NEXT: vmovdqa 1376(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vpbroadcastd 904(%rdi), %xmm15
+; AVX2-SLOW-NEXT: vmovdqa 928(%rdi), %xmm0
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm4[0],ymm1[0],ymm4[2],ymm1[2]
-; AVX2-SLOW-NEXT: vpbroadcastd 1548(%rdi), %ymm13
+; AVX2-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm7[0],ymm1[0],ymm7[2],ymm1[2]
+; AVX2-SLOW-NEXT: vpbroadcastd 1100(%rdi), %ymm13
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm13 = ymm6[0],mem[1],ymm6[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1,2],mem[3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm13 = ymm5[0],mem[1],ymm5[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm5[0,1,2],mem[3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[3,2,2,3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm9[0,2],ymm5[1,3],ymm9[4,6],ymm5[5,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm10[0,2],ymm4[1,3],ymm10[4,6],ymm4[5,7]
; AVX2-SLOW-NEXT: vbroadcastss 432(%rdi), %ymm14
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm13[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = ymm5[0],mem[1],ymm5[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],mem[3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = ymm4[0],mem[1],ymm4[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm4[0,1,2],mem[3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[3,1,1,0,7,5,5,4]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[0,2],ymm3[1,3],ymm6[4,6],ymm3[5,7]
+; AVX2-SLOW-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7]
; AVX2-SLOW-NEXT: vbroadcastss 880(%rdi), %ymm13
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm13[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm6[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,2,2,3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,2],ymm8[1,3],ymm15[4,6],ymm8[5,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm13[0,2],ymm14[1,3],ymm13[4,6],ymm14[5,7]
; AVX2-SLOW-NEXT: vbroadcastss 1328(%rdi), %ymm6
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm4[0,2],mem[1,3],ymm4[4,6],mem[5,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1776(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
@@ -9157,37 +9167,25 @@
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[3,1,1,0,7,5,5,4]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = ymm14[0,2],mem[1,3],ymm14[4,6],mem[5,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1776(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm15[0,2],mem[1,3],ymm15[4,6],mem[5,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1552(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendd $253, (%rsp), %ymm7, %ymm2 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm2 = mem[0],ymm7[1],mem[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0],ymm8[1],ymm9[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3]
; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[3,1,1,0,7,5,5,4]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,2],ymm1[1,3],ymm4[4,6],ymm1[5,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1552(%rdi), %ymm3
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,2],ymm1[1,3],ymm7[4,6],ymm1[5,7]
+; AVX2-SLOW-NEXT: vmovaps %ymm1, %ymm10
+; AVX2-SLOW-NEXT: vmovaps %ymm7, %ymm8
+; AVX2-SLOW-NEXT: vbroadcastss 1104(%rdi), %ymm3
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} xmm1 = xmm12[0,1,2],mem[3]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-SLOW-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovdqa %ymm11, %ymm8
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,2],ymm10[1,3],ymm11[4,6],ymm10[5,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1104(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
@@ -9196,9 +9194,8 @@
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2],ymm7[1,3],ymm6[4,6],ymm7[5,7]
+; AVX2-SLOW-NEXT: vmovdqa %ymm11, %ymm7
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,2],ymm11[1,3],ymm12[4,6],ymm11[5,7]
; AVX2-SLOW-NEXT: vbroadcastss 656(%rdi), %ymm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -9212,25 +9209,25 @@
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm5[1,3],ymm3[4,6],ymm5[5,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm4[1,3],ymm3[4,6],ymm4[5,7]
; AVX2-SLOW-NEXT: vbroadcastss 208(%rdi), %ymm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vbroadcastss 100(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %xmm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1,2],xmm0[3]
-; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm0 = [4,3,4,3]
-; AVX2-SLOW-NEXT: # xmm0 = mem[0,0]
+; AVX2-SLOW-NEXT: vbroadcastss 100(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
+; AVX2-SLOW-NEXT: vmovddup {{.*#+}} xmm5 = [4,3,4,3]
+; AVX2-SLOW-NEXT: # xmm5 = mem[0,0]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm5, %ymm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm9 = [0,7,0,7,0,7,0,7]
; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm9, %ymm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 212(%rdi), %ymm3
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
@@ -9238,101 +9235,100 @@
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm0, %ymm2
+; AVX2-SLOW-NEXT: vpermps %ymm1, %ymm5, %ymm2
; AVX2-SLOW-NEXT: vbroadcastss 324(%rdi), %xmm3
; AVX2-SLOW-NEXT: vmovaps 288(%rdi), %xmm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 436(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 436(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vbroadcastss 548(%rdi), %xmm5
+; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm5, %ymm3
+; AVX2-SLOW-NEXT: vbroadcastss 548(%rdi), %xmm4
; AVX2-SLOW-NEXT: vmovaps 512(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm5[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm5[2,3]
-; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm9, %ymm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-SLOW-NEXT: vpermps %ymm12, %ymm9, %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm7[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 660(%rdi), %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpermps %ymm3, %ymm5, %ymm4
; AVX2-SLOW-NEXT: vbroadcastss 772(%rdi), %xmm6
; AVX2-SLOW-NEXT: vmovaps 736(%rdi), %xmm3
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpermps %ymm11, %ymm9, %ymm6
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,3]
+; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 884(%rdi), %ymm7
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vbroadcastss 996(%rdi), %xmm6
-; AVX2-SLOW-NEXT: vmovaps 960(%rdi), %xmm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm9, %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm5, %ymm6
+; AVX2-SLOW-NEXT: vbroadcastss 996(%rdi), %xmm7
+; AVX2-SLOW-NEXT: vmovaps 960(%rdi), %xmm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm7 = xmm4[0,1,2],xmm7[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX2-SLOW-NEXT: vpermps %ymm8, %ymm9, %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 1108(%rdi), %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, (%rsp) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vbroadcastss 1220(%rdi), %xmm6
-; AVX2-SLOW-NEXT: vmovaps 1184(%rdi), %xmm13
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm13[0,1,2],xmm6[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm9, %ymm6
-; AVX2-SLOW-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1332(%rdi), %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vbroadcastss 1444(%rdi), %xmm6
-; AVX2-SLOW-NEXT: vmovaps 1408(%rdi), %xmm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm5, %ymm6
+; AVX2-SLOW-NEXT: vbroadcastss 1220(%rdi), %xmm8
+; AVX2-SLOW-NEXT: vmovaps 1184(%rdi), %xmm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm8 = xmm7[0,1,2],xmm8[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
+; AVX2-SLOW-NEXT: vpermps %ymm13, %ymm9, %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1332(%rdi), %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm5, %ymm6
+; AVX2-SLOW-NEXT: vbroadcastss 1444(%rdi), %xmm8
+; AVX2-SLOW-NEXT: vmovaps 1408(%rdi), %xmm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm8 = xmm13[0,1,2],xmm8[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
+; AVX2-SLOW-NEXT: vpermps %ymm15, %ymm9, %ymm8
+; AVX2-SLOW-NEXT: vmovaps %ymm15, %ymm14
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1556(%rdi), %ymm10
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm6 = ymm6[0,1,2,3],mem[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermps %ymm6, %ymm5, %ymm5
+; AVX2-SLOW-NEXT: vbroadcastss 1668(%rdi), %xmm6
+; AVX2-SLOW-NEXT: vmovaps 1632(%rdi), %xmm15
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
; AVX2-SLOW-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1556(%rdi), %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-SLOW-NEXT: vbroadcastss 1668(%rdi), %xmm6
-; AVX2-SLOW-NEXT: vmovaps 1632(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-SLOW-NEXT: vpermps %ymm14, %ymm9, %ymm6
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 1780(%rdi), %ymm8
@@ -9345,33 +9341,33 @@
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-SLOW-NEXT: vbroadcastss 216(%rdi), %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %xmm5
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-SLOW-NEXT: vmovaps 96(%rdi), %xmm6
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm6[0,1,2],xmm0[3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm8 = mem[1,0,2,3,5,4,6,7]
; AVX2-SLOW-NEXT: vextractf128 $1, %ymm8, %xmm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm8[0,1],xmm4[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %xmm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1,2],xmm1[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm8[0,1],xmm0[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 320(%rdi), %xmm0
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
-; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm4, %xmm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm4 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm4 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-SLOW-NEXT: vbroadcastss 440(%rdi), %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-SLOW-NEXT: vbroadcastss 440(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 544(%rdi), %xmm4
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1,2],xmm2[3]
+; AVX2-SLOW-NEXT: vmovaps 544(%rdi), %xmm5
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0,1,2],xmm2[3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm2 = mem[1,0,2,3,5,4,6,7]
@@ -9382,121 +9378,196 @@
; AVX2-SLOW-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-SLOW-NEXT: vbroadcastss 664(%rdi), %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-SLOW-NEXT: vbroadcastss 664(%rdi), %ymm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm8[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 768(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm3[3]
+; AVX2-SLOW-NEXT: vmovaps 768(%rdi), %xmm8
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm3[3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-SLOW-NEXT: vbroadcastss 888(%rdi), %ymm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 992(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm4[3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2]
; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm3 = mem[1,0,2,3,5,4,6,7]
; AVX2-SLOW-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX2-SLOW-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm3 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm3 = mem[0],ymm11[1],mem[2,3,4],ymm11[5],mem[6,7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; AVX2-SLOW-NEXT: vbroadcastss 888(%rdi), %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 992(%rdi), %xmm2
+; AVX2-SLOW-NEXT: vbroadcastss 1112(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1216(%rdi), %xmm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm7[3]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2]
-; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = mem[1,0,2,3,5,4,6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm6, %xmm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm6[0,1],xmm3[2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,0,3,3,5,4,7,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
-; AVX2-SLOW-NEXT: vbroadcastss 1112(%rdi), %ymm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm11 = ymm3[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 1216(%rdi), %xmm3
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm13[3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,3,2]
+; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-SLOW-NEXT: vbroadcastss 1336(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 1440(%rdi), %xmm3
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm13[3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7]
; AVX2-SLOW-NEXT: vextractf128 $1, %ymm7, %xmm7
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm7 = mem[0],ymm7[1],mem[2,3,4],ymm7[5],mem[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm12[0],ymm14[1],ymm12[2,3,4],ymm14[5],ymm12[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-SLOW-NEXT: vbroadcastss 1336(%rdi), %ymm8
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm8 = ymm6[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 1440(%rdi), %xmm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm7 = xmm6[0,1,2],xmm15[3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm7 = xmm7[0,1,3,2]
-; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm13 = mem[1,0,2,3,5,4,6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm13, %xmm13
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm7 = xmm13[0,1],xmm7[2,3]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm13 = mem[0],ymm12[1],mem[2,3,4],ymm12[5],mem[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm13[1,0,3,3,5,4,7,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,0,3]
-; AVX2-SLOW-NEXT: vbroadcastss 1560(%rdi), %ymm15
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm15[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps 1664(%rdi), %xmm15
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
-; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm13 = mem[1,0,2,3,5,4,6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm13, %xmm13
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm13[0,1],xmm0[2,3]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm13 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm13 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm13[1,0,3,3,5,4,7,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,0,3]
+; AVX2-SLOW-NEXT: vbroadcastss 1560(%rdi), %ymm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm7 = ymm4[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps 1664(%rdi), %xmm4
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm13 = xmm4[0,1,2],xmm15[3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm13 = xmm13[0,1,3,2]
+; AVX2-SLOW-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[1,0,2,3,5,4,6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm13 = xmm15[0,1],xmm13[2,3]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm15 = ymm15[1,0,3,3,5,4,7,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm15 = ymm15[0,1,0,3]
; AVX2-SLOW-NEXT: vbroadcastss 1784(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5,6],ymm12[7]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 136(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3]
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpermps 192(%rdi), %ymm9, %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 80(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm12[0,1,2],xmm5[3]
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm12[0,1],xmm5[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 360(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3]
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpermps 416(%rdi), %ymm9, %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 304(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm12 = xmm12[0,1,2],xmm14[3]
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1],xmm12[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm12[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 584(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3]
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpermps 640(%rdi), %ymm9, %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 528(%rdi), %ymm12
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 136(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vpermps 192(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 80(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1],xmm6[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 360(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vpermps 416(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 304(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3]
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm15[0,1],xmm0[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 584(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vpermps 640(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 528(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1,2],xmm5[3]
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1],xmm5[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 808(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vpermps 864(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 752(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1,2],xmm8[3]
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1032(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vpermps 1088(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 976(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm15[0,1,2],xmm1[3]
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1256(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vpermps 1312(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1200(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm15[0,1,2],xmm2[3]
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm15[0,1],xmm2[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1480(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vpermps 1536(%rdi), %ymm9, %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1424(%rdi), %ymm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm15[0,1,2],xmm3[3]
+; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-SLOW-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm15[0,1],xmm3[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1704(%rdi), %xmm12
+; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-SLOW-NEXT: vpermps 1760(%rdi), %ymm9, %ymm9
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm9 = ymm12[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 1648(%rdi), %ymm12
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
@@ -9504,82 +9575,7 @@
; AVX2-SLOW-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
; AVX2-SLOW-NEXT: vextractf128 $1, %ymm12, %xmm12
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm4 = xmm12[0,1],xmm4[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 808(%rdi), %xmm0
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3]
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vpermps 864(%rdi), %ymm9, %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 752(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm12[0,1,2],xmm1[3]
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1032(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3]
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-SLOW-NEXT: vpermps 1088(%rdi), %ymm9, %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 976(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm12[0,1,2],xmm2[3]
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm2 = xmm12[0,1],xmm2[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1256(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3]
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-SLOW-NEXT: vpermps 1312(%rdi), %ymm9, %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1200(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm12[0,1,2],xmm3[3]
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm3 = xmm12[0,1],xmm3[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1480(%rdi), %xmm3
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3]
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vpermps 1536(%rdi), %ymm9, %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1424(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm12[0,1,2],xmm6[3]
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm12[0,1],xmm6[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1704(%rdi), %xmm6
-; AVX2-SLOW-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: # xmm6 = xmm6[0],mem[1],xmm6[2,3]
-; AVX2-SLOW-NEXT: vpermps 1760(%rdi), %ymm9, %ymm9
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 1648(%rdi), %ymm9
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1,2],xmm15[3]
-; AVX2-SLOW-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-SLOW-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm9 = xmm12[0,1],xmm9[2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm9[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm9, 192(%rsi)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
@@ -9613,14 +9609,14 @@
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm9, 32(%rdx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm9, 192(%rcx)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm9, 128(%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm9, 64(%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm9, (%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm9, 192(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm9, 224(%rcx)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm9, 160(%rcx)
@@ -9644,17 +9640,17 @@
; AVX2-SLOW-NEXT: vmovaps %ymm9, 96(%r8)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm9, 32(%r8)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm9, 224(%r9)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm9, 192(%r9)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm9, 160(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm8, 224(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm8, 192(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm8, 160(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm8, 128(%r9)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm8, 96(%r9)
; AVX2-SLOW-NEXT: vmovups (%rsp), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm9, 128(%r9)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm9, 96(%r9)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm9, 64(%r9)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm9, 32(%r9)
@@ -9663,7 +9659,8 @@
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT: vmovaps %ymm13, 224(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm7, 192(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm8, 160(%rax)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm7, 160(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm11, 128(%rax)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm7, 96(%rax)
@@ -9674,29 +9671,31 @@
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm7, (%rax)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovaps %ymm6, 224(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm4, 224(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm3, 192(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm2, 160(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm1, 128(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm0, 96(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm4, 64(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm14, 32(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm5, (%rax)
-; AVX2-SLOW-NEXT: addq $2664, %rsp # imm = 0xA68
+; AVX2-SLOW-NEXT: vmovaps %ymm14, 96(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm5, 64(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm6, (%rax)
+; AVX2-SLOW-NEXT: addq $2680, %rsp # imm = 0xA78
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: load_i32_stride7_vf64:
; AVX2-FAST: # %bb.0:
-; AVX2-FAST-NEXT: subq $2648, %rsp # imm = 0xA58
+; AVX2-FAST-NEXT: subq $2680, %rsp # imm = 0xA78
; AVX2-FAST-NEXT: vmovdqa 1216(%rdi), %ymm8
+; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 1152(%rdi), %ymm11
; AVX2-FAST-NEXT: vmovdqa 1120(%rdi), %ymm5
-; AVX2-FAST-NEXT: vmovdqa 768(%rdi), %ymm9
-; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %ymm15
-; AVX2-FAST-NEXT: vmovdqa 672(%rdi), %ymm7
-; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 768(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovdqa 704(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovdqa 672(%rdi), %ymm14
+; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm10
+; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = <0,7,6,u>
@@ -9707,7 +9706,7 @@
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastq 304(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %xmm3
@@ -9718,12 +9717,13 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5],ymm15[6],ymm7[7]
-; AVX2-FAST-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm6[6],ymm14[7]
+; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm10
+; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastq 752(%rdi), %ymm2
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm7
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 800(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 832(%rdi), %xmm3
@@ -9735,8 +9735,6 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm11[6],ymm5[7]
-; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm9
-; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vpbroadcastq 1200(%rdi), %ymm2
@@ -9751,16 +9749,16 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 1600(%rdi), %ymm10
+; AVX2-FAST-NEXT: vmovdqa 1600(%rdi), %ymm2
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 1568(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm10[6],ymm1[7]
-; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vmovdqa 1664(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovdqu %ymm3, (%rsp) # 32-byte Spill
; AVX2-FAST-NEXT: vpbroadcastq 1648(%rdi), %ymm2
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm6
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqa 1696(%rdi), %xmm2
; AVX2-FAST-NEXT: vmovdqa 1728(%rdi), %xmm3
@@ -9812,10 +9810,11 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 928(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 896(%rdi), %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
+; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm15
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vmovdqa 992(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -9860,10 +9859,10 @@
; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm14
-; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1],ymm14[2,3],ymm6[4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1],ymm0[2,3],ymm8[4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} ymm1 = [5,6,5,6,5,6,5,6]
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm1, %ymm3
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = [1,0,7,7,5,4,7,7]
@@ -9887,8 +9886,7 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1],ymm12[2,3],ymm7[4,5],ymm12[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0],ymm14[1],ymm10[2,3,4],ymm14[5],ymm10[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm5, %ymm4
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
@@ -9905,11 +9903,11 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vmovdqa 1184(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1],ymm0[2,3],ymm8[4,5],ymm0[6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm8, %ymm13
-; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm0[2,3],ymm9[4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0],ymm9[1],ymm11[2,3,4],ymm9[5],ymm11[6,7]
+; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm5, %ymm4
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
@@ -9926,11 +9924,13 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vmovdqa 1632(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1],ymm0[2,3],ymm8[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1],ymm0[2,3],ymm6[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm14
+; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm4 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm4, %ymm5, %ymm4
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
@@ -9939,16 +9939,16 @@
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu %ymm0, (%rsp) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm0[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm15
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1],ymm15[2,3],ymm10[4,5],ymm15[6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm11[0,1],ymm0[2,3],ymm11[4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm3
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
@@ -9967,10 +9967,10 @@
; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm0[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm11
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm9[0,1],ymm11[2,3],ymm9[4,5],ymm11[6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 512(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1],ymm0[2,3],ymm13[4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm3
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
@@ -9994,9 +9994,8 @@
; AVX2-FAST-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
+; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm0, %ymm5, %ymm4
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2],ymm3[3,4,5,6,7]
@@ -10005,16 +10004,17 @@
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = mem[2,2,2,2]
; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
; AVX2-FAST-NEXT: vmovdqa 1504(%rdi), %ymm0
+; AVX2-FAST-NEXT: vmovdqa 1472(%rdi), %ymm10
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm0[12,13,14,15],ymm10[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm10[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm15
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 1472(%rdi), %ymm3
-; AVX2-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm3 = ymm0[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm0[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-FAST-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,0]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vmovdqa 1408(%rdi), %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0,1],ymm0[2,3],mem[4,5],ymm0[6,7]
+; AVX2-FAST-NEXT: vmovdqa 1408(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1],ymm6[2,3],ymm4[4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermd %ymm3, %ymm1, %ymm0
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
@@ -10025,11 +10025,13 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 304(%rdi), %xmm0
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm6[8,9,10,11,12,13,14,15],ymm14[0,1,2,3,4,5,6,7],ymm6[24,25,26,27,28,29,30,31],ymm14[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vpbroadcastd 232(%rdi), %xmm1
-; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %xmm6
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm6[1],xmm1[2,3]
+; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %xmm2
+; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -10054,7 +10056,7 @@
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 1200(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm13[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm13[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm9[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vpbroadcastd 1128(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovdqa 1152(%rdi), %xmm2
@@ -10070,21 +10072,35 @@
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 1648(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm14[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vpbroadcastd 1576(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovdqa 1600(%rdi), %xmm2
+; AVX2-FAST-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-FAST-NEXT: vpbroadcastd 1772(%rdi), %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-NEXT: vpbroadcastd 1772(%rdi), %ymm2
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa 1424(%rdi), %xmm0
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm4[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vpbroadcastd 1352(%rdi), %xmm1
+; AVX2-FAST-NEXT: vmovdqa 1376(%rdi), %xmm2
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm10[0],ymm15[0],ymm10[2],ymm15[2]
+; AVX2-FAST-NEXT: vpbroadcastd 1548(%rdi), %ymm7
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm7[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa 80(%rdi), %xmm0
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm10[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm11[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vpbroadcastd 8(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %xmm4
@@ -10092,73 +10108,58 @@
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpunpcklqdq (%rsp), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-FAST-NEXT: vpbroadcastd 204(%rdi), %ymm4
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm4[7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 528(%rdi), %xmm0
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm9[8,9,10,11,12,13,14,15],ymm11[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm11[16,17,18,19,20,21,22,23]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 456(%rdi), %xmm1
-; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %xmm4
-; AVX2-FAST-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-FAST-NEXT: vpbroadcastd 652(%rdi), %ymm15
+; AVX2-FAST-NEXT: vpbroadcastd 204(%rdi), %ymm15
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 976(%rdi), %xmm0
+; AVX2-FAST-NEXT: vmovdqa 528(%rdi), %xmm0
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm1 = ymm13[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm13[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 904(%rdi), %xmm15
-; AVX2-FAST-NEXT: vmovdqa 928(%rdi), %xmm1
+; AVX2-FAST-NEXT: vpbroadcastd 456(%rdi), %xmm15
+; AVX2-FAST-NEXT: vmovdqa 480(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm7[0],ymm10[0],ymm7[2],ymm10[2]
-; AVX2-FAST-NEXT: vpbroadcastd 1100(%rdi), %ymm14
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm9[0],mem[0],ymm9[2],mem[2]
+; AVX2-FAST-NEXT: vpbroadcastd 652(%rdi), %ymm14
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 1424(%rdi), %xmm0
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqa 976(%rdi), %xmm0
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm14 = ymm9[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
+; AVX2-FAST-NEXT: vpalignr {{.*#+}} ymm14 = ymm10[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0,1,2],ymm0[3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT: vpbroadcastd 1352(%rdi), %xmm15
-; AVX2-FAST-NEXT: vmovdqa 1376(%rdi), %xmm0
+; AVX2-FAST-NEXT: vpbroadcastd 904(%rdi), %xmm15
+; AVX2-FAST-NEXT: vmovdqa 928(%rdi), %xmm0
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm4[0],ymm1[0],ymm4[2],ymm1[2]
-; AVX2-FAST-NEXT: vpbroadcastd 1548(%rdi), %ymm13
+; AVX2-FAST-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm7[0],ymm1[0],ymm7[2],ymm1[2]
+; AVX2-FAST-NEXT: vpbroadcastd 1100(%rdi), %ymm13
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm11 = ymm14[0,1,2,3,4],ymm13[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendd $253, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm13 = mem[0],ymm11[1],mem[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm6 = xmm6[0,1,2],mem[3]
-; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[3,2,2,3]
-; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm13 = ymm4[0],mem[1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm6 = xmm4[0,1,2],mem[3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm6 = xmm6[3,2,2,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm13 = ymm15[0,2],ymm3[1,3],ymm15[4,6],ymm3[5,7]
; AVX2-FAST-NEXT: vbroadcastss 432(%rdi), %ymm14
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm13[5,6,7]
-; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm13[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm6 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
@@ -10176,38 +10177,55 @@
; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,2,2,3]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm11[0,2],ymm13[1,3],ymm11[4,6],ymm13[5,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7]
; AVX2-FAST-NEXT: vbroadcastss 1328(%rdi), %ymm6
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm4 = ymm4[0,2],mem[1,3],ymm4[4,6],mem[5,7]
+; AVX2-FAST-NEXT: vbroadcastss 1776(%rdi), %ymm5
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],mem[3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[3,1,1,0,7,5,5,4]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7]
-; AVX2-FAST-NEXT: vbroadcastss 1776(%rdi), %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm12[0,2],mem[1,3],ymm12[4,6],mem[5,7]
+; AVX2-FAST-NEXT: vbroadcastss 1552(%rdi), %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0],ymm8[1],ymm9[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0],ymm8[1],ymm10[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3]
; AVX2-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[3,1,1,0,7,5,5,4]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,2],ymm1[1,3],ymm4[4,6],ymm1[5,7]
-; AVX2-FAST-NEXT: vbroadcastss 1552(%rdi), %ymm3
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,2],ymm1[1,3],ymm7[4,6],ymm1[5,7]
+; AVX2-FAST-NEXT: vmovaps %ymm1, %ymm11
+; AVX2-FAST-NEXT: vmovaps %ymm7, %ymm8
+; AVX2-FAST-NEXT: vbroadcastss 1104(%rdi), %ymm3
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10219,24 +10237,8 @@
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqa %ymm10, %ymm8
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm7[0,2],ymm10[1,3],ymm7[4,6],ymm10[5,7]
-; AVX2-FAST-NEXT: vmovaps %ymm7, %ymm9
-; AVX2-FAST-NEXT: vbroadcastss 1104(%rdi), %ymm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2],ymm7[1,3],ymm6[4,6],ymm7[5,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm9[0,2],ymm7[1,3],ymm9[4,6],ymm7[5,7]
; AVX2-FAST-NEXT: vbroadcastss 656(%rdi), %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -10250,7 +10252,7 @@
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm4[1,3],ymm3[4,6],ymm4[5,7]
; AVX2-FAST-NEXT: vbroadcastss 208(%rdi), %ymm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
@@ -10282,12 +10284,12 @@
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
; AVX2-FAST-NEXT: vpermps %ymm15, %ymm10, %ymm3
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
; AVX2-FAST-NEXT: vbroadcastss 436(%rdi), %ymm4
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
@@ -10296,7 +10298,7 @@
; AVX2-FAST-NEXT: vmovaps 512(%rdi), %xmm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
-; AVX2-FAST-NEXT: vpermps %ymm6, %ymm10, %ymm4
+; AVX2-FAST-NEXT: vpermps %ymm9, %ymm10, %ymm4
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FAST-NEXT: vbroadcastss 660(%rdi), %ymm6
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm6[7]
@@ -10310,8 +10312,8 @@
; AVX2-FAST-NEXT: vmovaps 736(%rdi), %xmm3
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-NEXT: vpermps %ymm12, %ymm10, %ymm6
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-NEXT: vpermps %ymm15, %ymm10, %ymm6
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm14[6,7]
; AVX2-FAST-NEXT: vbroadcastss 884(%rdi), %ymm7
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7]
@@ -10325,8 +10327,8 @@
; AVX2-FAST-NEXT: vmovaps 960(%rdi), %xmm4
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm7 = xmm4[0,1,2],xmm7[3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
-; AVX2-FAST-NEXT: vpermps %ymm9, %ymm10, %ymm7
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm10, %ymm7
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm11[6,7]
; AVX2-FAST-NEXT: vbroadcastss 1108(%rdi), %ymm8
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -10339,8 +10341,9 @@
; AVX2-FAST-NEXT: vmovaps 1184(%rdi), %xmm7
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm8 = xmm7[0,1,2],xmm8[3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm10, %ymm8
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = ymm8[0,1,2,3,4,5],mem[6,7]
; AVX2-FAST-NEXT: vbroadcastss 1332(%rdi), %ymm9
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
@@ -10353,9 +10356,9 @@
; AVX2-FAST-NEXT: vmovaps 1408(%rdi), %xmm8
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm8[0,1,2],xmm9[3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm9[2,3]
-; AVX2-FAST-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm9 = ymm9[0,1,2,3,4,5],mem[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm12, %ymm10, %ymm9
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm12[6,7]
; AVX2-FAST-NEXT: vbroadcastss 1556(%rdi), %ymm11
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6],ymm11[7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm9[4,5,6,7]
@@ -10376,7 +10379,7 @@
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm9[7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm5 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
; AVX2-FAST-NEXT: vmovaps {{.*#+}} ymm14 = [1,0,3,3,1,0,7,7]
@@ -10391,52 +10394,68 @@
; AVX2-FAST-NEXT: vextractf128 $1, %ymm6, %xmm6
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm6[0,1],xmm0[2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 320(%rdi), %xmm0
-; AVX2-FAST-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
+; AVX2-FAST-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 320(%rdi), %xmm6
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm6[0,1,2],xmm1[3]
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm1 = mem[1,0,2,3,5,4,6,7]
; AVX2-FAST-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[0],ymm1[1],mem[2,3,4],ymm1[5],mem[6,7]
; AVX2-FAST-NEXT: vpermps %ymm1, %ymm14, %ymm1
; AVX2-FAST-NEXT: vbroadcastss 440(%rdi), %ymm5
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm5[7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 544(%rdi), %xmm0
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm2[3]
+; AVX2-FAST-NEXT: vmovaps 544(%rdi), %xmm5
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm5[0,1,2],xmm2[3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
+; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm1[0],mem[1],ymm1[2,3,4],mem[5],ymm1[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm1, %ymm14, %ymm1
+; AVX2-FAST-NEXT: vbroadcastss 664(%rdi), %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 768(%rdi), %xmm0
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm3[3]
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm2 = mem[1,0,2,3,5,4,6,7]
; AVX2-FAST-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-FAST-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm2 = mem[0],ymm15[1],mem[2,3,4],ymm15[5],mem[6,7]
; AVX2-FAST-NEXT: vpermps %ymm2, %ymm14, %ymm2
-; AVX2-FAST-NEXT: vbroadcastss 664(%rdi), %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-NEXT: vbroadcastss 888(%rdi), %ymm3
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovaps 768(%rdi), %xmm1
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm3[3]
+; AVX2-FAST-NEXT: vmovaps 992(%rdi), %xmm1
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm4[3]
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2]
; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm3 = mem[1,0,2,3,5,4,6,7]
; AVX2-FAST-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX2-FAST-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm3 = mem[0],ymm12[1],mem[2,3,4],ymm12[5],mem[6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
; AVX2-FAST-NEXT: vpermps %ymm3, %ymm14, %ymm3
-; AVX2-FAST-NEXT: vbroadcastss 888(%rdi), %ymm6
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 992(%rdi), %xmm2
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm4[3]
+; AVX2-FAST-NEXT: vbroadcastss 1112(%rdi), %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 1216(%rdi), %xmm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm7[3]
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2]
; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7]
@@ -10446,235 +10465,219 @@
; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
; AVX2-FAST-NEXT: vpermps %ymm4, %ymm14, %ymm4
-; AVX2-FAST-NEXT: vbroadcastss 1112(%rdi), %ymm12
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm12[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1216(%rdi), %xmm3
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm7 = xmm3[0,1,2],xmm7[3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm7 = xmm7[0,1,3,2]
-; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm12 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm7 = xmm12[0,1],xmm7[2,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm12 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm12, %ymm14, %ymm12
-; AVX2-FAST-NEXT: vbroadcastss 1336(%rdi), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5,6],ymm15[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1440(%rdi), %xmm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm8 = xmm15[0,1,2],xmm8[3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm8 = xmm8[0,1,3,2]
-; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm12 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm8 = xmm12[0,1],xmm8[2,3]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm12 = mem[0],ymm5[1],mem[2,3,4],ymm5[5],mem[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm12, %ymm14, %ymm12
-; AVX2-FAST-NEXT: vbroadcastss 1560(%rdi), %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm12[0,1,2,3,4,5,6],ymm5[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 1664(%rdi), %xmm12
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm12[0,1,2],xmm11[3]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm5 = xmm5[0,1,3,2]
-; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm11 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-NEXT: vextractf128 $1, %ymm11, %xmm11
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm11[0,1],xmm5[2,3]
-; AVX2-FAST-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm11 = mem[0],ymm13[1],mem[2,3,4],ymm13[5],mem[6,7]
-; AVX2-FAST-NEXT: vpermps %ymm11, %ymm14, %ymm11
-; AVX2-FAST-NEXT: vbroadcastss 1784(%rdi), %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,6],ymm14[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm5[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastss 136(%rdi), %xmm5
-; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
-; AVX2-FAST-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-NEXT: vpermps 192(%rdi), %ymm10, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 80(%rdi), %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm14[0,1,2],xmm9[3]
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7]
-; AVX2-FAST-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm14[0,1],xmm9[2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastss 360(%rdi), %xmm5
-; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
-; AVX2-FAST-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-NEXT: vpermps 416(%rdi), %ymm10, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 304(%rdi), %ymm14
-; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm13 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm13 = xmm14[0,1,2],mem[3]
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7]
-; AVX2-FAST-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm13 = xmm14[0,1],xmm13[2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastss 584(%rdi), %xmm5
-; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
-; AVX2-FAST-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-NEXT: vpermps 640(%rdi), %ymm10, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 528(%rdi), %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm14[0,1,2],xmm0[3]
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7]
-; AVX2-FAST-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm14[0,1],xmm0[2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastss 808(%rdi), %xmm0
-; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3]
-; AVX2-FAST-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpermps 864(%rdi), %ymm10, %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 752(%rdi), %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0,1,2],xmm1[3]
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
-; AVX2-FAST-NEXT: vextractf128 $1, %ymm5, %xmm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastss 1032(%rdi), %xmm1
-; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3]
-; AVX2-FAST-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-NEXT: vpermps 1088(%rdi), %ymm10, %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 976(%rdi), %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1,2],xmm2[3]
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
-; AVX2-FAST-NEXT: vextractf128 $1, %ymm5, %xmm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastss 1256(%rdi), %xmm2
-; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3]
-; AVX2-FAST-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vpermps 1312(%rdi), %ymm10, %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 1200(%rdi), %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm5[0,1,2],xmm3[3]
-; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
-; AVX2-FAST-NEXT: vextractf128 $1, %ymm5, %xmm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm5[0,1],xmm3[2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastss 1480(%rdi), %xmm3
-; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3]
-; AVX2-FAST-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-FAST-NEXT: vpermps 1536(%rdi), %ymm10, %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 1424(%rdi), %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm15[3]
+; AVX2-FAST-NEXT: vbroadcastss 1336(%rdi), %ymm7
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1440(%rdi), %xmm3
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm8[3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm8[0,1],xmm4[2,3]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm8 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm8, %ymm14, %ymm8
+; AVX2-FAST-NEXT: vbroadcastss 1560(%rdi), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm8 = ymm4[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT: vmovaps 1664(%rdi), %xmm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm11 = xmm4[0,1,2],xmm11[3]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm11 = xmm11[0,1,3,2]
+; AVX2-FAST-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm11 = xmm15[0,1],xmm11[2,3]
+; AVX2-FAST-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[0],ymm13[1],mem[2,3,4],ymm13[5],mem[6,7]
+; AVX2-FAST-NEXT: vpermps %ymm15, %ymm14, %ymm14
+; AVX2-FAST-NEXT: vbroadcastss 1784(%rdi), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastss 136(%rdi), %xmm14
+; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm14 = xmm14[0],mem[1],xmm14[2,3]
+; AVX2-FAST-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-FAST-NEXT: vpermps 192(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vbroadcastss 80(%rdi), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm15[0,1,2],xmm9[3]
; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
; AVX2-FAST-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1],xmm5[2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastss 1704(%rdi), %xmm5
-; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX2-FAST-NEXT: # xmm5 = xmm5[0],mem[1],xmm5[2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm9 = xmm15[0,1],xmm9[2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastss 360(%rdi), %xmm14
+; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm14 = xmm14[0],mem[1],xmm14[2,3]
+; AVX2-FAST-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-FAST-NEXT: vpermps 416(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vbroadcastss 304(%rdi), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm13 = xmm15[0,1,2],xmm6[3]
+; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm13 = xmm15[0,1],xmm13[2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastss 584(%rdi), %xmm14
+; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm14 = xmm14[0],mem[1],xmm14[2,3]
+; AVX2-FAST-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-FAST-NEXT: vpermps 640(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vbroadcastss 528(%rdi), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm12 = xmm15[0,1,2],xmm5[3]
+; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm12 = xmm15[0,1],xmm12[2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm12[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastss 808(%rdi), %xmm12
+; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT: vpermps 864(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vbroadcastss 752(%rdi), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3]
+; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm0 = xmm15[0,1],xmm0[2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastss 1032(%rdi), %xmm12
+; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT: vpermps 1088(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vbroadcastss 976(%rdi), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm15[0,1,2],xmm1[3]
+; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastss 1256(%rdi), %xmm12
+; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT: vpermps 1312(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vbroadcastss 1200(%rdi), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm15[0,1,2],xmm2[3]
+; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm2 = xmm15[0,1],xmm2[2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastss 1480(%rdi), %xmm12
+; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT: vpermps 1536(%rdi), %ymm10, %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vbroadcastss 1424(%rdi), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm15[0,1,2],xmm3[3]
+; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm3 = xmm15[0,1],xmm3[2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastss 1704(%rdi), %xmm12
+; AVX2-FAST-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
; AVX2-FAST-NEXT: vpermps 1760(%rdi), %ymm10, %ymm10
-; AVX2-FAST-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 1648(%rdi), %ymm10
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0,1,2],xmm12[3]
+; AVX2-FAST-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-NEXT: vbroadcastss 1648(%rdi), %ymm12
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
; AVX2-FAST-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
; AVX2-FAST-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm10 = xmm12[0,1],xmm10[2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 192(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, (%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 224(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 160(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 192(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, (%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 224(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 160(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rdx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 192(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, (%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 224(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 160(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rcx)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, (%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 192(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 224(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 160(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%r8)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%r8)
+; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm4 = xmm12[0,1],xmm4[2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 192(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 128(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 64(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, (%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 224(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 160(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 96(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 32(%rsi)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 192(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 128(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 64(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, (%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 224(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 160(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 96(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 32(%rdx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 128(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 64(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, (%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 192(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 224(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 160(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 96(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 32(%rcx)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, (%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 64(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 128(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 192(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 224(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 160(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 96(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 32(%r8)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm5, 224(%r9)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
@@ -10685,26 +10688,28 @@
; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%r9)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%r9)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%r9)
-; AVX2-FAST-NEXT: vmovups (%rsp), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%r9)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm5, (%r9)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 64(%r9)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 32(%r9)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, (%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-NEXT: vmovaps %ymm11, 224(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm8, 192(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm7, 160(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm4, 128(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm6, 96(%rax)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 64(%rax)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, 32(%rax)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm4, (%rax)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 128(%rax)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 96(%rax)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 64(%rax)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, 32(%rax)
+; AVX2-FAST-NEXT: vmovups (%rsp), %ymm5 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm5, (%rax)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovaps %ymm10, 224(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm4, 224(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm3, 192(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm2, 160(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm1, 128(%rax)
@@ -10712,17 +10717,18 @@
; AVX2-FAST-NEXT: vmovaps %ymm14, 64(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm13, 32(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm9, (%rax)
-; AVX2-FAST-NEXT: addq $2648, %rsp # imm = 0xA58
+; AVX2-FAST-NEXT: addq $2680, %rsp # imm = 0xA78
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
;
; AVX2-FAST-PERLANE-LABEL: load_i32_stride7_vf64:
; AVX2-FAST-PERLANE: # %bb.0:
-; AVX2-FAST-PERLANE-NEXT: subq $2664, %rsp # imm = 0xA68
+; AVX2-FAST-PERLANE-NEXT: subq $2680, %rsp # imm = 0xA78
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1216(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1152(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1120(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 768(%rdi), %ymm13
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 704(%rdi), %ymm6
@@ -10769,12 +10775,10 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm4[6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 1200(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm12, %ymm7
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1248(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1280(%rdi), %xmm3
@@ -10785,11 +10789,11 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 1600(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 1568(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 1600(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 1568(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1,2,3,4,5],ymm1[6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1664(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10826,16 +10830,16 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 448(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm2[6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 448(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1,2,3,4,5],ymm2[6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm15
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 544(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 528(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 576(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 608(%rdi), %xmm3
@@ -10846,19 +10850,16 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 928(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 896(%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm2[6],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm13
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 928(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 896(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm14[6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 992(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 992(%rdi), %ymm7
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 976(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1024(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1056(%rdi), %xmm3
@@ -10876,7 +10877,7 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermd %ymm1, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1440(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq 1424(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3,4,5,6,7]
@@ -10935,20 +10936,21 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 1280(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 1280(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1248(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm8[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm8[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1184(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3],ymm7[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3],ymm9[4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0],ymm5[1],mem[2,3,4],ymm5[5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -10966,13 +10968,12 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1632(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3],ymm6[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3],ymm5[4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0],ymm12[1],mem[2,3,4],ymm12[5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -10990,11 +10991,10 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 512(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd $51, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1],ymm1[2,3],mem[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3],ymm6[4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0],ymm5[1],ymm15[2,3,4],ymm5[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0],ymm13[1],ymm15[2,3,4],ymm13[5],ymm15[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -11012,10 +11012,11 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vmovdqa 960(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm12[0,1],ymm1[2,3],ymm12[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3],ymm7[4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,3,2,3]
@@ -11025,16 +11026,17 @@
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1504(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1472(%rdi), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm1[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm1[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 1408(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm13 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm13[0,1],ymm1[2,3],ymm13[4,5],ymm1[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 1408(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm15[0,1],ymm7[2,3],ymm15[4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -11048,20 +11050,18 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm13
; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm2[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm13[12,13,14,15],ymm3[0,1,2,3,4,5,6,7,8,9,10,11],ymm13[28,29,30,31],ymm3[16,17,18,19,20,21,22,23,24,25,26,27]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm14
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,0]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1],ymm2[2,3],ymm10[4,5],ymm2[6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1],ymm6[2,3],ymm8[4,5],ymm6[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,2,2,3,5,6,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -11082,9 +11082,9 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm9[0],ymm5[0],ymm9[2],ymm5[2]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm10[0],ymm4[0],ymm10[2],ymm4[2]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 428(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -11099,54 +11099,68 @@
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm3[0],mem[0],ymm3[2],mem[2]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 876(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1200(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm7[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm9[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1128(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1152(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm8[0],ymm1[2],ymm8[2]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1324(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1648(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm6[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm6[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm5[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1576(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa 1600(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1772(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1772(%rdi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 1424(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm15[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm15[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1352(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 1376(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm12[0],ymm1[2],ymm12[2]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1548(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm7[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 80(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm10[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm1 = ymm8[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 8(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %xmm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm12[0],ymm14[0],ymm12[2],ymm14[2]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 204(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm14[0],ymm13[0],ymm14[2],ymm13[2]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 204(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa 528(%rdi), %xmm0
@@ -11154,89 +11168,90 @@
; AVX2-FAST-PERLANE-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 456(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %xmm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 652(%rdi), %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 976(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 904(%rdi), %xmm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 928(%rdi), %xmm12
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm12[1],xmm15[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 456(%rdi), %xmm15
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 480(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm1[1],xmm15[2,3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1100(%rdi), %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 652(%rdi), %ymm14
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5,6],ymm14[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm14[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 1424(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm14 = ymm13[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm13[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 976(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpalignr {{.*#+}} ymm14 = ymm9[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0,1,2],ymm0[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1352(%rdi), %xmm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 1376(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 904(%rdi), %xmm15
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 928(%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0],xmm0[1],xmm15[2,3]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm4[0],ymm1[0],ymm4[2],ymm1[2]
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1548(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT: vpunpcklqdq {{.*#+}} ymm15 = ymm7[0],ymm1[0],ymm7[2],ymm1[2]
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastd 1100(%rdi), %ymm13
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3,4,5,6],ymm13[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3,4],ymm13[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = ymm6[0],mem[1],ymm6[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm13 = ymm5[0],mem[1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm5[0,1,2],mem[3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[3,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm13[3,1,1,0,7,5,5,4]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1],ymm13[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm9[0,2],ymm5[1,3],ymm9[4,6],ymm5[5,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm10[0,2],ymm4[1,3],ymm10[4,6],ymm4[5,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 432(%rdi), %ymm14
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm14[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm13[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm5[0],mem[1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm4[0],mem[1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm4[0,1,2],mem[3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm5[3,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[3,1,1,0,7,5,5,4]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[0,2],ymm3[1,3],ymm6[4,6],ymm3[5,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm3[0,2],mem[1,3],ymm3[4,6],mem[5,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 880(%rdi), %ymm13
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm13[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4],ymm6[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm4 = xmm4[3,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[3,1,1,0,7,5,5,4]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm15[0,2],ymm8[1,3],ymm15[4,6],ymm8[5,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm13[0,2],ymm14[1,3],ymm13[4,6],ymm14[5,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1328(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4],ymm5[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm3[0],mem[1],ymm3[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],mem[3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,2,2,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm4[3,1,1,0,7,5,5,4]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm4[0,2],mem[1,3],ymm4[4,6],mem[5,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1776(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm4[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
@@ -11245,37 +11260,25 @@
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[3,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[3,1,1,0,7,5,5,4]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm14[0,2],mem[1,3],ymm14[4,6],mem[5,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1776(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps $216, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm15[0,2],mem[1,3],ymm15[4,6],mem[5,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1552(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendd $253, (%rsp), %ymm7, %ymm2 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0],ymm7[1],mem[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0],ymm8[1],ymm9[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],mem[3]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[3,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[3,1,1,0,7,5,5,4]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm4[0,2],ymm1[1,3],ymm4[4,6],ymm1[5,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1552(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm7[0,2],ymm1[1,3],ymm7[4,6],ymm1[5,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1104(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} xmm1 = xmm12[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[3,2,2,3]
-; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm11[0,2],ymm10[1,3],ymm11[4,6],ymm10[5,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1104(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4,5,6,7]
@@ -11284,9 +11287,8 @@
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,2,2,3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm6[0,2],ymm7[1,3],ymm6[4,6],ymm7[5,7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,2],ymm11[1,3],ymm12[4,6],ymm11[5,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 656(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
@@ -11300,25 +11302,25 @@
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,1,1,0,7,5,5,4]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm5[1,3],ymm3[4,6],ymm5[5,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[0,2],ymm4[1,3],ymm3[4,6],ymm4[5,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 208(%rdi), %ymm2
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 100(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %xmm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1,2],xmm0[3]
-; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm0 = [4,3,4,3]
-; AVX2-FAST-PERLANE-NEXT: # xmm0 = mem[0,0]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 100(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vmovddup {{.*#+}} xmm5 = [4,3,4,3]
+; AVX2-FAST-PERLANE-NEXT: # xmm5 = mem[0,0]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm5, %ymm2
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm9 = [0,7,0,7,0,7,0,7]
; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm9, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 212(%rdi), %ymm3
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
@@ -11326,101 +11328,100 @@
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm0, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm1, %ymm5, %ymm2
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 324(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vmovaps 288(%rdi), %xmm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm1[0,1,2],xmm3[3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 436(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 436(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 548(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm5, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 548(%rdi), %xmm4
; AVX2-FAST-PERLANE-NEXT: vmovaps 512(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0,1,2],xmm5[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm5[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm9, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm2[0,1,2],xmm4[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm12, %ymm9, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm7[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 660(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm3, %ymm5, %ymm4
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 772(%rdi), %xmm6
; AVX2-FAST-PERLANE-NEXT: vmovaps 736(%rdi), %xmm3
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm11, %ymm9, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 884(%rdi), %ymm7
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 996(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vmovaps 960(%rdi), %xmm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1,2],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm9, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm5, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 996(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT: vmovaps 960(%rdi), %xmm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm7 = xmm4[0,1,2],xmm7[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm8, %ymm9, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1108(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, (%rsp) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1220(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1184(%rdi), %xmm13
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm13[0,1,2],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm9, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1332(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1444(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1408(%rdi), %xmm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm5, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1220(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1184(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm8 = xmm7[0,1,2],xmm8[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm13, %ymm9, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1332(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm5, %ymm6
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1444(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1408(%rdi), %xmm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm8 = xmm13[0,1,2],xmm8[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm15, %ymm9, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm15, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1556(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $240, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0,1,2,3],mem[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermps %ymm6, %ymm5, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1668(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1632(%rdi), %xmm15
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
; AVX2-FAST-PERLANE-NEXT: vpermps {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1556(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm0, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1668(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1632(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps %ymm14, %ymm9, %ymm6
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1780(%rdi), %ymm8
@@ -11433,33 +11434,33 @@
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 216(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %xmm5
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 96(%rdi), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm6[0,1,2],xmm0[3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[1,0,2,3,5,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm8, %xmm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm8[0,1],xmm4[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %xmm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1,2],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm8[0,1],xmm0[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 320(%rdi), %xmm0
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm0[0,1,2],xmm1[3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm4, %xmm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1],xmm1[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm4 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 440(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm11[0],mem[1],ymm11[2,3,4],mem[5],ymm11[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm5 = ymm5[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 440(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 544(%rdi), %xmm4
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm4[0,1,2],xmm2[3]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 544(%rdi), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm5[0,1,2],xmm2[3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[1,0,2,3,5,4,6,7]
@@ -11470,121 +11471,196 @@
; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 664(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 664(%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm8[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 768(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 768(%rdi), %xmm8
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm8[0,1,2],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 888(%rdi), %ymm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 992(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1,2],xmm4[3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,1,3,2]
; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[1,0,2,3,5,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm3 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm3 = mem[0],ymm11[1],mem[2,3,4],ymm11[5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm3 = ymm3[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 888(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 992(%rdi), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1112(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5,6],ymm4[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1216(%rdi), %xmm2
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm2[0,1,2],xmm7[3]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm3 = xmm3[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm6, %xmm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm6[0,1],xmm3[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm6 = ymm6[0],mem[1],ymm6[2,3,4],mem[5],ymm6[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm6[1,0,3,3,5,4,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1112(%rdi), %ymm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5,6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm11 = ymm3[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1216(%rdi), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm13[3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm4 = ymm4[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1336(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1440(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1,2],xmm13[3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm4 = xmm4[0,1,3,2]
; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[1,0,2,3,5,4,6,7]
; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm7, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm7 = mem[0],ymm7[1],mem[2,3,4],ymm7[5],mem[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1],xmm4[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm12[0],ymm14[1],ymm12[2,3,4],ymm14[5],ymm12[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,0,3,3,5,4,7,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1336(%rdi), %ymm8
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm8 = ymm6[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1440(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm7 = xmm6[0,1,2],xmm15[3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm7 = xmm7[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm13, %xmm13
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm7 = xmm13[0,1],xmm7[2,3]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps $221, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = mem[0],ymm12[1],mem[2,3,4],ymm12[5],mem[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm13[1,0,3,3,5,4,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,0,3]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1560(%rdi), %ymm15
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5,6],ymm15[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 1664(%rdi), %xmm15
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1,3,2]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = mem[1,0,2,3,5,4,6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm13, %xmm13
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm13[0,1],xmm0[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm13 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm13 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm13[1,0,3,3,5,4,7,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,0,3]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1560(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm7 = ymm4[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 1664(%rdi), %xmm4
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm13 = xmm4[0,1,2],xmm15[3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm13 = xmm13[0,1,3,2]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $225, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[1,0,2,3,5,4,6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm13 = xmm15[0,1],xmm13[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm10[0],mem[1],ymm10[2,3,4],mem[5],ymm10[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm15 = ymm15[1,0,3,3,5,4,7,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm15 = ymm15[0,1,0,3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1784(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2,3,4,5,6],ymm12[7]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 136(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermps 192(%rdi), %ymm9, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 80(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm12[0,1,2],xmm5[3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm12[0,1],xmm5[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 360(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermps 416(%rdi), %ymm9, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 304(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm12 = xmm12[0,1,2],xmm14[3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm14, %xmm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1],xmm12[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm12[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 584(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermps 640(%rdi), %ymm9, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 528(%rdi), %ymm12
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 136(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermps 192(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 80(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1],xmm6[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 360(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermps 416(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 304(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm15[0,1,2],xmm0[3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm15[0,1],xmm0[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 584(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermps 640(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 528(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1,2],xmm5[3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1],xmm5[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 808(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermps 864(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 752(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1,2],xmm8[3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1032(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermps 1088(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 976(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm15[0,1,2],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1256(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermps 1312(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1200(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm15[0,1,2],xmm2[3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm15[0,1],xmm2[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1480(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpermps 1536(%rdi), %ymm9, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1424(%rdi), %ymm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm15[0,1,2],xmm3[3]
+; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = mem[2,3,2,3,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm15 = ymm15[0],mem[1],ymm15[2,3,4],mem[5],ymm15[6,7]
+; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm15[0,1],xmm3[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1704(%rdi), %xmm12
+; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # xmm12 = xmm12[0],mem[1],xmm12[2,3]
+; AVX2-FAST-PERLANE-NEXT: vpermps 1760(%rdi), %ymm9, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm9 = ymm12[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1648(%rdi), %ymm12
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm12[0,1,2],xmm4[3]
; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
@@ -11592,82 +11668,7 @@
; AVX2-FAST-PERLANE-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm12, %xmm12
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm4 = xmm12[0,1],xmm4[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 808(%rdi), %xmm0
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpermps 864(%rdi), %ymm9, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 752(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm12[0,1,2],xmm1[3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1032(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm1 = xmm1[0],mem[1],xmm1[2,3]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpermps 1088(%rdi), %ymm9, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 976(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm12[0,1,2],xmm2[3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm2 = xmm12[0,1],xmm2[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1256(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm2 = xmm2[0],mem[1],xmm2[2,3]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpermps 1312(%rdi), %ymm9, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1200(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm12[0,1,2],xmm3[3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm3 = xmm12[0,1],xmm3[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1480(%rdi), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm3 = xmm3[0],mem[1],xmm3[2,3]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpermps 1536(%rdi), %ymm9, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1424(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm12[0,1,2],xmm6[3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm12[0,1],xmm6[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1704(%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # xmm6 = xmm6[0],mem[1],xmm6[2,3]
-; AVX2-FAST-PERLANE-NEXT: vpermps 1760(%rdi), %ymm9, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 1648(%rdi), %ymm9
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1,2],xmm15[3]
-; AVX2-FAST-PERLANE-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm12 = ymm12[0],mem[1],ymm12[2,3,4],mem[5],ymm12[6,7]
-; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm12, %xmm12
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm9 = xmm12[0,1],xmm9[2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm9[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm9[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 192(%rsi)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
@@ -11701,14 +11702,14 @@
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 32(%rdx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 192(%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 128(%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 64(%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, (%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 192(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 224(%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 160(%rcx)
@@ -11732,17 +11733,17 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 96(%r8)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 32(%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 224(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 192(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 160(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 224(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 192(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 160(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 128(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 96(%r9)
; AVX2-FAST-PERLANE-NEXT: vmovups (%rsp), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 128(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 96(%r9)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 64(%r9)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 32(%r9)
@@ -11751,7 +11752,8 @@
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, 224(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, 192(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm8, 160(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, 160(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm11, 128(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, 96(%rax)
@@ -11762,15 +11764,15 @@
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm7, (%rax)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 224(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 224(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 192(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, 160(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm1, 128(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 96(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 64(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, 32(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, (%rax)
-; AVX2-FAST-PERLANE-NEXT: addq $2664, %rsp # imm = 0xA68
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, 96(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm5, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, (%rax)
+; AVX2-FAST-PERLANE-NEXT: addq $2680, %rsp # imm = 0xA78
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
index 7e73980..b161f24 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
@@ -909,55 +909,55 @@
; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 480(%rdi), %xmm11
; SSE-NEXT: movaps 448(%rdi), %xmm3
-; SSE-NEXT: movaps 160(%rdi), %xmm8
-; SSE-NEXT: movaps 128(%rdi), %xmm15
+; SSE-NEXT: movaps 160(%rdi), %xmm6
+; SSE-NEXT: movaps 128(%rdi), %xmm13
; SSE-NEXT: movaps 224(%rdi), %xmm5
; SSE-NEXT: movaps 192(%rdi), %xmm10
-; SSE-NEXT: movaps %xmm10, %xmm6
-; SSE-NEXT: unpcklps {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; SSE-NEXT: movaps %xmm15, %xmm9
-; SSE-NEXT: unpcklps {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
-; SSE-NEXT: movaps %xmm9, %xmm13
+; SSE-NEXT: movaps %xmm10, %xmm8
+; SSE-NEXT: unpcklps {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
+; SSE-NEXT: movaps %xmm13, %xmm9
+; SSE-NEXT: unpcklps {{.*#+}} xmm9 = xmm9[0],xmm6[0],xmm9[1],xmm6[1]
+; SSE-NEXT: movaps %xmm9, %xmm14
; SSE-NEXT: movaps %xmm9, %xmm12
-; SSE-NEXT: movlhps {{.*#+}} xmm13 = xmm13[0],xmm6[0]
-; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movlhps {{.*#+}} xmm14 = xmm14[0],xmm8[0]
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm3, %xmm9
; SSE-NEXT: unpcklps {{.*#+}} xmm9 = xmm9[0],xmm11[0],xmm9[1],xmm11[1]
-; SSE-NEXT: movaps %xmm4, %xmm13
-; SSE-NEXT: unpcklps {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm6[1]
+; SSE-NEXT: movaps %xmm4, %xmm14
+; SSE-NEXT: unpcklps {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm8[1]
; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm13, %xmm0
+; SSE-NEXT: movaps %xmm14, %xmm0
; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm9[0]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm9[1]
-; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm7, %xmm6
-; SSE-NEXT: unpcklps {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm9[1]
+; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm7, %xmm8
+; SSE-NEXT: unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
; SSE-NEXT: movaps 256(%rdi), %xmm9
; SSE-NEXT: movaps %xmm9, %xmm0
; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movaps %xmm1, %xmm4
; SSE-NEXT: movaps %xmm0, %xmm1
-; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm6[0]
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm8[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm6[1]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm8[1]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 96(%rdi), %xmm6
+; SSE-NEXT: movaps 96(%rdi), %xmm8
; SSE-NEXT: movaps 64(%rdi), %xmm12
-; SSE-NEXT: movaps %xmm12, %xmm13
-; SSE-NEXT: unpcklps {{.*#+}} xmm13 = xmm13[0],xmm6[0],xmm13[1],xmm6[1]
+; SSE-NEXT: movaps %xmm12, %xmm14
+; SSE-NEXT: unpcklps {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1]
; SSE-NEXT: movaps (%rdi), %xmm2
-; SSE-NEXT: movaps 32(%rdi), %xmm14
+; SSE-NEXT: movaps 32(%rdi), %xmm15
; SSE-NEXT: movaps %xmm2, %xmm1
-; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
+; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
; SSE-NEXT: movaps %xmm1, %xmm0
-; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm13[0]
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm14[0]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm13[1]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm14[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhps {{.*#+}} xmm10 = xmm10[2],xmm5[2],xmm10[3],xmm5[3]
-; SSE-NEXT: unpckhps {{.*#+}} xmm15 = xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm13 = xmm13[2],xmm6[2],xmm13[3],xmm6[3]
; SSE-NEXT: unpckhps {{.*#+}} xmm3 = xmm3[2],xmm11[2],xmm3[3],xmm11[3]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
@@ -965,13 +965,13 @@
; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT: # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
; SSE-NEXT: unpckhps {{.*#+}} xmm9 = xmm9[2],xmm4[2],xmm9[3],xmm4[3]
-; SSE-NEXT: unpckhps {{.*#+}} xmm12 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
-; SSE-NEXT: unpckhps {{.*#+}} xmm2 = xmm2[2],xmm14[2],xmm2[3],xmm14[3]
-; SSE-NEXT: movaps %xmm15, %xmm0
+; SSE-NEXT: unpckhps {{.*#+}} xmm12 = xmm12[2],xmm8[2],xmm12[3],xmm8[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm2 = xmm2[2],xmm15[2],xmm2[3],xmm15[3]
+; SSE-NEXT: movaps %xmm13, %xmm0
; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm10[0]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm10[1]
-; SSE-NEXT: movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm10[1]
+; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm11, %xmm0
; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm3[0]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -989,8 +989,8 @@
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 240(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 208(%rdi), %xmm15
-; SSE-NEXT: movaps %xmm15, %xmm0
+; SSE-NEXT: movaps 208(%rdi), %xmm12
+; SSE-NEXT: movaps %xmm12, %xmm0
; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movaps 176(%rdi), %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1002,36 +1002,36 @@
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 496(%rdi), %xmm0
+; SSE-NEXT: movaps 368(%rdi), %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 464(%rdi), %xmm3
+; SSE-NEXT: movaps 336(%rdi), %xmm3
; SSE-NEXT: movaps %xmm3, %xmm1
; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT: movaps 432(%rdi), %xmm0
+; SSE-NEXT: movaps 304(%rdi), %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 400(%rdi), %xmm10
+; SSE-NEXT: movaps 272(%rdi), %xmm10
; SSE-NEXT: movaps %xmm10, %xmm8
; SSE-NEXT: unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; SSE-NEXT: movaps %xmm8, %xmm0
; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm1[1]
-; SSE-NEXT: movaps 368(%rdi), %xmm12
-; SSE-NEXT: movaps 336(%rdi), %xmm2
+; SSE-NEXT: movaps 496(%rdi), %xmm14
+; SSE-NEXT: movaps 464(%rdi), %xmm2
; SSE-NEXT: movaps %xmm2, %xmm1
-; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
-; SSE-NEXT: movaps 304(%rdi), %xmm11
-; SSE-NEXT: movaps 272(%rdi), %xmm13
-; SSE-NEXT: movaps %xmm13, %xmm7
-; SSE-NEXT: unpcklps {{.*#+}} xmm7 = xmm7[0],xmm11[0],xmm7[1],xmm11[1]
+; SSE-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
+; SSE-NEXT: movaps 432(%rdi), %xmm13
+; SSE-NEXT: movaps 400(%rdi), %xmm15
+; SSE-NEXT: movaps %xmm15, %xmm7
+; SSE-NEXT: unpcklps {{.*#+}} xmm7 = xmm7[0],xmm13[0],xmm7[1],xmm13[1]
; SSE-NEXT: movaps %xmm7, %xmm0
; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm1[1]
-; SSE-NEXT: movaps 112(%rdi), %xmm14
+; SSE-NEXT: movaps 112(%rdi), %xmm11
; SSE-NEXT: movaps 80(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm0
-; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
; SSE-NEXT: movaps 16(%rdi), %xmm6
; SSE-NEXT: movaps 48(%rdi), %xmm9
; SSE-NEXT: movaps %xmm6, %xmm5
@@ -1040,31 +1040,31 @@
; SSE-NEXT: movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
; SSE-NEXT: movaps %xmm4, (%rsp) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm0[1]
-; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT: # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
-; SSE-NEXT: unpckhps {{.*#+}} xmm2 = xmm2[2],xmm12[2],xmm2[3],xmm12[3]
-; SSE-NEXT: unpckhps {{.*#+}} xmm13 = xmm13[2],xmm11[2],xmm13[3],xmm11[3]
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT: # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
; SSE-NEXT: unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
; SSE-NEXT: # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
-; SSE-NEXT: unpckhps {{.*#+}} xmm1 = xmm1[2],xmm14[2],xmm1[3],xmm14[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm2 = xmm2[2],xmm14[2],xmm2[3],xmm14[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm15 = xmm15[2],xmm13[2],xmm15[3],xmm13[3]
+; SSE-NEXT: unpckhps {{.*#+}} xmm1 = xmm1[2],xmm11[2],xmm1[3],xmm11[3]
; SSE-NEXT: unpckhps {{.*#+}} xmm6 = xmm6[2],xmm9[2],xmm6[3],xmm9[3]
-; SSE-NEXT: movaps %xmm4, %xmm0
-; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm15[0]
-; SSE-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm15[1]
-; SSE-NEXT: movaps %xmm4, %xmm9
-; SSE-NEXT: movaps %xmm13, %xmm4
-; SSE-NEXT: movlhps {{.*#+}} xmm4 = xmm4[0],xmm2[0]
-; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm2[1]
-; SSE-NEXT: movaps %xmm10, %xmm2
-; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE-NEXT: movaps %xmm0, %xmm4
+; SSE-NEXT: movlhps {{.*#+}} xmm4 = xmm4[0],xmm12[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm12[1]
+; SSE-NEXT: movaps %xmm0, %xmm9
+; SSE-NEXT: movaps %xmm10, %xmm0
+; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm3[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm3[1]
-; SSE-NEXT: movaps %xmm6, %xmm3
-; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT: movaps %xmm15, %xmm3
+; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm2[1]
+; SSE-NEXT: movaps %xmm6, %xmm2
+; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm1[1]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 32(%rsi)
@@ -1107,19 +1107,19 @@
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movaps %xmm7, 32(%rax)
-; SSE-NEXT: movaps %xmm8, 48(%rax)
-; SSE-NEXT: movaps %xmm5, (%rax)
+; SSE-NEXT: movaps %xmm7, 48(%rax)
+; SSE-NEXT: movaps %xmm8, 32(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%rax)
+; SSE-NEXT: movaps %xmm5, (%rax)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movaps %xmm2, 48(%rax)
-; SSE-NEXT: movaps %xmm4, 32(%rax)
-; SSE-NEXT: movaps %xmm0, 16(%rax)
-; SSE-NEXT: movaps %xmm3, (%rax)
+; SSE-NEXT: movaps %xmm3, 48(%rax)
+; SSE-NEXT: movaps %xmm0, 32(%rax)
+; SSE-NEXT: movaps %xmm4, 16(%rax)
+; SSE-NEXT: movaps %xmm2, (%rax)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movaps %xmm10, 48(%rax)
-; SSE-NEXT: movaps %xmm13, 32(%rax)
+; SSE-NEXT: movaps %xmm15, 48(%rax)
+; SSE-NEXT: movaps %xmm10, 32(%rax)
; SSE-NEXT: movaps %xmm9, 16(%rax)
; SSE-NEXT: movaps %xmm6, (%rax)
; SSE-NEXT: addq $296, %rsp # imm = 0x128
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll
index fb01a0a..9df756a 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-2.ll
@@ -154,30 +154,30 @@
; SSE-NEXT: movaps 16(%rdi), %xmm1
; SSE-NEXT: movaps 32(%rdi), %xmm2
; SSE-NEXT: movaps 48(%rdi), %xmm3
-; SSE-NEXT: movaps 80(%rdi), %xmm4
-; SSE-NEXT: movaps 64(%rdi), %xmm5
-; SSE-NEXT: movaps 112(%rdi), %xmm6
-; SSE-NEXT: movaps 96(%rdi), %xmm7
+; SSE-NEXT: movaps 112(%rdi), %xmm4
+; SSE-NEXT: movaps 96(%rdi), %xmm5
+; SSE-NEXT: movaps 80(%rdi), %xmm6
+; SSE-NEXT: movaps 64(%rdi), %xmm7
; SSE-NEXT: movaps %xmm7, %xmm8
; SSE-NEXT: movlhps {{.*#+}} xmm8 = xmm8[0],xmm6[0]
; SSE-NEXT: movaps %xmm5, %xmm9
; SSE-NEXT: movlhps {{.*#+}} xmm9 = xmm9[0],xmm4[0]
-; SSE-NEXT: movaps %xmm2, %xmm10
-; SSE-NEXT: movlhps {{.*#+}} xmm10 = xmm10[0],xmm3[0]
-; SSE-NEXT: movaps %xmm0, %xmm11
-; SSE-NEXT: movlhps {{.*#+}} xmm11 = xmm11[0],xmm1[0]
+; SSE-NEXT: movaps %xmm0, %xmm10
+; SSE-NEXT: movlhps {{.*#+}} xmm10 = xmm10[0],xmm1[0]
+; SSE-NEXT: movaps %xmm2, %xmm11
+; SSE-NEXT: movlhps {{.*#+}} xmm11 = xmm11[0],xmm3[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm6[1]
; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm4[1]
-; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT: movaps %xmm9, 32(%rsi)
-; SSE-NEXT: movaps %xmm8, 48(%rsi)
-; SSE-NEXT: movaps %xmm11, (%rsi)
-; SSE-NEXT: movaps %xmm10, 16(%rsi)
-; SSE-NEXT: movaps %xmm5, 32(%rdx)
-; SSE-NEXT: movaps %xmm7, 48(%rdx)
-; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
+; SSE-NEXT: movaps %xmm11, 16(%rsi)
+; SSE-NEXT: movaps %xmm10, (%rsi)
+; SSE-NEXT: movaps %xmm9, 48(%rsi)
+; SSE-NEXT: movaps %xmm8, 32(%rsi)
; SSE-NEXT: movaps %xmm2, 16(%rdx)
+; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: movaps %xmm5, 48(%rdx)
+; SSE-NEXT: movaps %xmm7, 32(%rdx)
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i64_stride2_vf8:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
index 88e8d2c..3fca052 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-3.ll
@@ -358,104 +358,104 @@
; SSE-LABEL: load_i64_stride3_vf16:
; SSE: # %bb.0:
; SSE-NEXT: subq $24, %rsp
-; SSE-NEXT: movapd 128(%rdi), %xmm0
-; SSE-NEXT: movapd 176(%rdi), %xmm1
-; SSE-NEXT: movapd 224(%rdi), %xmm4
-; SSE-NEXT: movapd 272(%rdi), %xmm3
-; SSE-NEXT: movapd 80(%rdi), %xmm2
-; SSE-NEXT: movapd 96(%rdi), %xmm5
-; SSE-NEXT: movapd 112(%rdi), %xmm11
-; SSE-NEXT: movapd 144(%rdi), %xmm6
-; SSE-NEXT: movapd 160(%rdi), %xmm14
-; SSE-NEXT: movapd 192(%rdi), %xmm7
+; SSE-NEXT: movapd 272(%rdi), %xmm0
+; SSE-NEXT: movapd 224(%rdi), %xmm2
+; SSE-NEXT: movapd 368(%rdi), %xmm1
+; SSE-NEXT: movapd 320(%rdi), %xmm3
+; SSE-NEXT: movapd 128(%rdi), %xmm4
+; SSE-NEXT: movapd 240(%rdi), %xmm5
+; SSE-NEXT: movapd 256(%rdi), %xmm10
+; SSE-NEXT: movapd 192(%rdi), %xmm6
; SSE-NEXT: movapd 208(%rdi), %xmm12
-; SSE-NEXT: movapd 240(%rdi), %xmm10
-; SSE-NEXT: movapd 256(%rdi), %xmm13
-; SSE-NEXT: movapd 48(%rdi), %xmm9
-; SSE-NEXT: movapd 64(%rdi), %xmm15
+; SSE-NEXT: movapd 336(%rdi), %xmm7
+; SSE-NEXT: movapd 352(%rdi), %xmm14
+; SSE-NEXT: movapd 288(%rdi), %xmm11
+; SSE-NEXT: movapd 304(%rdi), %xmm15
+; SSE-NEXT: movapd 96(%rdi), %xmm9
+; SSE-NEXT: movapd 112(%rdi), %xmm13
; SSE-NEXT: movapd %xmm15, %xmm8
-; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm9[0],xmm8[1]
+; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm11[0],xmm8[1]
; SSE-NEXT: movapd %xmm8, (%rsp) # 16-byte Spill
-; SSE-NEXT: shufpd {{.*#+}} xmm9 = xmm9[1],xmm2[0]
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm15[0],xmm2[1]
-; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufpd {{.*#+}} xmm11 = xmm11[1],xmm3[0]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm15[0],xmm3[1]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movapd %xmm14, %xmm15
-; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm6[0],xmm15[1]
-; SSE-NEXT: shufpd {{.*#+}} xmm6 = xmm6[1],xmm1[0]
-; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm7[0],xmm15[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm7 = xmm7[1],xmm1[0]
+; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd %xmm11, %xmm14
-; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm5[0],xmm14[1]
+; SSE-NEXT: movapd %xmm12, %xmm14
+; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm6[0],xmm14[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm6 = xmm6[1],xmm2[0]
+; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm12[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movapd %xmm10, %xmm12
+; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm5[0],xmm12[1]
; SSE-NEXT: shufpd {{.*#+}} xmm5 = xmm5[1],xmm0[0]
; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm11[0],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm10[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd %xmm13, %xmm11
-; SSE-NEXT: movsd {{.*#+}} xmm11 = xmm10[0],xmm11[1]
-; SSE-NEXT: shufpd {{.*#+}} xmm10 = xmm10[1],xmm3[0]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm13[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd %xmm12, %xmm13
-; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm7[0],xmm13[1]
-; SSE-NEXT: shufpd {{.*#+}} xmm7 = xmm7[1],xmm4[0]
-; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm12[0],xmm4[1]
+; SSE-NEXT: movapd %xmm13, %xmm10
+; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm9[0],xmm10[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm9 = xmm9[1],xmm4[0]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm13[0],xmm4[1]
; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd 336(%rdi), %xmm12
-; SSE-NEXT: movapd 352(%rdi), %xmm2
-; SSE-NEXT: movapd %xmm2, %xmm7
-; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm12[0],xmm7[1]
-; SSE-NEXT: movapd 368(%rdi), %xmm4
-; SSE-NEXT: shufpd {{.*#+}} xmm12 = xmm12[1],xmm4[0]
-; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
-; SSE-NEXT: movapd 288(%rdi), %xmm2
-; SSE-NEXT: movapd 304(%rdi), %xmm5
-; SSE-NEXT: movapd %xmm5, %xmm3
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
-; SSE-NEXT: movapd 320(%rdi), %xmm0
-; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1],xmm0[0]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
-; SSE-NEXT: movapd (%rdi), %xmm5
-; SSE-NEXT: movapd 16(%rdi), %xmm8
-; SSE-NEXT: movapd %xmm8, %xmm6
-; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm5[0],xmm6[1]
-; SSE-NEXT: movapd 32(%rdi), %xmm1
-; SSE-NEXT: shufpd {{.*#+}} xmm5 = xmm5[1],xmm1[0]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
-; SSE-NEXT: movapd %xmm3, 96(%rsi)
-; SSE-NEXT: movapd %xmm14, 32(%rsi)
-; SSE-NEXT: movapd %xmm7, 112(%rsi)
-; SSE-NEXT: movapd %xmm15, 48(%rsi)
-; SSE-NEXT: movapd %xmm13, 64(%rsi)
-; SSE-NEXT: movapd %xmm6, (%rsi)
-; SSE-NEXT: movapd %xmm11, 80(%rsi)
+; SSE-NEXT: movapd 144(%rdi), %xmm13
+; SSE-NEXT: movapd 160(%rdi), %xmm1
+; SSE-NEXT: movapd %xmm1, %xmm8
+; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm13[0],xmm8[1]
+; SSE-NEXT: movapd 176(%rdi), %xmm6
+; SSE-NEXT: shufpd {{.*#+}} xmm13 = xmm13[1],xmm6[0]
+; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1]
+; SSE-NEXT: movapd 48(%rdi), %xmm1
+; SSE-NEXT: movapd 64(%rdi), %xmm4
+; SSE-NEXT: movapd %xmm4, %xmm3
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSE-NEXT: movapd 80(%rdi), %xmm2
+; SSE-NEXT: shufpd {{.*#+}} xmm1 = xmm1[1],xmm2[0]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm4[0],xmm2[1]
+; SSE-NEXT: movapd (%rdi), %xmm4
+; SSE-NEXT: movapd 16(%rdi), %xmm7
+; SSE-NEXT: movapd %xmm7, %xmm5
+; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; SSE-NEXT: movapd 32(%rdi), %xmm0
+; SSE-NEXT: shufpd {{.*#+}} xmm4 = xmm4[1],xmm0[0]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1]
+; SSE-NEXT: movapd %xmm12, 80(%rsi)
+; SSE-NEXT: movapd %xmm3, 16(%rsi)
+; SSE-NEXT: movapd %xmm14, 64(%rsi)
+; SSE-NEXT: movapd %xmm5, (%rsi)
+; SSE-NEXT: movapd %xmm15, 112(%rsi)
+; SSE-NEXT: movapd %xmm8, 48(%rsi)
; SSE-NEXT: movaps (%rsp), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 16(%rsi)
-; SSE-NEXT: movapd %xmm2, 96(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 32(%rdx)
-; SSE-NEXT: movapd %xmm12, 112(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 64(%rdx)
-; SSE-NEXT: movapd %xmm5, (%rdx)
-; SSE-NEXT: movapd %xmm10, 80(%rdx)
-; SSE-NEXT: movapd %xmm9, 16(%rdx)
-; SSE-NEXT: movapd %xmm0, 96(%rcx)
-; SSE-NEXT: movapd %xmm4, 112(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 64(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 80(%rcx)
+; SSE-NEXT: movaps %xmm3, 96(%rsi)
+; SSE-NEXT: movapd %xmm10, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 80(%rdx)
+; SSE-NEXT: movapd %xmm1, 16(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 64(%rdx)
+; SSE-NEXT: movapd %xmm4, (%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 112(%rdx)
+; SSE-NEXT: movapd %xmm13, 48(%rdx)
+; SSE-NEXT: movapd %xmm11, 96(%rdx)
+; SSE-NEXT: movapd %xmm9, 32(%rdx)
+; SSE-NEXT: movapd %xmm2, 16(%rcx)
+; SSE-NEXT: movapd %xmm0, (%rcx)
+; SSE-NEXT: movapd %xmm6, 48(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 32(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rcx)
-; SSE-NEXT: movapd %xmm1, (%rcx)
+; SSE-NEXT: movaps %xmm0, 80(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%rcx)
+; SSE-NEXT: movaps %xmm0, 64(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 112(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 96(%rcx)
; SSE-NEXT: addq $24, %rsp
; SSE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll
index 108aa91..480e382 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-4.ll
@@ -1328,7 +1328,7 @@
;
; AVX1-ONLY-LABEL: load_i64_stride4_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1112, %rsp # imm = 0x458
+; AVX1-ONLY-NEXT: subq $1096, %rsp # imm = 0x448
; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -1338,65 +1338,41 @@
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm11[0],xmm10[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm9[0],xmm8[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm9[1],xmm8[1]
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm6[0],xmm4[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm5[0]
; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm7[1],xmm6[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm6[1],xmm4[1]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm11[0],xmm10[0]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm9[0],xmm8[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
+; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm5[1]
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm9[1],xmm8[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
@@ -1425,42 +1401,30 @@
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 432(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 688(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 656(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm1
@@ -1468,93 +1432,127 @@
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX1-ONLY-NEXT: vmovaps 944(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 912(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm15[0],xmm1[0]
+; AVX1-ONLY-NEXT: vmovaps 912(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm13
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm13[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
-; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm10[0]
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps 432(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm15[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
-; AVX1-ONLY-NEXT: vmovaps 560(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm6[0]
+; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps 688(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 656(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX1-ONLY-NEXT: vmovaps 816(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 784(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps 816(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 784(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm9[0],ymm10[0],ymm9[2],ymm10[2]
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm7[0],xmm8[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm14
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm14[0],ymm6[0],ymm14[2],ymm6[2]
+; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm11 = xmm4[0],xmm5[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm3[0],ymm11[2],ymm3[2]
+; AVX1-ONLY-NEXT: vmovaps 560(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = mem[0,1],xmm14[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = xmm13[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = mem[0,1],xmm14[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm12[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm14[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = mem[0,1],xmm12[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm12[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = xmm15[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm14[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = xmm13[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm13[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm9[1],xmm10[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm6[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = xmm12[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm9[1],ymm10[1],ymm9[3],ymm10[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],xmm8[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm14[1],ymm6[1],ymm14[3],ymm6[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],xmm5[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm3[1],ymm11[3],ymm3[3]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm1, 208(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm1, 192(%rsi)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm1, 128(%rsi)
@@ -1585,6 +1583,14 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm1, 112(%rsi)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm1, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm1, 208(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm1, 16(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm1, 64(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm1, 80(%rdx)
@@ -1593,13 +1599,13 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm1, 144(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rdx)
+; AVX1-ONLY-NEXT: vmovaps %xmm1, 224(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm1, 16(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %xmm1, 240(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm1, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %xmm1, 160(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm1, 208(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %xmm1, 176(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm1, 96(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -1608,16 +1614,6 @@
; AVX1-ONLY-NEXT: vmovaps %xmm1, 32(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm1, 48(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm1, 160(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm1, 176(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm1, 224(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm1, 240(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 192(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 128(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -1625,25 +1621,25 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 192(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 160(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm0, 192(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm9, 64(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm14, (%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm15, 224(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 160(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 128(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm7, (%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm12, 224(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm13, 160(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm15, 96(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
-; AVX1-ONLY-NEXT: addq $1112, %rsp # imm = 0x458
+; AVX1-ONLY-NEXT: addq $1096, %rsp # imm = 0x448
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -1659,59 +1655,41 @@
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm5
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %xmm9
-; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %xmm10
-; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %xmm11
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm11[0],xmm10[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm9[0],xmm8[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm9[1],xmm8[1]
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm6
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm6[0],xmm4[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm5[0]
; AVX2-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm7[1],xmm6[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
+; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm9
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm11
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm6[1],xmm4[1]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm11[0],xmm10[0]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm11[1],xmm10[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm9[0],xmm8[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
+; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm5[1]
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm9[1],xmm8[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
@@ -1740,12 +1718,42 @@
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm8[2,3],ymm3[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
@@ -1754,9 +1762,9 @@
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm8[2,3],ymm3[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1766,202 +1774,186 @@
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm8[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm7[2,3]
+; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm8[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm13[0],ymm14[0],ymm13[2],ymm14[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm9[0],ymm10[0],ymm9[2],ymm10[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm6[0],ymm8[0],ymm6[2],ymm8[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm15[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm15 = ymm7[1],mem[1],ymm7[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm7[2,3],ymm15[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm15[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm15[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm15[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm13[1],ymm14[1],ymm13[3],ymm14[3]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm14
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm9[0],ymm7[2],ymm9[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm14[0],ymm10[0],ymm14[2],ymm10[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm15
+; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm15[0],ymm6[0],ymm15[2],ymm6[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm8[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm8[0],ymm2[0],ymm8[2],ymm2[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm13[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm13[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm9[1],ymm10[1],ymm9[3],ymm10[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm9[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm6[1],ymm8[1],ymm6[3],ymm8[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm5[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm13[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm13[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm11 = ymm11[1],mem[1],ymm11[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm11[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm12[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm7[1],ymm9[1],ymm7[3],ymm9[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm14[1],ymm10[1],ymm14[3],ymm10[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm9[2,3],ymm7[2,3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm15[1],ymm6[1],ymm15[3],ymm6[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm4[2,3],ymm3[2,3]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 208(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 192(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 128(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 64(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, (%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 144(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 80(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 16(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 240(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 224(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 160(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 96(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 48(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 176(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 112(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 64(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 80(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 128(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 144(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, (%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 16(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 192(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 208(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 96(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 112(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 48(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 160(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 176(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 224(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 240(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 192(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 128(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 64(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 160(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm8[1],ymm2[1],ymm8[3],ymm2[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = ymm2[1],mem[1],ymm2[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 208(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 192(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 128(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 64(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 144(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 80(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 16(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 240(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 224(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 160(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 96(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 48(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 176(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 112(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 192(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 208(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 16(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 64(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 80(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 128(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 144(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 224(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 240(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 160(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 176(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 96(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 112(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm2, 48(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 128(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 64(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 192(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 160(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 96(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 224(%rcx)
; AVX2-ONLY-NEXT: vmovaps %ymm0, 192(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm4, 128(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm9, 64(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm13, (%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm15, 224(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm0, 160(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm0, 96(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 128(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm6, 64(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm7, (%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 224(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm11, 160(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm13, 96(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
; AVX2-ONLY-NEXT: addq $1224, %rsp # imm = 0x4C8
@@ -2776,75 +2768,147 @@
;
; AVX1-ONLY-LABEL: load_i64_stride4_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $2680, %rsp # imm = 0xA78
-; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm1
+; AVX1-ONLY-NEXT: subq $2632, %rsp # imm = 0xA48
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1760(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm5
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1440(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 1696(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 1952(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 1920(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps 2016(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vmovaps 1984(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm11[0],xmm10[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm9[0],xmm8[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm9[1],xmm8[1]
+; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm11[0],xmm10[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm6
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm8[0],xmm7[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm6[0],xmm9[0]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm8[1],xmm7[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm6[1],xmm9[1]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm4
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm6[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm8[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm6[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm8[1]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm2
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1312(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1568(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1888(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1856(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1824(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1792(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1184(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1440(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1760(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1696(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 2016(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1984(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1952(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1920(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
@@ -2873,182 +2937,122 @@
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1888(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1856(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1824(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1792(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1568(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1312(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 432(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps 560(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 688(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 816(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 656(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 784(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 944(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 1072(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 912(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1040(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 1200(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1328(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1168(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 1296(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 1456(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1584(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1424(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 1552(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1760(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 1888(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 1856(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 1712(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1840(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1680(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 1808(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -3065,93 +3069,97 @@
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm6[0],ymm10[0],ymm6[2],ymm10[2]
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm3[0],xmm7[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1760(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm11[0],ymm14[0],ymm11[2],ymm14[2]
-; AVX1-ONLY-NEXT: vmovaps 304(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm4[0],xmm5[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 560(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm8[0],xmm9[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm12[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 816(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 784(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm12[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1712(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1680(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 1504(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 1072(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1040(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 1456(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1424(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 1328(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1296(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 1200(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1168(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1632(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 1584(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1552(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vmovaps 944(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vmovaps 912(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm12[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm9[0],ymm10[0],ymm9[2],ymm10[2]
+; AVX1-ONLY-NEXT: vmovaps 688(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 656(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm7[0],xmm8[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm14
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm14[0],ymm6[0],ymm14[2],ymm6[2]
+; AVX1-ONLY-NEXT: vmovaps 432(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm4[0],xmm5[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1888(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1856(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm13[0],ymm2[0],ymm13[2],ymm2[2]
+; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 144(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm15 = xmm0[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm15[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm13[1],ymm2[1],ymm13[3],ymm2[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 1840(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1808(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm0[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm6[1],ymm10[1],ymm6[3],ymm10[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm7[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm14[1],ymm6[1],ymm14[3],ymm6[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm5[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
@@ -3160,8 +3168,8 @@
; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm14[1],ymm11[3],ymm14[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm5[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm9[1],ymm10[1],ymm9[3],ymm10[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm8[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
@@ -3169,45 +3177,33 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm15[1],mem[1],ymm15[3],mem[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm8[1],xmm9[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm11[1],xmm12[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vunpckhpd (%rsp), %xmm12, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm12[1],mem[1]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
@@ -3221,335 +3217,263 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = ymm2[1],mem[1],ymm2[3],mem[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1],xmm3[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm6[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 464(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 448(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 256(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 384(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 320(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 192(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 128(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, (%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 272(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 400(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 336(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 208(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 144(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 80(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 16(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 496(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 480(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 416(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 352(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 288(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 224(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 160(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 48(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 432(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 368(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 304(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 240(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 176(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 112(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 128(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 144(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 256(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 272(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 80(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, (%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 16(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 192(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 208(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 320(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 336(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 384(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 400(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 448(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 464(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 112(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 48(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 160(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 176(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 224(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 240(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 288(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 304(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 352(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 368(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 416(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 432(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 480(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm6, 496(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 448(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 384(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 320(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 256(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 192(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 128(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 480(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 416(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 352(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 288(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 224(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 160(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 496(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 480(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 432(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 416(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 368(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 352(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 288(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 224(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 160(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 48(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 304(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 240(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 176(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 112(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 464(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 448(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 256(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 384(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 320(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 192(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 128(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, (%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 272(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 400(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 336(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 208(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 144(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 80(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 16(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 224(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 240(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 160(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 176(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 112(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 48(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 480(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 496(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 416(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 432(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 352(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 368(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 288(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 304(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 448(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 464(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 384(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 400(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 320(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 336(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 208(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, (%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 16(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 80(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 128(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 144(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 256(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 272(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 288(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 352(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 416(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 480(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 448(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 384(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 320(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 256(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm3, 480(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 448(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 448(%r8)
; AVX1-ONLY-NEXT: vmovaps %ymm1, 416(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 384(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 352(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm4, 320(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm15, 288(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm14, 256(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm12, 224(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm11, 192(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm9, 160(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 128(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 96(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 64(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 384(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 352(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm6, 320(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm11, 288(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm10, 256(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm9, 224(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 160(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm13, (%r8)
-; AVX1-ONLY-NEXT: addq $2680, %rsp # imm = 0xA78
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 128(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm14, 96(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm13, 32(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm15, (%r8)
+; AVX1-ONLY-NEXT: addq $2632, %rsp # imm = 0xA48
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-ONLY-LABEL: load_i64_stride4_vf64:
; AVX2-ONLY: # %bb.0:
; AVX2-ONLY-NEXT: subq $3016, %rsp # imm = 0xBC8
-; AVX2-ONLY-NEXT: vmovaps 1248(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1504(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %xmm3
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1760(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %xmm5
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1440(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovaps 1696(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovaps 1952(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %xmm9
-; AVX2-ONLY-NEXT: vmovaps 2016(%rdi), %xmm10
-; AVX2-ONLY-NEXT: vmovaps 1984(%rdi), %xmm11
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm11[0],xmm10[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm9[0],xmm8[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm9[1],xmm8[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vmovaps 1120(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm8[0],xmm7[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %xmm6
+; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %xmm9
+; AVX2-ONLY-NEXT: vmovaps 1056(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %xmm11
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm8[1],xmm7[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm11[0],xmm10[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm11[1],xmm10[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %xmm7
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm8[0],xmm7[0]
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm7[0],xmm9[0]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm8[1],xmm7[1]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm7[1],xmm9[1]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm4
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm6[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm6[1]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1184(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %xmm2
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1888(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1856(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1824(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1792(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1632(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1568(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
@@ -3566,56 +3490,122 @@
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 1632(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 1568(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1888(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1856(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1824(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1792(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1248(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1184(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1504(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1440(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1760(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1696(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2016(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1984(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1952(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1120(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1056(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
@@ -3626,73 +3616,109 @@
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1184(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 1056(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1248(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 1120(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1440(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 1312(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1504(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 1376(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1568(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1632(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1824(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1792(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1888(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1856(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1952(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2016(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1984(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
@@ -3707,389 +3733,343 @@
; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1952(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 1440(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2016(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 1504(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1984(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 1184(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 1248(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm15[0],ymm10[2],ymm15[2]
+; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm11[0],ymm10[2],ymm11[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1056(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1120(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm9[0],ymm7[2],ymm9[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %ymm13
+; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm13[0],ymm9[0],ymm13[2],ymm9[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1312(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1376(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm4[0],ymm2[2],ymm4[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm12[0],ymm5[0],ymm12[2],ymm5[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm3[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1568(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovaps 1632(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm11[0],ymm14[0],ymm11[2],ymm14[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1824(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovaps 1792(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vmovaps 1888(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovaps 1856(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm13[0],ymm8[0],ymm13[2],ymm8[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm14[0],ymm3[0],ymm14[2],ymm3[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm15[2,3],ymm6[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm6 = ymm6[1],mem[1],ymm6[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm15[2,3],ymm6[2,3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm14[1],ymm3[1],ymm14[3],ymm3[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm4[1],ymm2[3],ymm4[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm12[1],ymm5[1],ymm12[3],ymm5[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm13[1],ymm9[1],ymm13[3],ymm9[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm11[1],ymm10[3],ymm11[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm15[1],ymm10[3],ymm15[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm10 = ymm10[1],mem[1],ymm10[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm10[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm10 = ymm10[1],mem[1],ymm10[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm10[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm9[1],ymm7[3],ymm9[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm7[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm9 = ymm9[1],mem[1],ymm9[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm9[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm6[1],ymm5[3],ymm6[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm5[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm5[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm11[1],ymm14[1],ymm11[3],ymm14[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm3[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm3[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm13[1],ymm8[1],ymm13[3],ymm8[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm2 = ymm2[1],mem[1],ymm2[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 464(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 448(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 256(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 384(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 320(%rsi)
-; AVX2-ONLY-NEXT: vmovaps (%rsp), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 192(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 128(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 64(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 272(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 400(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 336(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 208(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 144(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 80(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 16(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 496(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 480(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 416(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 352(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 288(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 224(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 160(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 96(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 48(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 432(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 368(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 304(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 240(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 176(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 112(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 128(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 144(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 256(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 272(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 64(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 80(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 16(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 192(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 208(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 320(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 336(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 384(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 400(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 448(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 464(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 96(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 112(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 48(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 160(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 176(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 224(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 240(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 288(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 304(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 352(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 368(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 416(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 432(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 480(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm2, 496(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 448(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 384(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 320(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 256(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 192(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 128(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 64(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 480(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 416(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 352(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 288(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 224(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 160(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 96(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = ymm4[1],mem[1],ymm4[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm4[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 496(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 480(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 432(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 416(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 368(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 352(%rsi)
+; AVX2-ONLY-NEXT: vmovaps (%rsp), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 288(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 224(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 160(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 96(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 48(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 304(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 240(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 176(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 112(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 464(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 448(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 256(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 384(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 320(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 192(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 128(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 64(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, (%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 272(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 400(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 336(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 208(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 144(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 80(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 16(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 224(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 240(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 160(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 176(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 96(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 112(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 48(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 480(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 496(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 416(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 432(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 352(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 368(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 288(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 304(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 448(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 464(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 384(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 400(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 320(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 336(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 192(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 208(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, (%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 16(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 64(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 80(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 128(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 144(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 256(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm4, 272(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 96(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 160(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 224(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 288(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 352(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 416(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 480(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 448(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 384(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 320(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 256(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 192(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 128(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rcx)
; AVX2-ONLY-NEXT: vmovaps %ymm0, 480(%r8)
; AVX2-ONLY-NEXT: vmovaps %ymm1, 448(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm4, 416(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 384(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm12, 352(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 320(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm9, 288(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm7, 256(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm10, 224(%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm15, 192(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 416(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 384(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm0, 160(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm0, 128(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm0, 96(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm0, 64(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm0, (%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm0, 352(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm13, 320(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm11, 288(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm10, 256(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm9, 224(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm8, 192(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm7, 160(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm12, 128(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm15, 64(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm14, 32(%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm6, (%r8)
; AVX2-ONLY-NEXT: addq $3016, %rsp # imm = 0xBC8
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-5.ll
index a1c9ad9..d096726 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-5.ll
@@ -356,80 +356,80 @@
define void @load_i64_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4) nounwind {
; SSE-LABEL: load_i64_stride5_vf8:
; SSE: # %bb.0:
-; SSE-NEXT: movapd 304(%rdi), %xmm2
-; SSE-NEXT: movapd 64(%rdi), %xmm1
-; SSE-NEXT: movapd 224(%rdi), %xmm0
-; SSE-NEXT: movapd 256(%rdi), %xmm4
-; SSE-NEXT: movapd 176(%rdi), %xmm3
-; SSE-NEXT: movapd 288(%rdi), %xmm6
-; SSE-NEXT: movapd 208(%rdi), %xmm5
+; SSE-NEXT: movapd 224(%rdi), %xmm2
+; SSE-NEXT: movapd 304(%rdi), %xmm1
+; SSE-NEXT: movapd 64(%rdi), %xmm0
+; SSE-NEXT: movapd 176(%rdi), %xmm4
+; SSE-NEXT: movapd 256(%rdi), %xmm3
+; SSE-NEXT: movapd 208(%rdi), %xmm6
+; SSE-NEXT: movapd 288(%rdi), %xmm7
; SSE-NEXT: movapd (%rdi), %xmm8
-; SSE-NEXT: movapd 16(%rdi), %xmm7
-; SSE-NEXT: movapd 32(%rdi), %xmm13
+; SSE-NEXT: movapd 16(%rdi), %xmm5
+; SSE-NEXT: movapd 32(%rdi), %xmm14
; SSE-NEXT: movapd 48(%rdi), %xmm9
-; SSE-NEXT: movapd 240(%rdi), %xmm10
-; SSE-NEXT: movapd 272(%rdi), %xmm14
-; SSE-NEXT: movapd 160(%rdi), %xmm11
-; SSE-NEXT: movapd 192(%rdi), %xmm15
-; SSE-NEXT: movapd %xmm15, %xmm12
-; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm11[0],xmm12[1]
-; SSE-NEXT: shufpd {{.*#+}} xmm11 = xmm11[1],xmm5[0]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm3[0],xmm5[1]
-; SSE-NEXT: shufpd {{.*#+}} xmm3 = xmm3[1],xmm0[0]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm15[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd %xmm13, %xmm15
-; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm8[0],xmm15[1]
-; SSE-NEXT: shufpd {{.*#+}} xmm8 = xmm8[1],xmm9[0]
-; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm7[0],xmm9[1]
-; SSE-NEXT: shufpd {{.*#+}} xmm7 = xmm7[1],xmm1[0]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm13[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movapd 160(%rdi), %xmm10
+; SSE-NEXT: movapd 192(%rdi), %xmm12
+; SSE-NEXT: movapd 240(%rdi), %xmm11
+; SSE-NEXT: movapd 272(%rdi), %xmm15
; SSE-NEXT: movapd %xmm14, %xmm13
-; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm10[0],xmm13[1]
+; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm8[0],xmm13[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm8 = xmm8[1],xmm9[0]
+; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm5[0],xmm9[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm5 = xmm5[1],xmm0[0]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movapd %xmm15, %xmm14
+; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm11[0],xmm14[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm11 = xmm11[1],xmm7[0]
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm3[0],xmm7[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm3 = xmm3[1],xmm1[0]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm15[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movapd %xmm12, %xmm15
+; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm10[0],xmm15[1]
; SSE-NEXT: shufpd {{.*#+}} xmm10 = xmm10[1],xmm6[0]
; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
; SSE-NEXT: shufpd {{.*#+}} xmm4 = xmm4[1],xmm2[0]
; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm14[0],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm12[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd 80(%rdi), %xmm14
+; SSE-NEXT: movapd 80(%rdi), %xmm12
; SSE-NEXT: movapd 112(%rdi), %xmm4
; SSE-NEXT: movapd %xmm4, %xmm3
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm14[0],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm12[0],xmm3[1]
; SSE-NEXT: movapd 128(%rdi), %xmm0
-; SSE-NEXT: shufpd {{.*#+}} xmm14 = xmm14[1],xmm0[0]
+; SSE-NEXT: shufpd {{.*#+}} xmm12 = xmm12[1],xmm0[0]
; SSE-NEXT: movapd 96(%rdi), %xmm1
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE-NEXT: movapd 144(%rdi), %xmm2
; SSE-NEXT: shufpd {{.*#+}} xmm1 = xmm1[1],xmm2[0]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm4[0],xmm2[1]
; SSE-NEXT: movapd %xmm3, 16(%rsi)
-; SSE-NEXT: movapd %xmm13, 48(%rsi)
-; SSE-NEXT: movapd %xmm15, (%rsi)
-; SSE-NEXT: movapd %xmm12, 32(%rsi)
-; SSE-NEXT: movapd %xmm14, 16(%rdx)
-; SSE-NEXT: movapd %xmm10, 48(%rdx)
+; SSE-NEXT: movapd %xmm15, 32(%rsi)
+; SSE-NEXT: movapd %xmm14, 48(%rsi)
+; SSE-NEXT: movapd %xmm13, (%rsi)
+; SSE-NEXT: movapd %xmm12, 16(%rdx)
+; SSE-NEXT: movapd %xmm10, 32(%rdx)
+; SSE-NEXT: movapd %xmm11, 48(%rdx)
; SSE-NEXT: movapd %xmm8, (%rdx)
-; SSE-NEXT: movapd %xmm11, 32(%rdx)
; SSE-NEXT: movapd %xmm0, 16(%rcx)
-; SSE-NEXT: movapd %xmm6, 48(%rcx)
+; SSE-NEXT: movapd %xmm6, 32(%rcx)
+; SSE-NEXT: movapd %xmm7, 48(%rcx)
; SSE-NEXT: movapd %xmm9, (%rcx)
-; SSE-NEXT: movapd %xmm5, 32(%rcx)
; SSE-NEXT: movapd %xmm1, 16(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%r8)
-; SSE-NEXT: movapd %xmm7, (%r8)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 32(%r8)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 48(%r8)
+; SSE-NEXT: movapd %xmm5, (%r8)
; SSE-NEXT: movapd %xmm2, 16(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%r9)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 48(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%r9)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%r9)
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i64_stride5_vf8:
@@ -712,54 +712,54 @@
; SSE-LABEL: load_i64_stride5_vf16:
; SSE: # %bb.0:
; SSE-NEXT: subq $280, %rsp # imm = 0x118
-; SSE-NEXT: movapd 224(%rdi), %xmm1
-; SSE-NEXT: movapd 144(%rdi), %xmm2
-; SSE-NEXT: movapd 64(%rdi), %xmm3
+; SSE-NEXT: movapd 224(%rdi), %xmm0
+; SSE-NEXT: movapd 144(%rdi), %xmm1
+; SSE-NEXT: movapd 64(%rdi), %xmm2
+; SSE-NEXT: movapd 96(%rdi), %xmm3
; SSE-NEXT: movapd 176(%rdi), %xmm4
-; SSE-NEXT: movapd 96(%rdi), %xmm5
-; SSE-NEXT: movapd 208(%rdi), %xmm6
-; SSE-NEXT: movapd 128(%rdi), %xmm8
-; SSE-NEXT: movapd (%rdi), %xmm10
-; SSE-NEXT: movapd 16(%rdi), %xmm7
-; SSE-NEXT: movapd 32(%rdi), %xmm14
+; SSE-NEXT: movapd 128(%rdi), %xmm6
+; SSE-NEXT: movapd 208(%rdi), %xmm7
+; SSE-NEXT: movapd (%rdi), %xmm8
+; SSE-NEXT: movapd 16(%rdi), %xmm5
+; SSE-NEXT: movapd 32(%rdi), %xmm13
; SSE-NEXT: movapd 48(%rdi), %xmm9
+; SSE-NEXT: movapd 80(%rdi), %xmm10
+; SSE-NEXT: movapd 112(%rdi), %xmm14
; SSE-NEXT: movapd 160(%rdi), %xmm11
-; SSE-NEXT: movapd 192(%rdi), %xmm13
-; SSE-NEXT: movapd 80(%rdi), %xmm12
-; SSE-NEXT: movapd 112(%rdi), %xmm0
-; SSE-NEXT: movapd %xmm14, %xmm15
-; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm10[0],xmm15[1]
-; SSE-NEXT: movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufpd {{.*#+}} xmm10 = xmm10[1],xmm9[0]
-; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm7[0],xmm9[1]
-; SSE-NEXT: movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufpd {{.*#+}} xmm7 = xmm7[1],xmm3[0]
-; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm14[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd %xmm0, %xmm3
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm12[0],xmm3[1]
-; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufpd {{.*#+}} xmm12 = xmm12[1],xmm8[0]
+; SSE-NEXT: movapd 192(%rdi), %xmm15
+; SSE-NEXT: movapd %xmm13, %xmm12
+; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm8[0],xmm12[1]
; SSE-NEXT: movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm5[0],xmm8[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm8 = xmm8[1],xmm9[0]
; SSE-NEXT: movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm5[0],xmm9[1]
+; SSE-NEXT: movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufpd {{.*#+}} xmm5 = xmm5[1],xmm2[0]
; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm13[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd %xmm13, %xmm0
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm11[0],xmm0[1]
-; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufpd {{.*#+}} xmm11 = xmm11[1],xmm6[0]
-; SSE-NEXT: movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
+; SSE-NEXT: movapd %xmm14, %xmm2
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm10[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufpd {{.*#+}} xmm10 = xmm10[1],xmm6[0]
+; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm3[0],xmm6[1]
; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufpd {{.*#+}} xmm4 = xmm4[1],xmm1[0]
-; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm13[0],xmm1[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm3 = xmm3[1],xmm1[0]
+; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movapd %xmm15, %xmm1
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm11[0],xmm1[1]
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufpd {{.*#+}} xmm11 = xmm11[1],xmm7[0]
+; SSE-NEXT: movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm4[0],xmm7[1]
+; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufpd {{.*#+}} xmm4 = xmm4[1],xmm0[0]
+; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm15[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movapd 240(%rdi), %xmm2
; SSE-NEXT: movapd 272(%rdi), %xmm0
; SSE-NEXT: movapd %xmm0, %xmm1
@@ -770,19 +770,19 @@
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movapd 256(%rdi), %xmm2
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movapd 304(%rdi), %xmm1
; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd 320(%rdi), %xmm14
+; SSE-NEXT: movapd 320(%rdi), %xmm15
; SSE-NEXT: movapd 352(%rdi), %xmm0
; SSE-NEXT: movapd %xmm0, %xmm1
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm15[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movapd 368(%rdi), %xmm1
-; SSE-NEXT: shufpd {{.*#+}} xmm14 = xmm14[1],xmm1[0]
+; SSE-NEXT: shufpd {{.*#+}} xmm15 = xmm15[1],xmm1[0]
; SSE-NEXT: movapd 336(%rdi), %xmm2
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -790,82 +790,82 @@
; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd 400(%rdi), %xmm11
-; SSE-NEXT: movapd 432(%rdi), %xmm0
-; SSE-NEXT: movapd %xmm0, %xmm15
-; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm11[0],xmm15[1]
+; SSE-NEXT: movapd %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT: movapd 400(%rdi), %xmm8
+; SSE-NEXT: movapd 432(%rdi), %xmm1
+; SSE-NEXT: movapd %xmm1, %xmm13
+; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm8[0],xmm13[1]
; SSE-NEXT: movapd 448(%rdi), %xmm12
-; SSE-NEXT: shufpd {{.*#+}} xmm11 = xmm11[1],xmm12[0]
-; SSE-NEXT: movapd 416(%rdi), %xmm13
-; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm13[0],xmm12[1]
-; SSE-NEXT: movapd 464(%rdi), %xmm1
-; SSE-NEXT: shufpd {{.*#+}} xmm13 = xmm13[1],xmm1[0]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd 480(%rdi), %xmm2
-; SSE-NEXT: movapd 512(%rdi), %xmm5
-; SSE-NEXT: movapd %xmm5, %xmm7
-; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm2[0],xmm7[1]
-; SSE-NEXT: movapd 528(%rdi), %xmm3
-; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1],xmm3[0]
-; SSE-NEXT: movapd 496(%rdi), %xmm8
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm8[0],xmm3[1]
-; SSE-NEXT: movapd 544(%rdi), %xmm9
-; SSE-NEXT: shufpd {{.*#+}} xmm8 = xmm8[1],xmm9[0]
-; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm5[0],xmm9[1]
-; SSE-NEXT: movapd 560(%rdi), %xmm5
-; SSE-NEXT: movapd 592(%rdi), %xmm10
-; SSE-NEXT: movapd %xmm10, %xmm6
-; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm5[0],xmm6[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm8 = xmm8[1],xmm12[0]
+; SSE-NEXT: movapd 416(%rdi), %xmm14
+; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm14[0],xmm12[1]
+; SSE-NEXT: movapd 464(%rdi), %xmm0
+; SSE-NEXT: shufpd {{.*#+}} xmm14 = xmm14[1],xmm0[0]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movapd 480(%rdi), %xmm3
+; SSE-NEXT: movapd 512(%rdi), %xmm6
+; SSE-NEXT: movapd %xmm6, %xmm9
+; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm3[0],xmm9[1]
+; SSE-NEXT: movapd 528(%rdi), %xmm5
+; SSE-NEXT: shufpd {{.*#+}} xmm3 = xmm3[1],xmm5[0]
+; SSE-NEXT: movapd 496(%rdi), %xmm4
+; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; SSE-NEXT: movapd 544(%rdi), %xmm10
+; SSE-NEXT: shufpd {{.*#+}} xmm4 = xmm4[1],xmm10[0]
+; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm6[0],xmm10[1]
+; SSE-NEXT: movapd 560(%rdi), %xmm6
+; SSE-NEXT: movapd 592(%rdi), %xmm11
+; SSE-NEXT: movapd %xmm11, %xmm7
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm6[0],xmm7[1]
; SSE-NEXT: movapd 608(%rdi), %xmm0
-; SSE-NEXT: shufpd {{.*#+}} xmm5 = xmm5[1],xmm0[0]
+; SSE-NEXT: shufpd {{.*#+}} xmm6 = xmm6[1],xmm0[0]
; SSE-NEXT: movapd 576(%rdi), %xmm1
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd 624(%rdi), %xmm4
-; SSE-NEXT: shufpd {{.*#+}} xmm1 = xmm1[1],xmm4[0]
-; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm10[0],xmm4[1]
-; SSE-NEXT: movapd %xmm7, 96(%rsi)
+; SSE-NEXT: movapd 624(%rdi), %xmm2
+; SSE-NEXT: shufpd {{.*#+}} xmm1 = xmm1[1],xmm2[0]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm11[0],xmm2[1]
+; SSE-NEXT: movapd %xmm13, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: movaps %xmm11, 16(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: movaps %xmm11, 64(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: movaps %xmm11, (%rsi)
+; SSE-NEXT: movapd %xmm7, 112(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movaps %xmm7, 48(%rsi)
+; SSE-NEXT: movapd %xmm9, 96(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; SSE-NEXT: movaps %xmm7, 32(%rsi)
-; SSE-NEXT: movapd %xmm6, 112(%rsi)
+; SSE-NEXT: movapd %xmm8, 80(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movaps %xmm7, 16(%rdx)
+; SSE-NEXT: movapd %xmm15, 64(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: movaps %xmm7, (%rdx)
+; SSE-NEXT: movapd %xmm6, 112(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: movaps %xmm6, 48(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: movaps %xmm6, 64(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: movaps %xmm6, (%rsi)
-; SSE-NEXT: movapd %xmm15, 80(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: movaps %xmm6, 16(%rsi)
-; SSE-NEXT: movapd %xmm2, 96(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 32(%rdx)
-; SSE-NEXT: movapd %xmm5, 112(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 48(%rdx)
-; SSE-NEXT: movapd %xmm14, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, (%rdx)
-; SSE-NEXT: movapd %xmm11, 80(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 16(%rdx)
-; SSE-NEXT: movapd %xmm3, 96(%rcx)
-; SSE-NEXT: movapd %xmm0, 112(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 64(%rcx)
+; SSE-NEXT: movaps %xmm6, 48(%rdx)
+; SSE-NEXT: movapd %xmm3, 96(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 16(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, (%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 32(%rcx)
; SSE-NEXT: movapd %xmm12, 80(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rcx)
-; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 64(%rcx)
+; SSE-NEXT: movapd %xmm0, 112(%rcx)
+; SSE-NEXT: movapd %xmm5, 96(%rcx)
; SSE-NEXT: movapd %xmm1, 112(%r8)
-; SSE-NEXT: movapd %xmm8, 96(%r8)
-; SSE-NEXT: movapd %xmm13, 80(%r8)
+; SSE-NEXT: movapd %xmm4, 96(%r8)
+; SSE-NEXT: movapd %xmm14, 80(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 64(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -876,11 +876,11 @@
; SSE-NEXT: movaps %xmm0, 16(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%r8)
-; SSE-NEXT: movapd %xmm4, 112(%r9)
-; SSE-NEXT: movapd %xmm9, 96(%r9)
+; SSE-NEXT: movapd %xmm2, 112(%r9)
+; SSE-NEXT: movapd %xmm10, 96(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 80(%r9)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 64(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 48(%r9)
@@ -895,7 +895,7 @@
;
; AVX1-ONLY-LABEL: load_i64_stride5_vf16:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $376, %rsp # imm = 0x178
+; AVX1-ONLY-NEXT: subq $360, %rsp # imm = 0x168
; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 64(%rdi), %ymm5
@@ -915,28 +915,29 @@
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm1[0,1,2],ymm7[3]
; AVX1-ONLY-NEXT: vmovapd 512(%rdi), %xmm6
; AVX1-ONLY-NEXT: vmovapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm8 = xmm11[0],xmm6[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm8[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm7 = xmm9[0],xmm6[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm7[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm5[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd (%rdi), %xmm14
+; AVX1-ONLY-NEXT: vmovapd (%rdi), %xmm13
; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm9 = xmm14[0],xmm0[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm9[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm11 = xmm13[0],xmm0[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm11[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm3
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm3[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm13 = xmm0[0],xmm12[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm13[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm14 = xmm0[0],xmm6[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm14[0,1],ymm15[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm15
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm15
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm15[0],ymm2[3],ymm15[2]
; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm15
; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = xmm4[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
@@ -946,123 +947,122 @@
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm4
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[3],ymm4[2]
; AVX1-ONLY-NEXT: vmovdqa 528(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm11 = xmm11[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm9 = xmm9[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm11
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm11[0],ymm5[3],ymm11[2]
-; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm14 = xmm14[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm14[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm9[0],ymm5[3],ymm9[2]
+; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm13 = xmm13[8,9,10,11,12,13,14,15],xmm9[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm13[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm5
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm5[0],ymm3[3],ymm5[2]
; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm8
; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm14[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm13 = xmm9[0,1,2,3],xmm15[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm14 = xmm11[0,1,2,3],xmm15[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovdqa 496(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm13[0,1,2,3],xmm7[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 496(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm5[0,1,2,3],xmm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm7[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm7[0,1],xmm6[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm6[0,1,2,3],xmm11[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm11[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm9 = xmm10[0,1,2,3],xmm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm9[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm2[0,1],xmm14[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm8 = xmm10[0,1,2,3],xmm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm8[0,1,2,3],ymm14[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm2[0,1],xmm13[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm8 = xmm12[0,1,2,3],xmm8[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm8[0,1,2,3],ymm13[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
; AVX1-ONLY-NEXT: vmovapd 288(%rdi), %ymm8
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm8[0],ymm4[3],ymm8[2]
-; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm9 = xmm9[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm9[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm9 = xmm11[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm4[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[3],ymm4[2]
-; AVX1-ONLY-NEXT: vmovdqa 544(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm9 = xmm13[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm11[0],ymm3[3],ymm11[2]
+; AVX1-ONLY-NEXT: vmovdqa 544(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = xmm5[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm3
; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm7
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm7[0],ymm3[3],ymm7[2]
-; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = xmm6[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = xmm10[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm10[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovapd 448(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm6[0],ymm2[3],ymm6[2]
+; AVX1-ONLY-NEXT: vmovapd 448(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm10[0],ymm2[3],ymm10[2]
; AVX1-ONLY-NEXT: vmovdqa 384(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = xmm10[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm10[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm12 = xmm12[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm12[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm8 = mem[0,1,2],ymm8[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2,3],xmm5[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = mem[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3],xmm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = mem[0,1,2],ymm11[3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,2,3],xmm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm8 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm8 = mem[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2],ymm7[3]
; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = mem[0,1,2],ymm7[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,2,3],xmm13[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = mem[0,1,2],ymm6[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm0 = xmm12[0],xmm0[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovaps %ymm14, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovaps %ymm15, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rcx)
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm13, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm14, (%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm15, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 32(%rcx)
; AVX1-ONLY-NEXT: vmovapd %ymm2, 64(%r8)
; AVX1-ONLY-NEXT: vmovapd %ymm3, (%r8)
-; AVX1-ONLY-NEXT: vmovapd %ymm9, 96(%r8)
-; AVX1-ONLY-NEXT: vmovapd %ymm11, 32(%r8)
+; AVX1-ONLY-NEXT: vmovapd %ymm5, 96(%r8)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, 32(%r8)
+; AVX1-ONLY-NEXT: vmovapd %ymm1, (%r9)
; AVX1-ONLY-NEXT: vmovapd %ymm0, 64(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm4, (%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm1, 96(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm5, 32(%r9)
-; AVX1-ONLY-NEXT: addq $376, %rsp # imm = 0x178
+; AVX1-ONLY-NEXT: vmovapd %ymm4, 96(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 32(%r9)
+; AVX1-ONLY-NEXT: addq $360, %rsp # imm = 0x168
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -1072,55 +1072,55 @@
; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm14
; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm4
; AVX2-ONLY-NEXT: vmovdqa 576(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm2[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm12[6,7]
; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %xmm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm5[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm5[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm13[6,7]
; AVX2-ONLY-NEXT: vmovdqa 512(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 480(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm9 = xmm8[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 480(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm9 = xmm7[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %xmm11
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vmovdqa (%rdi), %xmm9
; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm9 = xmm11[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm10 = xmm9[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm12[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm10[0,1,2,3,4,5],ymm11[6,7]
; AVX2-ONLY-NEXT: vmovdqa 352(%rdi), %xmm6
; AVX2-ONLY-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %xmm0
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm15 = xmm0[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 208(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = xmm5[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm2 = ymm2[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm2[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,1,2,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 528(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = xmm8[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm1[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vmovdqa 208(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = xmm5[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm1[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,2,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 48(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = xmm11[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 528(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = xmm7[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm3 = ymm3[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm3[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,1,2,1]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 48(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = xmm9[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm3
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm4 = ymm4[8,9,10,11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,2,1]
@@ -1129,7 +1129,7 @@
; AVX2-ONLY-NEXT: vmovdqa 368(%rdi), %xmm4
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = xmm0[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm5 = ymm12[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],ymm12[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm5 = ymm10[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,1,2,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1161,7 +1161,7 @@
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm4[2,3],ymm5[2,3]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm4 = ymm10[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm4 = ymm12[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm12[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,3]
; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %xmm15
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = mem[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
@@ -1169,35 +1169,35 @@
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm4 = ymm13[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm13[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = mem[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm5 = ymm14[8,9,10,11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23]
; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,1,0,3]
; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm12 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm12[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm12 = ymm9[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm12 = ymm12[0,1,0,3]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm10[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm11[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[0,1,0,3]
; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm11 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm9 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm12[0,1,2,3,4,5],ymm8[6,7]
; AVX2-ONLY-NEXT: vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm10 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm10 = mem[0,1],xmm15[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm10[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm10[0,1,2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm13[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm8 = mem[0,1],xmm8[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm14[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm7 = mem[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm11[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, 64(%rsi)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -1222,15 +1222,15 @@
; AVX2-ONLY-NEXT: vmovaps %ymm2, 96(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rcx)
-; AVX2-ONLY-NEXT: vmovdqa %ymm11, 64(%r8)
+; AVX2-ONLY-NEXT: vmovdqa %ymm9, 64(%r8)
; AVX2-ONLY-NEXT: vmovdqa %ymm5, (%r8)
; AVX2-ONLY-NEXT: vmovdqa %ymm4, 96(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%r8)
-; AVX2-ONLY-NEXT: vmovdqa %ymm0, 64(%r9)
; AVX2-ONLY-NEXT: vmovdqa %ymm1, (%r9)
+; AVX2-ONLY-NEXT: vmovdqa %ymm0, 64(%r9)
; AVX2-ONLY-NEXT: vmovdqa %ymm6, 96(%r9)
-; AVX2-ONLY-NEXT: vmovdqa %ymm7, 32(%r9)
+; AVX2-ONLY-NEXT: vmovdqa %ymm8, 32(%r9)
; AVX2-ONLY-NEXT: addq $360, %rsp # imm = 0x168
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
@@ -1837,401 +1837,395 @@
;
; AVX1-ONLY-LABEL: load_i64_stride5_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1384, %rsp # imm = 0x568
-; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: subq $1336, %rsp # imm = 0x538
+; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm4
; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 576(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 544(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovapd 576(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 544(%rdi), %ymm14
; AVX1-ONLY-NEXT: vmovapd 256(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm11[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm1 = xmm12[0],xmm1[1]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm9[0,1,2],ymm2[3]
-; AVX1-ONLY-NEXT: vmovapd 512(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm2 = xmm6[0],xmm2[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1,2],ymm3[3]
+; AVX1-ONLY-NEXT: vmovapd 512(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm1 = xmm15[0],xmm1[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm14[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1216(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1184(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm11[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 1216(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 1184(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovapd 1152(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1120(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm2 = xmm13[0],xmm2[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 1120(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm2 = xmm7[0],xmm2[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm12
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm12[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm15[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm8[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm3 = xmm1[0],xmm0[1]
+; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm3 = xmm5[0],xmm0[1]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 736(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm3[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm3[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovapd 672(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovapd 640(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm5 = xmm2[0],xmm0[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm10 = xmm2[0],xmm0[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm10[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm9[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovaps 992(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm4[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm10[0,1,2,3],ymm5[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm10
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[0],ymm10[0],ymm7[3],ymm10[2]
-; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm9[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm13[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm11[0],ymm13[0],ymm11[3],ymm13[2]
+; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm12 = xmm12[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm14[0],ymm13[0],ymm14[3],ymm13[2]
+; AVX1-ONLY-NEXT: vmovdqa 528(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm15 = xmm15[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm0[0],ymm13[0],ymm0[3],ymm13[2]
+; AVX1-ONLY-NEXT: vmovdqa 848(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = xmm8[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm9[0],ymm7[0],ymm9[3],ymm7[2]
-; AVX1-ONLY-NEXT: vmovdqa 528(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm6[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[3],ymm0[2]
-; AVX1-ONLY-NEXT: vmovdqa 848(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = xmm14[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa %xmm6, %xmm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm1 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm13[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm11[0],ymm7[0],ymm11[3],ymm7[2]
-; AVX1-ONLY-NEXT: vmovdqa 1168(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = xmm13[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm10[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[3],ymm7[2]
-; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = xmm12[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm15[0],ymm6[0],ymm15[3],ymm6[2]
-; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm1 = xmm1[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[3],ymm1[2]
-; AVX1-ONLY-NEXT: vmovdqa 688(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[3],ymm1[2]
-; AVX1-ONLY-NEXT: vmovdqa 1008(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm4[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm5[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovaps 176(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm8[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm5[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovaps 496(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm7[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm13[0,1],xmm9[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovdqa 816(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm9[0,1,2,3],xmm14[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm14[0,1],xmm0[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovaps 1136(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = xmm4[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm0[0,1,2,3],xmm11[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm1[0],ymm6[3],ymm1[2]
+; AVX1-ONLY-NEXT: vmovdqa 1168(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1],xmm12[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm0[0],ymm6[0],ymm0[3],ymm6[2]
+; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = xmm4[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm8[0],ymm4[0],ymm8[3],ymm4[2]
+; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm0[0,1,2,3],xmm15[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm15[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovaps 656(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vblendps $12, (%rsp), %xmm6, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm6[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vmovdqa 976(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm11 = xmm1[0,1,2,3],xmm10[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovapd 288(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm2[0],ymm12[0],ymm2[3],ymm12[2]
-; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = xmm8[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm8[0,1],ymm11[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm5[0],ymm2[0],ymm5[3],ymm2[2]
-; AVX1-ONLY-NEXT: vmovdqa 544(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm7[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovapd 928(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm8[0],ymm11[0],ymm8[3],ymm11[2]
-; AVX1-ONLY-NEXT: vmovdqa 864(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm9 = xmm9[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm9[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vmovapd 1248(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm8[0],ymm2[0],ymm8[3],ymm2[2]
-; AVX1-ONLY-NEXT: vmovdqa 1184(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = xmm4[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm4[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm5[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovapd 1088(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[3],ymm9[2]
-; AVX1-ONLY-NEXT: vmovdqa 1024(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm1 = xmm1[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[3],ymm0[2]
+; AVX1-ONLY-NEXT: vmovdqa 688(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovapd 768(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm0[0],ymm3[0],ymm0[3],ymm3[2]
-; AVX1-ONLY-NEXT: vmovdqa 704(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = xmm6[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[3],ymm2[2]
+; AVX1-ONLY-NEXT: vmovdqa 1008(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm9[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm10[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm13[0,1,2,3],xmm11[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1],xmm12[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vmovdqa 496(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm9 = xmm2[0,1,2,3],xmm14[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm9[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm3[0,1],xmm15[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vmovaps 816(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm5[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm9 = xmm5[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vmovdqa 1136(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm11[0,1,2,3],xmm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1056(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vmovdqa 976(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm8 = xmm0[0,1,2,3],xmm8[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = xmm8[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm7 = xmm9[0,1,2,3],xmm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm9[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT: vmovaps 336(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = xmm10[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm14[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT: vmovaps 656(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm8 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm8 = xmm12[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vmovapd 288(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm4[0],ymm15[0],ymm4[3],ymm15[2]
+; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm13 = xmm13[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm13[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm6[0],ymm4[0],ymm6[3],ymm4[2]
+; AVX1-ONLY-NEXT: vmovdqa 544(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovapd 928(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[3],ymm3[2]
+; AVX1-ONLY-NEXT: vmovdqa 864(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovapd 1248(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[3],ymm3[2]
+; AVX1-ONLY-NEXT: vmovdqa 1184(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm11[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovapd 1088(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm6[0],ymm1[3],ymm6[2]
+; AVX1-ONLY-NEXT: vmovdqa 1024(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vmovapd 448(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm8[0],ymm4[3],ymm8[2]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovapd 768(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm0[0],ymm4[0],ymm0[3],ymm4[2]
+; AVX1-ONLY-NEXT: vmovdqa 704(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm12 = xmm12[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm12[0,1],ymm11[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vmovapd 448(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm9[0],ymm2[0],ymm9[3],ymm2[2]
; AVX1-ONLY-NEXT: vmovdqa 384(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm0[0],ymm4[3],ymm0[2]
-; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,2,3],xmm13[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm4[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm12[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,2,3],xmm10[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm4[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = xmm10[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm10[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm12
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm9[0],ymm12[0],ymm9[3],ymm12[2]
+; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm9 = mem[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm0[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm15[3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,2,3],xmm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm7[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm2[3]
; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm1[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3],xmm5[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm3[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3],xmm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm1[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm11[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3],xmm7[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm1[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm9[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3],xmm14[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3],xmm8[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2,3],xmm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 192(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 128(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 160(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 192(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 128(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 160(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 192(%rcx)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 128(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 160(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, (%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 128(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 192(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 160(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 224(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm5, 192(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm4, 160(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm2, 128(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 96(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm8, 64(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm15, 32(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm13, (%r9)
-; AVX1-ONLY-NEXT: addq $1384, %rsp # imm = 0x568
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3],xmm13[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3],xmm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,1],xmm4[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 192(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 128(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 224(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 160(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 128(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 224(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 160(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 128(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 192(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 224(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 160(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm10, (%r8)
+; AVX1-ONLY-NEXT: vmovapd %ymm11, 64(%r8)
+; AVX1-ONLY-NEXT: vmovapd %ymm14, 128(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 192(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 224(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 160(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 96(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 224(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm0, 192(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 160(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm1, 128(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 96(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, 64(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, 32(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm12, (%r9)
+; AVX1-ONLY-NEXT: addq $1336, %rsp # imm = 0x538
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -2419,6 +2413,13 @@
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vmovaps 1056(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm0
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm1
@@ -2440,13 +2441,6 @@
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2,3],ymm0[4,5],mem[6,7]
-; AVX2-ONLY-NEXT: vmovaps 1056(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm0 = ymm13[8,9,10,11,12,13,14,15],ymm11[0,1,2,3,4,5,6,7],ymm13[24,25,26,27,28,29,30,31],ymm11[16,17,18,19,20,21,22,23]
@@ -2585,14 +2579,14 @@
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm6, 32(%rdx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 192(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm6, 128(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm6, 64(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm6, (%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm6, 192(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm6, 224(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm6, 160(%rcx)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-6.ll
index 05f0703..f14a54a 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-6.ll
@@ -403,56 +403,56 @@
; SSE: # %bb.0:
; SSE-NEXT: subq $24, %rsp
; SSE-NEXT: movaps 160(%rdi), %xmm8
-; SSE-NEXT: movaps 256(%rdi), %xmm9
-; SSE-NEXT: movaps 208(%rdi), %xmm0
+; SSE-NEXT: movaps 64(%rdi), %xmm9
; SSE-NEXT: movaps 352(%rdi), %xmm12
-; SSE-NEXT: movaps 304(%rdi), %xmm1
-; SSE-NEXT: movaps 64(%rdi), %xmm15
+; SSE-NEXT: movaps 304(%rdi), %xmm0
+; SSE-NEXT: movaps 256(%rdi), %xmm15
+; SSE-NEXT: movaps 208(%rdi), %xmm1
; SSE-NEXT: movaps (%rdi), %xmm3
; SSE-NEXT: movaps 16(%rdi), %xmm2
; SSE-NEXT: movaps 48(%rdi), %xmm10
; SSE-NEXT: movaps 144(%rdi), %xmm14
-; SSE-NEXT: movaps 96(%rdi), %xmm4
-; SSE-NEXT: movaps 240(%rdi), %xmm13
-; SSE-NEXT: movaps 192(%rdi), %xmm5
-; SSE-NEXT: movaps 336(%rdi), %xmm11
-; SSE-NEXT: movaps 288(%rdi), %xmm6
+; SSE-NEXT: movaps 96(%rdi), %xmm5
+; SSE-NEXT: movaps 336(%rdi), %xmm13
+; SSE-NEXT: movaps 288(%rdi), %xmm4
+; SSE-NEXT: movaps 240(%rdi), %xmm11
+; SSE-NEXT: movaps 192(%rdi), %xmm6
; SSE-NEXT: movaps %xmm6, %xmm7
; SSE-NEXT: movlhps {{.*#+}} xmm7 = xmm7[0],xmm11[0]
; SSE-NEXT: movaps %xmm7, (%rsp) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm11[1]
; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm5, %xmm11
+; SSE-NEXT: movaps %xmm4, %xmm11
; SSE-NEXT: movlhps {{.*#+}} xmm11 = xmm11[0],xmm13[0]
-; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm13[1]
-; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm4, %xmm13
-; SSE-NEXT: movlhps {{.*#+}} xmm13 = xmm13[0],xmm14[0]
-; SSE-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm14[1]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm13[1]
; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm5, %xmm13
+; SSE-NEXT: movlhps {{.*#+}} xmm13 = xmm13[0],xmm14[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm14[1]
+; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm3, %xmm14
; SSE-NEXT: movlhps {{.*#+}} xmm14 = xmm14[0],xmm10[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm10[1]
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm2, %xmm3
+; SSE-NEXT: movaps %xmm1, %xmm3
; SSE-NEXT: movlhps {{.*#+}} xmm3 = xmm3[0],xmm15[0]
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm15[1]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm1, %xmm15
-; SSE-NEXT: movlhps {{.*#+}} xmm15 = xmm15[0],xmm12[0]
-; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm12[1]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm15[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm0, %xmm12
-; SSE-NEXT: movlhps {{.*#+}} xmm12 = xmm12[0],xmm9[0]
-; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm9[1]
+; SSE-NEXT: movaps %xmm0, %xmm15
+; SSE-NEXT: movlhps {{.*#+}} xmm15 = xmm15[0],xmm12[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm12[1]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm2, %xmm12
+; SSE-NEXT: movlhps {{.*#+}} xmm12 = xmm12[0],xmm9[0]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm9[1]
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 112(%rdi), %xmm7
; SSE-NEXT: movaps %xmm7, %xmm9
; SSE-NEXT: movlhps {{.*#+}} xmm9 = xmm9[0],xmm8[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm8[1]
-; SSE-NEXT: movaps 80(%rdi), %xmm1
-; SSE-NEXT: movaps 32(%rdi), %xmm8
+; SSE-NEXT: movaps 272(%rdi), %xmm1
+; SSE-NEXT: movaps 224(%rdi), %xmm8
; SSE-NEXT: movaps %xmm8, %xmm10
; SSE-NEXT: movlhps {{.*#+}} xmm10 = xmm10[0],xmm1[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm1[1]
@@ -461,8 +461,8 @@
; SSE-NEXT: movaps %xmm3, %xmm6
; SSE-NEXT: movlhps {{.*#+}} xmm6 = xmm6[0],xmm1[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
-; SSE-NEXT: movaps 272(%rdi), %xmm1
-; SSE-NEXT: movaps 224(%rdi), %xmm4
+; SSE-NEXT: movaps 80(%rdi), %xmm1
+; SSE-NEXT: movaps 32(%rdi), %xmm4
; SSE-NEXT: movaps %xmm4, %xmm5
; SSE-NEXT: movlhps {{.*#+}} xmm5 = xmm5[0],xmm1[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm1[1]
@@ -472,39 +472,39 @@
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; SSE-NEXT: movaps %xmm13, 16(%rsi)
-; SSE-NEXT: movaps %xmm11, 32(%rsi)
-; SSE-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 48(%rsi)
; SSE-NEXT: movaps %xmm14, (%rsi)
+; SSE-NEXT: movaps %xmm11, 48(%rsi)
+; SSE-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT: movaps %xmm1, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%rdx)
+; SSE-NEXT: movaps %xmm1, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%rdx)
+; SSE-NEXT: movaps %xmm1, 32(%rdx)
; SSE-NEXT: movaps %xmm9, 16(%rcx)
-; SSE-NEXT: movaps %xmm12, 32(%rcx)
+; SSE-NEXT: movaps %xmm12, (%rcx)
; SSE-NEXT: movaps %xmm15, 48(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%rcx)
+; SSE-NEXT: movaps %xmm1, 32(%rcx)
; SSE-NEXT: movaps %xmm7, 16(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, 32(%r8)
+; SSE-NEXT: movaps %xmm1, (%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps %xmm1, 48(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movaps %xmm1, (%r8)
+; SSE-NEXT: movaps %xmm1, 32(%r8)
; SSE-NEXT: movaps %xmm2, 16(%r9)
-; SSE-NEXT: movaps %xmm5, 32(%r9)
+; SSE-NEXT: movaps %xmm5, (%r9)
; SSE-NEXT: movaps %xmm6, 48(%r9)
-; SSE-NEXT: movaps %xmm10, (%r9)
+; SSE-NEXT: movaps %xmm10, 32(%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT: movaps %xmm0, 16(%rax)
-; SSE-NEXT: movaps %xmm4, 32(%rax)
+; SSE-NEXT: movaps %xmm4, (%rax)
; SSE-NEXT: movaps %xmm3, 48(%rax)
-; SSE-NEXT: movaps %xmm8, (%rax)
+; SSE-NEXT: movaps %xmm8, 32(%rax)
; SSE-NEXT: addq $24, %rsp
; SSE-NEXT: retq
;
@@ -869,35 +869,35 @@
; SSE: # %bb.0:
; SSE-NEXT: subq $408, %rsp # imm = 0x198
; SSE-NEXT: movaps (%rdi), %xmm7
-; SSE-NEXT: movaps 624(%rdi), %xmm0
-; SSE-NEXT: movaps 576(%rdi), %xmm8
-; SSE-NEXT: movaps 240(%rdi), %xmm1
-; SSE-NEXT: movaps 192(%rdi), %xmm9
-; SSE-NEXT: movaps 720(%rdi), %xmm2
-; SSE-NEXT: movaps 672(%rdi), %xmm10
-; SSE-NEXT: movaps 336(%rdi), %xmm3
-; SSE-NEXT: movaps 288(%rdi), %xmm11
-; SSE-NEXT: movaps 432(%rdi), %xmm4
-; SSE-NEXT: movaps 384(%rdi), %xmm13
-; SSE-NEXT: movaps 528(%rdi), %xmm5
-; SSE-NEXT: movaps 480(%rdi), %xmm12
-; SSE-NEXT: movaps 144(%rdi), %xmm6
-; SSE-NEXT: movaps 96(%rdi), %xmm14
+; SSE-NEXT: movaps 528(%rdi), %xmm0
+; SSE-NEXT: movaps 480(%rdi), %xmm8
+; SSE-NEXT: movaps 144(%rdi), %xmm1
+; SSE-NEXT: movaps 96(%rdi), %xmm9
+; SSE-NEXT: movaps 432(%rdi), %xmm2
+; SSE-NEXT: movaps 384(%rdi), %xmm10
+; SSE-NEXT: movaps 720(%rdi), %xmm3
+; SSE-NEXT: movaps 672(%rdi), %xmm11
+; SSE-NEXT: movaps 336(%rdi), %xmm4
+; SSE-NEXT: movaps 288(%rdi), %xmm12
+; SSE-NEXT: movaps 624(%rdi), %xmm5
+; SSE-NEXT: movaps 576(%rdi), %xmm13
+; SSE-NEXT: movaps 240(%rdi), %xmm6
+; SSE-NEXT: movaps 192(%rdi), %xmm14
; SSE-NEXT: movaps %xmm14, %xmm15
; SSE-NEXT: movlhps {{.*#+}} xmm15 = xmm15[0],xmm6[0]
; SSE-NEXT: movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm6[1]
; SSE-NEXT: movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm12, %xmm6
+; SSE-NEXT: movaps %xmm13, %xmm6
; SSE-NEXT: movlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm5[1]
-; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps %xmm13, %xmm5
+; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm5[1]
+; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm12, %xmm5
; SSE-NEXT: movlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm4[1]
-; SSE-NEXT: movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm4[1]
+; SSE-NEXT: movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps %xmm11, %xmm4
; SSE-NEXT: movlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -949,7 +949,7 @@
; SSE-NEXT: movaps 304(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
-; SSE-NEXT: movaps %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 448(%rdi), %xmm0
@@ -960,36 +960,36 @@
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 544(%rdi), %xmm0
-; SSE-NEXT: movaps 496(%rdi), %xmm14
-; SSE-NEXT: movaps %xmm14, %xmm1
-; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm0[1]
-; SSE-NEXT: movaps 640(%rdi), %xmm0
-; SSE-NEXT: movaps 592(%rdi), %xmm12
+; SSE-NEXT: movaps 496(%rdi), %xmm12
; SSE-NEXT: movaps %xmm12, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
-; SSE-NEXT: movaps 736(%rdi), %xmm0
-; SSE-NEXT: movaps 688(%rdi), %xmm8
-; SSE-NEXT: movaps %xmm8, %xmm1
+; SSE-NEXT: movaps 640(%rdi), %xmm0
+; SSE-NEXT: movaps 592(%rdi), %xmm9
+; SSE-NEXT: movaps %xmm9, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm0[1]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm0[1]
+; SSE-NEXT: movaps 736(%rdi), %xmm0
+; SSE-NEXT: movaps 688(%rdi), %xmm15
+; SSE-NEXT: movaps %xmm15, %xmm1
+; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
; SSE-NEXT: movaps 80(%rdi), %xmm0
; SSE-NEXT: movaps 32(%rdi), %xmm1
; SSE-NEXT: movaps %xmm1, %xmm2
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
-; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps %xmm2, (%rsp) # 16-byte Spill
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movaps 176(%rdi), %xmm0
-; SSE-NEXT: movaps 128(%rdi), %xmm13
-; SSE-NEXT: movaps %xmm13, %xmm1
+; SSE-NEXT: movaps 128(%rdi), %xmm14
+; SSE-NEXT: movaps %xmm14, %xmm1
; SSE-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm0[1]
+; SSE-NEXT: unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm0[1]
; SSE-NEXT: movaps 272(%rdi), %xmm0
; SSE-NEXT: movaps 224(%rdi), %xmm11
; SSE-NEXT: movaps %xmm11, %xmm1
@@ -998,13 +998,13 @@
; SSE-NEXT: unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm0[1]
; SSE-NEXT: movaps 368(%rdi), %xmm0
; SSE-NEXT: movaps 320(%rdi), %xmm10
-; SSE-NEXT: movaps %xmm10, %xmm15
-; SSE-NEXT: movlhps {{.*#+}} xmm15 = xmm15[0],xmm0[0]
+; SSE-NEXT: movaps %xmm10, %xmm13
+; SSE-NEXT: movlhps {{.*#+}} xmm13 = xmm13[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm10 = xmm10[1],xmm0[1]
; SSE-NEXT: movaps 464(%rdi), %xmm0
; SSE-NEXT: movaps 416(%rdi), %xmm6
-; SSE-NEXT: movaps %xmm6, %xmm9
-; SSE-NEXT: movlhps {{.*#+}} xmm9 = xmm9[0],xmm0[0]
+; SSE-NEXT: movaps %xmm6, %xmm8
+; SSE-NEXT: movlhps {{.*#+}} xmm8 = xmm8[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm0[1]
; SSE-NEXT: movaps 560(%rdi), %xmm0
; SSE-NEXT: movaps 512(%rdi), %xmm5
@@ -1022,56 +1022,56 @@
; SSE-NEXT: movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 96(%rsi)
+; SSE-NEXT: movaps %xmm0, 80(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 112(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rsi)
+; SSE-NEXT: movaps %xmm0, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 64(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 80(%rsi)
+; SSE-NEXT: movaps %xmm0, 112(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%rsi)
+; SSE-NEXT: movaps %xmm0, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 96(%rdx)
+; SSE-NEXT: movaps %xmm0, 96(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 112(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: movaps %xmm0, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 80(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 96(%rcx)
+; SSE-NEXT: movaps %xmm0, 64(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 112(%rcx)
+; SSE-NEXT: movaps %xmm0, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 64(%rcx)
+; SSE-NEXT: movaps %xmm0, 112(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 80(%rcx)
+; SSE-NEXT: movaps %xmm0, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rcx)
-; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rcx)
+; SSE-NEXT: movaps %xmm0, 96(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%rcx)
-; SSE-NEXT: movaps %xmm8, 112(%r8)
-; SSE-NEXT: movaps %xmm12, 96(%r8)
-; SSE-NEXT: movaps %xmm14, 80(%r8)
+; SSE-NEXT: movaps %xmm0, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 80(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 64(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 112(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 96(%rcx)
+; SSE-NEXT: movaps %xmm15, 112(%r8)
+; SSE-NEXT: movaps %xmm9, 96(%r8)
+; SSE-NEXT: movaps %xmm12, 80(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 64(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -1085,13 +1085,13 @@
; SSE-NEXT: movaps %xmm2, 112(%r9)
; SSE-NEXT: movaps %xmm4, 96(%r9)
; SSE-NEXT: movaps %xmm7, 80(%r9)
-; SSE-NEXT: movaps %xmm9, 64(%r9)
-; SSE-NEXT: movaps %xmm15, 48(%r9)
+; SSE-NEXT: movaps %xmm8, 64(%r9)
+; SSE-NEXT: movaps %xmm13, 48(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 32(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%r9)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT: movaps %xmm1, 112(%rax)
@@ -1100,7 +1100,7 @@
; SSE-NEXT: movaps %xmm6, 64(%rax)
; SSE-NEXT: movaps %xmm10, 48(%rax)
; SSE-NEXT: movaps %xmm11, 32(%rax)
-; SSE-NEXT: movaps %xmm13, 16(%rax)
+; SSE-NEXT: movaps %xmm14, 16(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%rax)
; SSE-NEXT: addq $408, %rsp # imm = 0x198
@@ -1108,11 +1108,11 @@
;
; AVX1-ONLY-LABEL: load_i64_stride6_vf16:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $552, %rsp # imm = 0x228
+; AVX1-ONLY-NEXT: subq $536, %rsp # imm = 0x218
; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovups %ymm8, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm2
; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdi), %ymm0, %ymm3
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
@@ -1147,120 +1147,120 @@
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm5[1],xmm6[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd (%rsp), %ymm4, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm4[1],mem[1],ymm4[3],mem[3]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm9[1],xmm8[1]
; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm3
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm4 = ymm7[1],mem[1],ymm7[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm12[1],xmm11[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm12[1],xmm11[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm13[1],ymm10[1],ymm13[3],ymm10[3]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm15[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm8[0],ymm2[2],ymm8[2]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm5[0],xmm11[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm0, %ymm11
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm11[0],ymm2[2],ymm11[2]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm13[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 544(%rdi), %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm3[0],ymm6[0],ymm3[2],ymm6[2]
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm14[0],xmm13[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 544(%rdi), %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm8[0],xmm7[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm6[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%rdi), %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm15[0],ymm7[0],ymm15[2],ymm7[2]
-; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm4[0],xmm5[0]
+; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, 736(%rdi), %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm6[0],ymm14[0],ymm6[2],ymm14[2]
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm3[0],xmm15[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm10[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, 736(%rdi), %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, 352(%rdi), %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm5[0],ymm2[0],ymm5[2],ymm2[2]
+; AVX1-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 208(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm0[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm12[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm9[1],xmm11[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm8[1],ymm6[1],ymm8[3],ymm6[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm14[1],xmm13[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm9 = ymm9[1],ymm11[1],ymm9[3],ymm11[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm10[1],xmm13[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm9[1],ymm4[1],ymm9[3],ymm4[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm8[1],xmm7[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm6[1],ymm14[1],ymm6[3],ymm14[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm15[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm5[1],ymm2[1],ymm5[3],ymm2[3]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm15[1],ymm7[1],ymm15[3],ymm7[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm5[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 512(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, 128(%rdi), %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm14[0],ymm1[0],ymm14[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm0[0],xmm12[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 464(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm14[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 544(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, 512(%rdi), %ymm0, %ymm8
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm8[0],ymm9[0],ymm8[2],ymm9[2]
+; AVX1-ONLY-NEXT: vmovaps 464(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 416(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm3[0],xmm6[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vinsertf128 $1, 128(%rdi), %ymm0, %ymm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm9[0],ymm10[0],ymm9[2],ymm10[2]
-; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm12[0],xmm8[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, 320(%rdi), %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vinsertf128 $1, 320(%rdi), %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm10[0],ymm11[0],ymm10[2],ymm11[2]
+; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm7
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm4[0],xmm5[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm4[0],xmm7[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 736(%rdi), %ymm13
; AVX1-ONLY-NEXT: vinsertf128 $1, 704(%rdi), %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm13[0],ymm2[2],ymm13[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm2[0],ymm13[0],ymm2[2],ymm13[2]
; AVX1-ONLY-NEXT: vmovaps 656(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm15 = xmm0[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm15[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm13[1],ymm2[3],ymm13[3]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm11[1],xmm14[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm6[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm9[1],ymm10[1],ymm9[3],ymm10[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm12[1],xmm8[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm14[1],mem[1],ymm14[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm12[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm6[1],ymm7[1],ymm6[3],ymm7[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm5[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm10[1],ymm11[1],ymm10[3],ymm11[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm7[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rsi)
@@ -1274,15 +1274,15 @@
; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rcx)
@@ -1294,232 +1294,231 @@
; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, (%r8)
-; AVX1-ONLY-NEXT: vmovaps %ymm15, 96(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%r9)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%r9)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm13, 96(%rax)
-; AVX1-ONLY-NEXT: addq $552, %rsp # imm = 0x228
+; AVX1-ONLY-NEXT: addq $536, %rsp # imm = 0x218
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-ONLY-LABEL: load_i64_stride6_vf16:
; AVX2-ONLY: # %bb.0:
-; AVX2-ONLY-NEXT: subq $520, %rsp # imm = 0x208
-; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm5
+; AVX2-ONLY-NEXT: subq $488, %rsp # imm = 0x1E8
+; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm6
; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm12
; AVX2-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm9
; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm4
; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps 48(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm2[0]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm0[0],ymm4[2],ymm0[2]
-; AVX2-ONLY-NEXT: vmovaps %ymm0, %ymm11
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 48(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm3[0]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm9[0],ymm4[2],ymm9[2]
+; AVX2-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 432(%rdi), %xmm15
-; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm8[0],xmm15[0]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-ONLY-NEXT: vmovaps %ymm7, %ymm6
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm9 = ymm9[0,1,0,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm11
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm11[0],xmm15[0]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm5[0],ymm7[0],ymm5[2],ymm7[2]
+; AVX2-ONLY-NEXT: vmovaps %ymm7, %ymm5
+; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,0,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm8[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 240(%rdi), %xmm9
+; AVX2-ONLY-NEXT: vmovaps 240(%rdi), %xmm8
; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm10
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm10[0],xmm9[0]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm12[0],ymm5[0],ymm12[2],ymm5[2]
-; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm10[0],xmm8[0]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm12[0],ymm6[0],ymm12[2],ymm6[2]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,1,0,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm12[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm4
; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm13[0],ymm3[0],ymm13[2],ymm3[2]
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm13[0],ymm4[0],ymm13[2],ymm4[2]
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm12 = ymm12[0,1,0,3]
; AVX2-ONLY-NEXT: vmovaps 624(%rdi), %xmm14
; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm0[0],xmm14[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT: vbroadcastsd 104(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm11[1],ymm2[3],ymm11[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
+; AVX2-ONLY-NEXT: vbroadcastsd 104(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm9[1],ymm3[3],ymm9[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm8[1],xmm15[1]
-; AVX2-ONLY-NEXT: vbroadcastsd 488(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm6[1],ymm2[3],ymm6[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm11[1],xmm15[1]
+; AVX2-ONLY-NEXT: vbroadcastsd 488(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm5[1],ymm3[3],ymm5[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm10[1],xmm8[1]
+; AVX2-ONLY-NEXT: vbroadcastsd 296(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm6[1],ymm3[3],ymm6[3]
; AVX2-ONLY-NEXT: vmovaps %ymm6, %ymm15
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm10[1],xmm9[1]
-; AVX2-ONLY-NEXT: vbroadcastsd 296(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm5[1],ymm2[3],ymm5[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm14[1]
; AVX2-ONLY-NEXT: vbroadcastsd 680(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 160(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm0[0],ymm11[2],ymm0[2]
-; AVX2-ONLY-NEXT: vmovaps 16(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm10
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm10[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm9[0],ymm0[0],ymm9[2],ymm0[2]
+; AVX2-ONLY-NEXT: vmovaps 16(%rdi), %xmm14
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm14[0],xmm3[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 544(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm4[0],ymm0[0],ymm4[2],ymm0[2]
-; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %xmm6
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[2],ymm0[2]
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %xmm4
; AVX2-ONLY-NEXT: vmovaps 400(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm7[0],xmm6[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vbroadcastsd 352(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm14[0],ymm0[0],ymm14[2],ymm0[2]
-; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vmovaps 208(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm3[0],xmm8[0]
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm7[0],xmm4[0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 736(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm13[0],ymm0[0],ymm13[2],ymm0[2]
-; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovaps 592(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm0[0],xmm2[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm9[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm10[1]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm0[0],ymm13[2],ymm0[2]
+; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vmovaps 592(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm10[0],xmm8[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastsd 352(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm11[0],ymm0[0],ymm11[2],ymm0[2]
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 208(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm14[1],xmm3[1]
; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm10 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,2,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
-; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm4[1],ymm10[1],ymm4[3],ymm10[3]
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,2,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm0[1],xmm2[1]
-; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm13[1],ymm1[1],ymm13[3],ymm1[3]
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm8[1]
-; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm14[1],ymm11[1],ymm14[3],ymm11[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm9[1],ymm12[1],ymm9[3],ymm12[3]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,2,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm7[1],xmm4[1]
+; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm6[1],ymm9[1],ymm6[3],ymm9[3]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,2,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm10[1],xmm8[1]
+; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm13[1],ymm2[1],ymm13[3],ymm2[3]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm15[0],ymm10[0],ymm15[2],ymm10[2]
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovaps 464(%rdi), %xmm9
-; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm8[0],xmm9[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm3, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm0[1],xmm1[1]
+; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm11[1],ymm8[1],ymm11[3],ymm8[3]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,2,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm0[0],ymm12[0],ymm0[2],ymm12[2]
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovaps 80(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm7[0],xmm6[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm12[0],ymm0[2],ymm12[2]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm11
+; AVX2-ONLY-NEXT: vmovaps 80(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm11[0],xmm10[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm0[0],ymm11[0],ymm0[2],ymm11[2]
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovaps 272(%rdi), %xmm5
-; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm2[0],xmm5[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-ONLY-NEXT: vmovaps 464(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm3[0],xmm4[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm15[0],ymm8[0],ymm15[2],ymm8[2]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
+; AVX2-ONLY-NEXT: vmovaps 272(%rdi), %xmm6
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm5[0],xmm6[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,3]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[0,1,0,3]
; AVX2-ONLY-NEXT: vmovaps 656(%rdi), %xmm13
; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm15 = xmm0[0],xmm13[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm15[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0,1,2,3],ymm7[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd 712(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm15[1],ymm1[1],ymm15[3],ymm1[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm15[1],ymm2[1],ymm15[3],ymm2[3]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm13[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd 520(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm10[1],ymm0[3],ymm10[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm8[1],xmm9[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vbroadcastsd 136(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm12[1],ymm1[3],ymm12[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vbroadcastsd 328(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm11[1],ymm6[3],ymm11[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm2[1],xmm5[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%r8)
-; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm5, (%r8)
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 96(%r9)
-; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%r9)
-; AVX2-ONLY-NEXT: vmovaps %ymm14, (%r9)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%r9)
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm4[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd 136(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm12[1],ymm0[3],ymm12[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm11[1],xmm10[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd 328(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm8[1],ymm3[3],ymm8[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm6[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 96(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 64(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 96(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 64(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 96(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 64(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%r8)
+; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 96(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 64(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%r8)
+; AVX2-ONLY-NEXT: vmovaps %ymm7, 96(%r9)
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%r9)
+; AVX2-ONLY-NEXT: vmovaps %ymm14, 64(%r9)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, (%r9)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rax)
-; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rax)
-; AVX2-ONLY-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm0, (%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 64(%rax)
; AVX2-ONLY-NEXT: vmovaps %ymm13, 96(%rax)
-; AVX2-ONLY-NEXT: addq $520, %rsp # imm = 0x208
+; AVX2-ONLY-NEXT: addq $488, %rsp # imm = 0x1E8
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
;
@@ -2506,9 +2505,17 @@
; AVX1-ONLY-NEXT: vinsertf128 $1, 1504(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vmovaps 1360(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm10[0],xmm12[0]
+; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vmovaps 1360(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm12[0],xmm14[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vinsertf128 $1, 1312(%rdi), %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovaps 1168(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm8[0],xmm9[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm0
@@ -2523,27 +2530,19 @@
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vinsertf128 $1, 544(%rdi), %ymm0, %ymm11
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm14[0],ymm11[0],ymm14[2],ymm11[2]
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm8[0],xmm9[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, 928(%rdi), %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, 544(%rdi), %ymm0, %ymm6
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
-; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 784(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 400(%rdi), %xmm4
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm4[0],xmm5[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, 1312(%rdi), %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vmovaps 864(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, 928(%rdi), %ymm0, %ymm13
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm13[0],ymm3[2],ymm13[2]
-; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 1168(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 832(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 784(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm15 = xmm1[0],xmm2[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2574,8 +2573,12 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm10[1],xmm12[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm12[1],xmm14[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm10[1],ymm11[3],ymm10[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm9[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm13[1],ymm3[3],ymm13[3]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
@@ -2585,10 +2588,6 @@
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm5[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm14[1],ymm11[1],ymm14[3],ymm11[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm8[1],xmm9[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
@@ -2757,14 +2756,14 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 192(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 128(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 64(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 192(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, 160(%rcx)
@@ -2823,14 +2822,14 @@
; AVX2-ONLY-LABEL: load_i64_stride6_vf32:
; AVX2-ONLY: # %bb.0:
; AVX2-ONLY-NEXT: subq $1480, %rsp # imm = 0x5C8
-; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1056(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 1056(%rdi), %ymm6
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm5
@@ -2848,17 +2847,17 @@
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %xmm5
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm5[0],xmm0[0]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm6[0],ymm3[0],ymm6[2],ymm3[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1008(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm0[0]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm2[0],xmm0[0]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm6[0],ymm3[0],ymm6[2],ymm3[2]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,0,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2868,9 +2867,9 @@
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
; AVX2-ONLY-NEXT: vmovaps 1392(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %xmm13
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm13[0],xmm1[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm1[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -2880,40 +2879,40 @@
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm8
; AVX2-ONLY-NEXT: vmovaps 48(%rdi), %xmm9
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm8[0],xmm9[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm8[0],xmm9[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm6
; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm7[0],ymm0[2],ymm7[2]
-; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovaps 432(%rdi), %xmm10
-; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm11
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm11[0],xmm10[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[2],ymm6[2]
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovaps 816(%rdi), %xmm12
-; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %xmm15
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm15[0],xmm12[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovaps 432(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm11
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm11[0],xmm10[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovaps 1248(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovaps 1200(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovaps 816(%rdi), %xmm12
+; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %xmm15
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm15[0],xmm12[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 1248(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
+; AVX2-ONLY-NEXT: vmovaps 1200(%rdi), %xmm13
; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm1[0],xmm13[0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -2931,162 +2930,164 @@
; AVX2-ONLY-NEXT: # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm0 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm0 = xmm3[1],mem[1]
-; AVX2-ONLY-NEXT: vbroadcastsd 1064(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm0 = xmm2[1],mem[1]
+; AVX2-ONLY-NEXT: vbroadcastsd 1064(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = ymm2[1],mem[1],ymm2[3],mem[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm0 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm0 = xmm13[1],mem[1]
-; AVX2-ONLY-NEXT: vbroadcastsd 1448(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm0 = xmm7[1],mem[1]
+; AVX2-ONLY-NEXT: vbroadcastsd 1448(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = ymm2[1],mem[1],ymm2[3],mem[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm8[1],xmm9[1]
-; AVX2-ONLY-NEXT: vbroadcastsd 104(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd 104(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = ymm2[1],mem[1],ymm2[3],mem[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm11[1],xmm10[1]
-; AVX2-ONLY-NEXT: vbroadcastsd 488(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm7[1],ymm3[3],ymm7[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd 488(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm6[1],ymm2[3],ymm6[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm15[1],xmm12[1]
-; AVX2-ONLY-NEXT: vbroadcastsd 872(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm6[1],ymm3[3],ymm6[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd 872(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm4[1],ymm2[3],ymm4[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm13[1]
; AVX2-ONLY-NEXT: vbroadcastsd 1256(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 352(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm9
-; AVX2-ONLY-NEXT: vmovaps 208(%rdi), %xmm10
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm10[0],xmm9[0]
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 208(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 736(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vmovaps 592(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm7[0],xmm8[0]
+; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vmovaps 592(%rdi), %xmm6
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm6[0],xmm10[0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 1120(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovaps 976(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm4[0],xmm6[0]
+; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 976(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm1[0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastsd 1504(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovaps 1360(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm1[0],xmm2[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps 1360(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm3[0],xmm4[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vbroadcastsd 160(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vbroadcastsd 1312(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovaps 1168(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastsd 160(%rdi), %ymm7
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm0[0],ymm7[0],ymm0[2],ymm7[2]
; AVX2-ONLY-NEXT: vmovaps 16(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm5
-; AVX2-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm0[0],xmm5[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm0[0],xmm8[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm7[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vbroadcastsd 544(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
-; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 400(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm3[0],xmm0[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastsd 544(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm15[0],ymm7[0],ymm15[2],ymm7[2]
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %xmm12
+; AVX2-ONLY-NEXT: vmovaps 400(%rdi), %xmm9
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm9[0],xmm12[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm13[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vbroadcastsd 928(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm5[0],ymm11[0],ymm5[2],ymm11[2]
-; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %xmm11
-; AVX2-ONLY-NEXT: vmovaps 784(%rdi), %xmm13
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm15 = xmm13[0],xmm11[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm14[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vbroadcastsd 1312(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm3[0],ymm14[0],ymm3[2],ymm14[2]
-; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %xmm15
-; AVX2-ONLY-NEXT: vmovaps 1168(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm0[0],xmm15[0]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm14[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm10[1],xmm9[1]
-; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm10 = ymm10[1],ymm12[1],ymm10[3],ymm12[3]
-; AVX2-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,2,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm7[1],xmm8[1]
-; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vbroadcastsd 928(%rdi), %ymm13
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
-; AVX2-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm8 = ymm8[0,1,2,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],xmm6[1]
-; AVX2-ONLY-NEXT: vmovaps 1120(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm8[0],ymm13[0],ymm8[2],ymm13[2]
+; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %xmm14
+; AVX2-ONLY-NEXT: vmovaps 784(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm11 = xmm7[0],xmm14[0]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm11 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm13 = ymm13[1],ymm0[1],ymm13[3],ymm0[3]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm6[1],xmm10[1]
+; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm10 = ymm10[1],ymm11[1],ymm10[3],ymm11[3]
+; AVX2-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm10 = ymm10[0,1,2,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm5 = xmm5[1],mem[1]
+; AVX2-ONLY-NEXT: vmovaps 1120(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm7[1],ymm6[3],ymm7[3]
-; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm6[1],ymm0[1],ymm6[3],ymm0[3]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; AVX2-ONLY-NEXT: vmovaps 1504(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm0[1],ymm4[3],ymm0[3]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[0,1,2,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT: vmovaps 1504(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1312(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm4[1],ymm2[3],ymm4[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
+; AVX2-ONLY-NEXT: vmovaps %ymm3, %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[0,1,2,1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm15[1]
-; AVX2-ONLY-NEXT: vmovaps 1312(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm7[1],xmm14[1]
+; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm8[1],ymm2[1],ymm8[3],ymm2[3]
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm13[1],xmm11[1]
-; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm3[1],ymm5[3],ymm3[3]
-; AVX2-ONLY-NEXT: vmovaps %ymm3, %ymm4
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm0 = xmm0[1],mem[1]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm9[1],xmm12[1]
; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm15[1],ymm3[1],ymm15[3],ymm3[3]
; AVX2-ONLY-NEXT: vmovaps %ymm3, %ymm5
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
@@ -3112,7 +3113,8 @@
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm12[0],ymm0[2],ymm12[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
; AVX2-ONLY-NEXT: vmovaps 272(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3130,7 +3132,7 @@
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm11[0],ymm0[2],ymm11[2]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
; AVX2-ONLY-NEXT: vmovaps 656(%rdi), %xmm11
; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %xmm10
@@ -3138,7 +3140,7 @@
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
; AVX2-ONLY-NEXT: vmovaps 848(%rdi), %xmm9
; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %xmm8
@@ -3146,24 +3148,25 @@
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm7[0],ymm0[2],ymm7[2]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovaps 1040(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %xmm12
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm12[0],xmm7[0]
+; AVX2-ONLY-NEXT: vmovaps 1040(%rdi), %xmm12
+; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %xmm6
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm6[0],xmm12[0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
-; AVX2-ONLY-NEXT: vmovaps 1232(%rdi), %xmm6
+; AVX2-ONLY-NEXT: vmovaps 1232(%rdi), %xmm5
; AVX2-ONLY-NEXT: vmovaps 1184(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm4[0],xmm6[0]
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm4[0],xmm5[0]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,1,0,3]
; AVX2-ONLY-NEXT: vmovaps 1424(%rdi), %xmm2
; AVX2-ONLY-NEXT: vmovaps 1376(%rdi), %xmm1
@@ -3200,17 +3203,17 @@
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm8[1],xmm9[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd 1096(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm12[1],xmm7[1]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm7[1],ymm0[3],ymm7[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm6[1],xmm12[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd 1288(%rdi), %ymm0
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],xmm6[1]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm4[1],xmm5[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vbroadcastsd 1480(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = ymm4[1],mem[1],ymm4[3],mem[3]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -3246,14 +3249,14 @@
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rdx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 192(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, 128(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, 64(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 192(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, 224(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, 160(%rcx)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-7.ll
index c745750..d475581 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-7.ll
@@ -146,50 +146,50 @@
; SSE: # %bb.0:
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %r10
-; SSE-NEXT: movapd 208(%rdi), %xmm1
-; SSE-NEXT: movapd 96(%rdi), %xmm0
+; SSE-NEXT: movapd 96(%rdi), %xmm1
+; SSE-NEXT: movapd 208(%rdi), %xmm0
; SSE-NEXT: movapd 144(%rdi), %xmm2
-; SSE-NEXT: movapd 192(%rdi), %xmm4
-; SSE-NEXT: movapd 80(%rdi), %xmm3
+; SSE-NEXT: movapd 80(%rdi), %xmm4
+; SSE-NEXT: movapd 192(%rdi), %xmm3
; SSE-NEXT: movapd 128(%rdi), %xmm5
-; SSE-NEXT: movapd 176(%rdi), %xmm8
-; SSE-NEXT: movapd 64(%rdi), %xmm7
+; SSE-NEXT: movapd 64(%rdi), %xmm8
+; SSE-NEXT: movapd 176(%rdi), %xmm7
; SSE-NEXT: movapd (%rdi), %xmm10
; SSE-NEXT: movapd 16(%rdi), %xmm9
; SSE-NEXT: movapd 32(%rdi), %xmm6
-; SSE-NEXT: movapd 48(%rdi), %xmm11
-; SSE-NEXT: movapd 112(%rdi), %xmm12
+; SSE-NEXT: movapd 48(%rdi), %xmm12
+; SSE-NEXT: movapd 112(%rdi), %xmm11
; SSE-NEXT: movapd 160(%rdi), %xmm13
; SSE-NEXT: movapd %xmm13, %xmm14
-; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm12[0],xmm14[1]
-; SSE-NEXT: movapd %xmm11, %xmm15
+; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm11[0],xmm14[1]
+; SSE-NEXT: movapd %xmm12, %xmm15
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm10[0],xmm15[1]
-; SSE-NEXT: shufpd {{.*#+}} xmm10 = xmm10[1],xmm7[0]
-; SSE-NEXT: shufpd {{.*#+}} xmm12 = xmm12[1],xmm8[0]
-; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm9[0],xmm7[1]
-; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm5[0],xmm8[1]
-; SSE-NEXT: shufpd {{.*#+}} xmm9 = xmm9[1],xmm3[0]
-; SSE-NEXT: shufpd {{.*#+}} xmm5 = xmm5[1],xmm4[0]
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm6[0],xmm3[1]
-; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
-; SSE-NEXT: shufpd {{.*#+}} xmm6 = xmm6[1],xmm0[0]
-; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1],xmm1[0]
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm11[0],xmm0[1]
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm13[0],xmm1[1]
-; SSE-NEXT: movapd %xmm14, 16(%rsi)
+; SSE-NEXT: shufpd {{.*#+}} xmm11 = xmm11[1],xmm7[0]
+; SSE-NEXT: shufpd {{.*#+}} xmm10 = xmm10[1],xmm8[0]
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm5[0],xmm7[1]
+; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm9[0],xmm8[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm5 = xmm5[1],xmm3[0]
+; SSE-NEXT: shufpd {{.*#+}} xmm9 = xmm9[1],xmm4[0]
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
+; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm6[0],xmm4[1]
+; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1],xmm0[0]
+; SSE-NEXT: shufpd {{.*#+}} xmm6 = xmm6[1],xmm1[0]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm13[0],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm12[0],xmm1[1]
; SSE-NEXT: movapd %xmm15, (%rsi)
-; SSE-NEXT: movapd %xmm12, 16(%rdx)
+; SSE-NEXT: movapd %xmm14, 16(%rsi)
; SSE-NEXT: movapd %xmm10, (%rdx)
-; SSE-NEXT: movapd %xmm8, 16(%rcx)
-; SSE-NEXT: movapd %xmm7, (%rcx)
-; SSE-NEXT: movapd %xmm5, 16(%r8)
+; SSE-NEXT: movapd %xmm11, 16(%rdx)
+; SSE-NEXT: movapd %xmm8, (%rcx)
+; SSE-NEXT: movapd %xmm7, 16(%rcx)
; SSE-NEXT: movapd %xmm9, (%r8)
-; SSE-NEXT: movapd %xmm4, 16(%r9)
-; SSE-NEXT: movapd %xmm3, (%r9)
-; SSE-NEXT: movapd %xmm2, 16(%r10)
+; SSE-NEXT: movapd %xmm5, 16(%r8)
+; SSE-NEXT: movapd %xmm4, (%r9)
+; SSE-NEXT: movapd %xmm3, 16(%r9)
; SSE-NEXT: movapd %xmm6, (%r10)
-; SSE-NEXT: movapd %xmm1, 16(%rax)
-; SSE-NEXT: movapd %xmm0, (%rax)
+; SSE-NEXT: movapd %xmm2, 16(%r10)
+; SSE-NEXT: movapd %xmm1, (%rax)
+; SSE-NEXT: movapd %xmm0, 16(%rax)
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i64_stride7_vf4:
@@ -652,208 +652,208 @@
;
; AVX1-ONLY-LABEL: load_i64_stride7_vf8:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovapd 320(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, 384(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vmovapd 272(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm3 = mem[0],xmm1[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm3 = xmm10[0],xmm1[1]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, 384(%rdi), %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm0, %ymm3
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm7[0,1,2],ymm3[3]
-; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vmovapd 272(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm14 = xmm13[0],xmm4[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm13 = mem[0],xmm4[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm13[0,1],ymm5[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm8[0],ymm6[3],ymm8[2]
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm14 = mem[8,9,10,11,12,13,14,15],xmm12[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm14[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[0],ymm10[0],ymm7[3],ymm10[2]
-; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm13 = xmm13[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm13[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm14[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = xmm10[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[0],ymm9[0],ymm7[3],ymm9[2]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = mem[8,9,10,11,12,13,14,15],xmm12[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm10[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm13[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT: vmovapd 240(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm14 = xmm13[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm14[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm15
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm15[0,1,2],ymm9[3]
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm12 = xmm11[0,1,2,3],xmm12[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm12[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm14[0,1,2],ymm10[3]
-; AVX1-ONLY-NEXT: vmovapd 240(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm15 = xmm14[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm15[0,1],ymm10[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm11[1],ymm9[0],ymm11[2],ymm9[2]
-; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm13 = xmm13[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm13[2,3]
-; AVX1-ONLY-NEXT: vmovapd 304(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm14[1],ymm13[0],ymm14[2],ymm13[2]
-; AVX1-ONLY-NEXT: vmovdqa 416(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm12 = xmm12[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm14[0,1],ymm12[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm12[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT: vmovapd 304(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm13[1],ymm12[0],ymm13[2],ymm12[2]
+; AVX1-ONLY-NEXT: vmovdqa 416(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = xmm10[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm13[0,1],ymm10[2,3]
+; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm11[1],ymm13[0],ymm11[2],ymm13[2]
+; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm14 = xmm14[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm14[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm15
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm15[0,1,2],ymm14[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm9 = mem[0],xmm9[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm14[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovapd 256(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm13 = xmm1[0],xmm13[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm12 = xmm1[0],xmm12[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm14
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm13 = mem[0],xmm13[1]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm13[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm13
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm15[0],ymm13[0],ymm15[3],ymm13[2]
-; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = mem[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm15[2,3]
-; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm14[0],ymm15[0],ymm14[3],ymm15[2]
; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm0
; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm1 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm14[0],ymm15[0],ymm14[3],ymm15[2]
+; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = mem[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm14[2,3]
; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm13 = mem[0,1,2],ymm13[3]
-; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm14 = xmm14[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm14[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm13[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm15[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rsi)
-; AVX1-ONLY-NEXT: vmovapd %ymm7, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, (%rdx)
-; AVX1-ONLY-NEXT: vmovapd %ymm10, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm8, (%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm12, 32(%r8)
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm4 = xmm4[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, (%rdx)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, (%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm8, 32(%rcx)
; AVX1-ONLY-NEXT: vmovapd %ymm11, (%r8)
-; AVX1-ONLY-NEXT: vmovapd %ymm2, 32(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm9, (%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm10, 32(%r8)
+; AVX1-ONLY-NEXT: vmovapd %ymm2, (%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm12, 32(%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovapd %ymm1, 32(%rax)
; AVX1-ONLY-NEXT: vmovapd %ymm5, (%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm1, 32(%rax)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT: vmovapd %ymm3, (%rax)
; AVX1-ONLY-NEXT: vmovapd %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm13, (%rax)
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-ONLY-LABEL: load_i64_stride7_vf8:
; AVX2-ONLY: # %bb.0:
-; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovdqa 352(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 48(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = mem[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3,4,5],ymm5[6,7]
; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %xmm13
-; AVX2-ONLY-NEXT: vmovdqa 272(%rdi), %xmm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm14 = xmm13[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vmovdqa 272(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm13[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm5
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 48(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm14 = mem[0,1],xmm5[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm7 = ymm7[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %xmm14
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm15 = mem[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %xmm14
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm13 = xmm13[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm7[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm8 = ymm8[8,9,10,11,12,13,14,15],ymm11[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm11[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %xmm15
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm13 = xmm13[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm13[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq 128(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm9 = ymm13[1],ymm9[1],ymm13[3],ymm9[3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm13 = mem[0,1],xmm14[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm13[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq 352(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm11 = ymm13[1],ymm11[1],ymm13[3],ymm11[3]
-; AVX2-ONLY-NEXT: vmovdqa 240(%rdi), %xmm13
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm13 = xmm13[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %xmm13
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm14 = mem[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm14[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq 352(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm9 = ymm14[1],ymm9[1],ymm14[3],ymm9[3]
+; AVX2-ONLY-NEXT: vmovdqa 240(%rdi), %xmm14
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm14[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq 128(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm11 = ymm14[1],ymm11[1],ymm14[3],ymm11[3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm13 = mem[0,1],xmm13[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm13[0,1,2,3],ymm11[4,5,6,7]
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm12 = mem[8,9,10,11,12,13,14,15],ymm12[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm12[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %xmm13
+; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %xmm13
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm14 = mem[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm12 = ymm12[2,3],ymm14[2,3]
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = mem[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %xmm14
+; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %xmm14
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm15 = mem[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm15, %ymm0, %ymm15
; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm10 = ymm10[2,3],ymm15[2,3]
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm13, %ymm0, %ymm13
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm4[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %xmm15
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm15 = xmm15[0,1],mem[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm15[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 352(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm6 = xmm0[0,1],mem[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm14[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm4 = ymm4[8,9,10,11,12,13,14,15],ymm14[0,1,2,3,4,5,6,7],ymm4[24,25,26,27,28,29,30,31],ymm14[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm3[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %xmm15
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm15[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm0
+; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %xmm6
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm6 = xmm6[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm3 = ymm3[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm3[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm15 = xmm15[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm15[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm15
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm14 = ymm14[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23]
; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %xmm7
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm15 = ymm15[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm15[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm14 = ymm15[1],ymm14[1],ymm15[3],ymm14[3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm14[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq %xmm3, %ymm3
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm3 = ymm3[1],ymm7[1],ymm3[3],ymm7[3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rsi)
-; AVX2-ONLY-NEXT: vmovdqa %ymm8, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rdx)
-; AVX2-ONLY-NEXT: vmovdqa %ymm11, 32(%rcx)
-; AVX2-ONLY-NEXT: vmovdqa %ymm9, (%rcx)
-; AVX2-ONLY-NEXT: vmovdqa %ymm10, 32(%r8)
-; AVX2-ONLY-NEXT: vmovdqa %ymm12, (%r8)
-; AVX2-ONLY-NEXT: vmovdqa %ymm6, 32(%r9)
-; AVX2-ONLY-NEXT: vmovdqa %ymm13, (%r9)
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm14[4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm6 = ymm14[1],ymm6[1],ymm14[3],ymm6[3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq %xmm4, %ymm2
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm2 = ymm2[1],ymm15[1],ymm2[3],ymm15[3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm5[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovdqa %ymm8, (%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovdqa %ymm11, (%rcx)
+; AVX2-ONLY-NEXT: vmovdqa %ymm9, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovdqa %ymm10, (%r8)
+; AVX2-ONLY-NEXT: vmovdqa %ymm12, 32(%r8)
+; AVX2-ONLY-NEXT: vmovdqa %ymm1, (%r9)
+; AVX2-ONLY-NEXT: vmovdqa %ymm13, 32(%r9)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm0, 32(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm4, (%rax)
+; AVX2-ONLY-NEXT: vmovdqa %ymm7, (%rax)
+; AVX2-ONLY-NEXT: vmovdqa %ymm3, 32(%rax)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm1, 32(%rax)
; AVX2-ONLY-NEXT: vmovdqa %ymm2, (%rax)
+; AVX2-ONLY-NEXT: vmovdqa %ymm0, 32(%rax)
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
;
@@ -1106,52 +1106,52 @@
; SSE-NEXT: movapd 192(%rdi), %xmm3
; SSE-NEXT: movapd 80(%rdi), %xmm4
; SSE-NEXT: movapd 128(%rdi), %xmm5
-; SSE-NEXT: movapd 64(%rdi), %xmm8
-; SSE-NEXT: movapd 176(%rdi), %xmm9
+; SSE-NEXT: movapd 176(%rdi), %xmm7
+; SSE-NEXT: movapd 64(%rdi), %xmm9
; SSE-NEXT: movapd (%rdi), %xmm10
-; SSE-NEXT: movapd 16(%rdi), %xmm7
+; SSE-NEXT: movapd 16(%rdi), %xmm8
; SSE-NEXT: movapd 32(%rdi), %xmm6
-; SSE-NEXT: movapd 48(%rdi), %xmm14
-; SSE-NEXT: movapd 224(%rdi), %xmm11
-; SSE-NEXT: movapd 112(%rdi), %xmm12
-; SSE-NEXT: movapd 160(%rdi), %xmm15
-; SSE-NEXT: movapd %xmm14, %xmm13
+; SSE-NEXT: movapd 48(%rdi), %xmm15
+; SSE-NEXT: movapd 112(%rdi), %xmm11
+; SSE-NEXT: movapd 160(%rdi), %xmm14
+; SSE-NEXT: movapd 224(%rdi), %xmm12
+; SSE-NEXT: movapd %xmm15, %xmm13
; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm10[0],xmm13[1]
; SSE-NEXT: movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufpd {{.*#+}} xmm10 = xmm10[1],xmm8[0]
+; SSE-NEXT: shufpd {{.*#+}} xmm10 = xmm10[1],xmm9[0]
; SSE-NEXT: movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm7[0],xmm8[1]
+; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm8[0],xmm9[1]
+; SSE-NEXT: movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufpd {{.*#+}} xmm8 = xmm8[1],xmm4[0]
; SSE-NEXT: movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufpd {{.*#+}} xmm7 = xmm7[1],xmm4[0]
-; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm4 = xmm6[0],xmm4[1]
; SSE-NEXT: movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufpd {{.*#+}} xmm6 = xmm6[1],xmm1[0]
; SSE-NEXT: movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm15[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd %xmm15, %xmm1
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm12[0],xmm1[1]
+; SSE-NEXT: movapd %xmm14, %xmm1
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm11[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufpd {{.*#+}} xmm12 = xmm12[1],xmm9[0]
-; SSE-NEXT: movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm5[0],xmm9[1]
-; SSE-NEXT: movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufpd {{.*#+}} xmm11 = xmm11[1],xmm7[0]
+; SSE-NEXT: movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm5[0],xmm7[1]
+; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufpd {{.*#+}} xmm5 = xmm5[1],xmm3[0]
; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
; SSE-NEXT: movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1],xmm0[0]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm15[0],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movapd 272(%rdi), %xmm0
; SSE-NEXT: movapd %xmm0, %xmm1
-; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm11[0],xmm1[1]
+; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm12[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movapd 288(%rdi), %xmm1
-; SSE-NEXT: shufpd {{.*#+}} xmm11 = xmm11[1],xmm1[0]
-; SSE-NEXT: movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufpd {{.*#+}} xmm12 = xmm12[1],xmm1[0]
+; SSE-NEXT: movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movapd 240(%rdi), %xmm2
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1210,15 +1210,15 @@
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd 560(%rdi), %xmm13
+; SSE-NEXT: movapd 560(%rdi), %xmm11
; SSE-NEXT: movapd 608(%rdi), %xmm2
; SSE-NEXT: movapd %xmm2, %xmm0
-; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm13[0],xmm0[1]
+; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm11[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd 624(%rdi), %xmm14
-; SSE-NEXT: shufpd {{.*#+}} xmm13 = xmm13[1],xmm14[0]
+; SSE-NEXT: movapd 624(%rdi), %xmm15
+; SSE-NEXT: shufpd {{.*#+}} xmm11 = xmm11[1],xmm15[0]
; SSE-NEXT: movapd 576(%rdi), %xmm1
-; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm1[0],xmm14[1]
+; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
; SSE-NEXT: movapd 640(%rdi), %xmm0
; SSE-NEXT: shufpd {{.*#+}} xmm1 = xmm1[1],xmm0[0]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1230,79 +1230,79 @@
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movapd 672(%rdi), %xmm2
+; SSE-NEXT: movapd 672(%rdi), %xmm5
; SSE-NEXT: movapd 720(%rdi), %xmm4
-; SSE-NEXT: movapd %xmm4, %xmm8
-; SSE-NEXT: movsd {{.*#+}} xmm8 = xmm2[0],xmm8[1]
-; SSE-NEXT: movapd 736(%rdi), %xmm6
-; SSE-NEXT: shufpd {{.*#+}} xmm2 = xmm2[1],xmm6[0]
-; SSE-NEXT: movapd 688(%rdi), %xmm9
-; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm9[0],xmm6[1]
-; SSE-NEXT: movapd 752(%rdi), %xmm12
-; SSE-NEXT: shufpd {{.*#+}} xmm9 = xmm9[1],xmm12[0]
-; SSE-NEXT: movapd 704(%rdi), %xmm15
-; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm15[0],xmm12[1]
+; SSE-NEXT: movapd %xmm4, %xmm12
+; SSE-NEXT: movsd {{.*#+}} xmm12 = xmm5[0],xmm12[1]
+; SSE-NEXT: movapd 736(%rdi), %xmm7
+; SSE-NEXT: shufpd {{.*#+}} xmm5 = xmm5[1],xmm7[0]
+; SSE-NEXT: movapd 688(%rdi), %xmm8
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm8[0],xmm7[1]
+; SSE-NEXT: movapd 752(%rdi), %xmm13
+; SSE-NEXT: shufpd {{.*#+}} xmm8 = xmm8[1],xmm13[0]
+; SSE-NEXT: movapd 704(%rdi), %xmm14
+; SSE-NEXT: movsd {{.*#+}} xmm13 = xmm14[0],xmm13[1]
; SSE-NEXT: movapd 768(%rdi), %xmm0
-; SSE-NEXT: shufpd {{.*#+}} xmm15 = xmm15[1],xmm0[0]
+; SSE-NEXT: shufpd {{.*#+}} xmm14 = xmm14[1],xmm0[0]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
; SSE-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movapd 784(%rdi), %xmm4
-; SSE-NEXT: movapd 832(%rdi), %xmm11
-; SSE-NEXT: movapd %xmm11, %xmm5
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; SSE-NEXT: movapd 832(%rdi), %xmm10
+; SSE-NEXT: movapd %xmm10, %xmm3
+; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm4[0],xmm3[1]
; SSE-NEXT: movapd 848(%rdi), %xmm0
; SSE-NEXT: shufpd {{.*#+}} xmm4 = xmm4[1],xmm0[0]
; SSE-NEXT: movapd 800(%rdi), %xmm1
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT: movapd 864(%rdi), %xmm3
-; SSE-NEXT: shufpd {{.*#+}} xmm1 = xmm1[1],xmm3[0]
-; SSE-NEXT: movapd 816(%rdi), %xmm7
-; SSE-NEXT: movsd {{.*#+}} xmm3 = xmm7[0],xmm3[1]
-; SSE-NEXT: movapd 880(%rdi), %xmm10
-; SSE-NEXT: shufpd {{.*#+}} xmm7 = xmm7[1],xmm10[0]
-; SSE-NEXT: movsd {{.*#+}} xmm10 = xmm11[0],xmm10[1]
-; SSE-NEXT: movapd %xmm8, 96(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movaps %xmm8, 32(%rsi)
-; SSE-NEXT: movapd %xmm5, 112(%rsi)
-; SSE-NEXT: movaps (%rsp), %xmm5 # 16-byte Reload
-; SSE-NEXT: movaps %xmm5, 48(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movaps %xmm5, 64(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movaps %xmm5, (%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movaps %xmm5, 80(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: movaps %xmm5, 16(%rsi)
-; SSE-NEXT: movapd %xmm2, 96(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 32(%rdx)
+; SSE-NEXT: movapd 864(%rdi), %xmm2
+; SSE-NEXT: shufpd {{.*#+}} xmm1 = xmm1[1],xmm2[0]
+; SSE-NEXT: movapd 816(%rdi), %xmm6
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm6[0],xmm2[1]
+; SSE-NEXT: movapd 880(%rdi), %xmm9
+; SSE-NEXT: shufpd {{.*#+}} xmm6 = xmm6[1],xmm9[0]
+; SSE-NEXT: movsd {{.*#+}} xmm9 = xmm10[0],xmm9[1]
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movaps %xmm10, 80(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movaps %xmm10, 16(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movaps %xmm10, 64(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movaps %xmm10, (%rsi)
+; SSE-NEXT: movapd %xmm3, 112(%rsi)
+; SSE-NEXT: movaps (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 48(%rsi)
+; SSE-NEXT: movapd %xmm12, 96(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 32(%rsi)
+; SSE-NEXT: movapd %xmm11, 80(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 16(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 64(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, (%rdx)
; SSE-NEXT: movapd %xmm4, 112(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 48(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 64(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, (%rdx)
-; SSE-NEXT: movapd %xmm13, 80(%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 16(%rdx)
-; SSE-NEXT: movapd %xmm6, 96(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 48(%rdx)
+; SSE-NEXT: movapd %xmm5, 96(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 16(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, (%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 48(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 32(%rcx)
+; SSE-NEXT: movapd %xmm15, 80(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 64(%rcx)
; SSE-NEXT: movapd %xmm0, 112(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 64(%rcx)
-; SSE-NEXT: movapd %xmm14, 80(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 16(%rcx)
+; SSE-NEXT: movapd %xmm7, 96(%rcx)
; SSE-NEXT: movapd %xmm1, 112(%r8)
-; SSE-NEXT: movapd %xmm9, 96(%r8)
+; SSE-NEXT: movapd %xmm8, 96(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 80(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -1315,8 +1315,8 @@
; SSE-NEXT: movaps %xmm0, 16(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%r8)
-; SSE-NEXT: movapd %xmm3, 112(%r9)
-; SSE-NEXT: movapd %xmm12, 96(%r9)
+; SSE-NEXT: movapd %xmm2, 112(%r9)
+; SSE-NEXT: movapd %xmm13, 96(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 80(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -1330,8 +1330,8 @@
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movapd %xmm7, 112(%rax)
-; SSE-NEXT: movapd %xmm15, 96(%rax)
+; SSE-NEXT: movapd %xmm6, 112(%rax)
+; SSE-NEXT: movapd %xmm14, 96(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 80(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -1345,7 +1345,7 @@
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, (%rax)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movapd %xmm10, 112(%rax)
+; SSE-NEXT: movapd %xmm9, 112(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 96(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -1417,181 +1417,181 @@
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[3],ymm4[2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[3],ymm2[2]
-; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = mem[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm6[0],ymm1[3],ymm6[2]
+; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = mem[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[2]
-; AVX1-ONLY-NEXT: vmovdqa 512(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm7[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[3],ymm9[2]
+; AVX1-ONLY-NEXT: vmovdqa 512(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm1 = xmm7[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm3[0,1,2],ymm5[3]
-; AVX1-ONLY-NEXT: vmovapd 240(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm7 = xmm3[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm7[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 800(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3]
-; AVX1-ONLY-NEXT: vmovapd 688(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm9 = xmm5[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm9[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm9[0,1,2],ymm2[3]
-; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm9[0,1,2,3],xmm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 576(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm6[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovdqa 464(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm8 = xmm6[0,1,2,3],xmm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm8[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 752(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm5[1],ymm15[0],ymm5[2],ymm15[2]
-; AVX1-ONLY-NEXT: vmovdqa 864(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm8[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 304(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[1],ymm7[0],ymm3[2],ymm7[2]
-; AVX1-ONLY-NEXT: vmovdqa 416(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = xmm0[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT: vmovapd 688(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm1 = xmm4[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm9[1],ymm11[0],ymm9[2],ymm11[2]
-; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = xmm4[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm8[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT: vmovapd 240(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm1 = xmm5[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm6[0,1,2,3],xmm8[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 576(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT: vmovdqa 464(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm9 = xmm12[0,1,2,3],xmm10[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm9[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 752(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm4[1],ymm11[0],ymm4[2],ymm11[2]
+; AVX1-ONLY-NEXT: vmovdqa 864(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm9[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 304(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[1],ymm7[0],ymm5[2],ymm7[2]
+; AVX1-ONLY-NEXT: vmovdqa 416(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm9[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm5[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm6[1],ymm8[0],ymm6[2],ymm8[2]
+; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm3[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm5[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 528(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[1],ymm4[0],ymm6[2],ymm4[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm12[1],ymm4[0],ymm12[2],ymm4[2]
; AVX1-ONLY-NEXT: vmovdqa 640(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = xmm10[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm6[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm1 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm10[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovapd 256(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm7 = xmm6[0],xmm7[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm7[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovapd 800(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm5[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm5
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm5[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm7 = xmm2[0],xmm15[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm7[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovapd 256(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm7 = xmm3[0],xmm7[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm7[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovapd 800(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm10[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm11 = xmm7[0],xmm11[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm11[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm9[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm6 = mem[0],xmm8[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm6[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovapd 576(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm4 = xmm15[0],xmm4[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm4[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm3 = mem[0],xmm11[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm3[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm10[0],ymm3[0],ymm10[3],ymm3[2]
-; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = xmm6[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 864(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm1[0],ymm5[3],ymm1[2]
+; AVX1-ONLY-NEXT: vmovapd 576(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm4 = xmm8[0],xmm4[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm4[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm4[0],ymm5[3],ymm4[2]
+; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm3[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm3[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vmovapd 864(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm10[0],ymm3[0],ymm10[3],ymm3[2]
; AVX1-ONLY-NEXT: vmovdqa 768(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm2[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vmovdqa 544(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm15[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovapd 640(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[0],ymm15[0],ymm7[3],ymm15[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm7
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm7[0],ymm4[3],ymm7[2]
-; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm14 = mem[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm14[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm10[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 544(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = xmm8[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 640(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm10[0],ymm1[3],ymm10[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm9[0],ymm1[0],ymm9[3],ymm1[2]
+; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm15 = mem[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm15[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm4 = mem[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3],xmm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3]
; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = mem[0,1,2],ymm3[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = mem[0,1,2,3],xmm10[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm10[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2,3],xmm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2],ymm7[3]
-; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm7 = xmm7[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2],ymm15[3]
-; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm10 = xmm10[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm10[0,1],ymm7[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm12, 64(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, (%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 32(%r8)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm10 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm10, 96(%r8)
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm1 = xmm1[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2],ymm10[3]
+; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm3 = xmm3[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm13, 64(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, (%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 32(%r8)
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%r8)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 64(%r9)
; AVX1-ONLY-NEXT: vmovapd %ymm11, (%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm8, 64(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm9, 96(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm13, 32(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm12, 96(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm14, 32(%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovapd %ymm4, (%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm2, 64(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm5, 96(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 32(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, (%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm8, 64(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, 96(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm5, 32(%rax)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovapd %ymm7, 64(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm1, (%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm0, 96(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm3, 32(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm1, 64(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm0, (%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm15, 96(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm2, 32(%rax)
; AVX1-ONLY-NEXT: addq $552, %rsp # imm = 0x228
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
@@ -1618,8 +1618,9 @@
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm2[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vmovdqa 672(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovdqa 720(%rdi), %xmm13
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm6[0,1],xmm13[2,3]
+; AVX2-ONLY-NEXT: vmovdqa 720(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm6[0,1],xmm7[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm7[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %xmm4
@@ -1636,9 +1637,8 @@
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0,1,2,3,4,5],ymm4[6,7]
; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovdqa 496(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm7[0,1],xmm8[2,3]
+; AVX2-ONLY-NEXT: vmovdqa 496(%rdi), %xmm12
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm7[0,1],xmm12[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm8[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %xmm4
@@ -1665,18 +1665,18 @@
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastq 352(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm5[1],ymm0[3],ymm5[3]
-; AVX2-ONLY-NEXT: vmovdqa 240(%rdi), %xmm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpbroadcastq 800(%rdi), %ymm0
; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm4[1],ymm0[3],ymm4[3]
; AVX2-ONLY-NEXT: vmovdqa 688(%rdi), %xmm4
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpbroadcastq 352(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm5[1],ymm0[3],ymm5[3]
+; AVX2-ONLY-NEXT: vmovdqa 240(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpbroadcastq 128(%rdi), %ymm0
; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = mem[0,1],xmm3[2,3]
@@ -1703,8 +1703,8 @@
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm0
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm0 = mem[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %xmm11
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = mem[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],ymm4[2,3]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1715,112 +1715,112 @@
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
; AVX2-ONLY-NEXT: vperm2i128 {{.*#+}} ymm15 = ymm0[2,3],ymm5[2,3]
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vmovdqa 352(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm8[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vmovdqa 352(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %xmm9
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm9[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vmovdqa 800(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 704(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm7[0,1],mem[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm12 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vmovdqa 576(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 480(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm4[0,1],mem[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 800(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 704(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm8[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm0
+; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %xmm5
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm6 = ymm6[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],ymm6[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = xmm8[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm0
+; AVX2-ONLY-NEXT: vmovdqa 576(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 480(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm14 = xmm0[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm14[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm7 = ymm7[8,9,10,11,12,13,14,15],ymm14[0,1,2,3,4,5,6,7],ymm7[24,25,26,27,28,29,30,31],ymm14[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm9 = xmm9[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm9[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 864(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm6 = ymm6[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm6[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vmovdqa 768(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = xmm8[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm8[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 864(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm3 = ymm3[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm3[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 768(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm4[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 640(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm9 = ymm9[8,9,10,11,12,13,14,15],ymm4[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm4[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm3[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm11 = ymm11[8,9,10,11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %xmm14
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm14 = mem[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm14[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm2 = ymm14[1],ymm2[1],ymm14[3],ymm2[3]
-; AVX2-ONLY-NEXT: vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm0 = mem[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm14 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
-; AVX2-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm1[1],ymm4[1],ymm1[3],ymm4[3]
+; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 640(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm1[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm10[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm10[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %xmm13
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm13 = mem[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm13[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm13 = ymm13[1],ymm14[1],ymm13[3],ymm14[3]
+; AVX2-ONLY-NEXT: vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm2 = mem[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm14 = ymm2[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm2 = ymm2[1],ymm9[1],ymm2[3],ymm9[3]
+; AVX2-ONLY-NEXT: vpblendd $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm3 = mem[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
; AVX2-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 64(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 96(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 64(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 96(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 64(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 96(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rcx)
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm2 = ymm2[1],ymm8[1],ymm2[3],ymm8[3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm12[0,1],mem[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 96(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 96(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 96(%rcx)
; AVX2-ONLY-NEXT: vmovdqa %ymm15, 64(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 96(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, (%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 96(%r8)
+; AVX2-ONLY-NEXT: vmovdqa %ymm4, 64(%r9)
; AVX2-ONLY-NEXT: vmovdqa %ymm5, (%r9)
-; AVX2-ONLY-NEXT: vmovdqa %ymm10, 64(%r9)
-; AVX2-ONLY-NEXT: vmovdqa %ymm12, 96(%r9)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%r9)
+; AVX2-ONLY-NEXT: vmovdqa %ymm11, 96(%r9)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%r9)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm11, (%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm9, 64(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm7, 96(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm6, 32(%rax)
-; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT: vmovdqa %ymm10, (%rax)
; AVX2-ONLY-NEXT: vmovdqa %ymm1, 64(%rax)
+; AVX2-ONLY-NEXT: vmovdqa %ymm6, 96(%rax)
+; AVX2-ONLY-NEXT: vmovdqa %ymm7, 32(%rax)
+; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; AVX2-ONLY-NEXT: vmovdqa %ymm2, 64(%rax)
; AVX2-ONLY-NEXT: vmovdqa %ymm0, (%rax)
; AVX2-ONLY-NEXT: vmovdqa %ymm13, 96(%rax)
; AVX2-ONLY-NEXT: vmovdqa %ymm14, 32(%rax)
@@ -2779,492 +2779,492 @@
; AVX1-ONLY-NEXT: subq $1720, %rsp # imm = 0x6B8
; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 320(%rdi), %ymm10
; AVX1-ONLY-NEXT: vinsertf128 $1, 384(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm3[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, 832(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps 720(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm9[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovaps 672(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 720(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm4[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, 1280(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovaps 1120(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 1168(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm6[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1168(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %ymm5
; AVX1-ONLY-NEXT: vinsertf128 $1, 1728(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps 1568(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vmovaps 1616(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm14[0,1],xmm2[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovaps 1568(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovaps 1616(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm7[0,1],xmm0[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm14
; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm8[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm2 = mem[0],xmm1[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm14[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm8 = mem[0],xmm0[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm8[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 544(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vmovapd 544(%rdi), %ymm9
; AVX1-ONLY-NEXT: vinsertf128 $1, 608(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm9[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovapd 448(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vmovapd 496(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm2 = xmm11[0],xmm1[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 496(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm8 = xmm11[0],xmm0[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm8[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 992(%rdi), %ymm12
; AVX1-ONLY-NEXT: vinsertf128 $1, 1056(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm12[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 896(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovapd 944(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm4 = xmm2[0],xmm1[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm12[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd 896(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vmovapd 944(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm8 = xmm13[0],xmm0[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm8[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1440(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovapd 1440(%rdi), %ymm15
; AVX1-ONLY-NEXT: vinsertf128 $1, 1504(%rdi), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm4[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 1344(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm15[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd 1344(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovapd 1392(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm13 = xmm7[0],xmm0[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm13[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm3[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm10[0],ymm0[0],ymm10[3],ymm0[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 736(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm1 = xmm9[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovapd 832(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm9[0],ymm3[3],ymm9[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 1184(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm6[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm3[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 384(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm10[0],ymm8[0],ymm10[3],ymm8[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 736(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm4[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 832(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[3],ymm3[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 1184(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm6[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovapd 1280(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm1[0],ymm6[3],ymm1[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 1632(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm14[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovapd 1728(%rdi), %ymm13
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm13[0],ymm5[3],ymm13[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm8[0],ymm3[0],ymm8[3],ymm3[2]
-; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = mem[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm15[0],ymm10[0],ymm15[3],ymm10[2]
-; AVX1-ONLY-NEXT: vmovdqa 512(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = xmm11[8,9,10,11,12,13,14,15],xmm14[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1056(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm12[0],ymm5[0],ymm12[3],ymm5[2]
-; AVX1-ONLY-NEXT: vmovdqa 960(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0],ymm1[0],ymm4[3],ymm1[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 1632(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm7[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 1728(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm4[0],ymm5[3],ymm4[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 160(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm14[0],ymm5[0],ymm14[3],ymm5[2]
+; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = mem[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 608(%rdi), %ymm14
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm9[0],ymm14[0],ymm9[3],ymm14[2]
+; AVX1-ONLY-NEXT: vmovdqa 512(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = xmm11[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 1056(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm12[0],ymm11[0],ymm12[3],ymm11[2]
+; AVX1-ONLY-NEXT: vmovdqa 960(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = xmm13[8,9,10,11,12,13,14,15],xmm12[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 1504(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm15[0],ymm13[0],ymm15[3],ymm13[2]
+; AVX1-ONLY-NEXT: vmovdqa 1408(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1504(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm4[0],ymm8[0],ymm4[3],ymm8[2]
-; AVX1-ONLY-NEXT: vmovdqa 1408(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = xmm7[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT: vmovapd 240(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm8 = xmm9[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm8[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 240(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm2 = xmm12[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 800(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3]
-; AVX1-ONLY-NEXT: vmovapd 688(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm2 = xmm11[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 1248(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovapd 1136(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm1 = xmm7[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1696(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm13[3]
-; AVX1-ONLY-NEXT: vmovapd 1584(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm1 = xmm13[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 800(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm2[0,1,2],ymm3[3]
+; AVX1-ONLY-NEXT: vmovapd 688(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm15 = xmm2[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm15[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 1136(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm15 = xmm1[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm15[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1696(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT: vmovapd 1584(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm4 = xmm15[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT: vmovdqa 1360(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm13[0,1,2,3],xmm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm3[3]
-; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm10[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 576(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm10[3]
-; AVX1-ONLY-NEXT: vmovdqa 464(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm10[0,1,2,3],xmm14[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 576(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT: vmovdqa 464(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm14[0,1,2,3],xmm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 1024(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT: vmovdqa 1024(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm0[0,1,2],ymm11[3]
; AVX1-ONLY-NEXT: vmovdqa 912(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0,1,2,3],xmm15[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 1472(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm8[3]
-; AVX1-ONLY-NEXT: vmovdqa 1360(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm2[0,1,2,3],xmm4[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 304(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm12[1],ymm8[0],ymm12[2],ymm8[2]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm0[0,1,2,3],xmm12[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 304(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm9[1],ymm7[0],ymm9[2],ymm7[2]
; AVX1-ONLY-NEXT: vmovdqa 416(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[8,9,10,11,12,13,14,15],xmm12[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = xmm6[8,9,10,11,12,13,14,15],xmm12[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 752(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm11[1],ymm5[0],ymm11[2],ymm5[2]
-; AVX1-ONLY-NEXT: vmovdqa 864(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1200(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm7[1],ymm1[0],ymm7[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovdqa 1312(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm9[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1648(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm13[1],ymm1[0],ymm13[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovdqa 1760(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1424(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm2[1],ymm9[0],ymm2[2],ymm9[2]
-; AVX1-ONLY-NEXT: vmovdqa 1536(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm15[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1],ymm5[0],ymm2[2],ymm5[2]
+; AVX1-ONLY-NEXT: vmovdqa 864(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = xmm8[8,9,10,11,12,13,14,15],xmm9[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 1200(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],ymm2[0],ymm1[2],ymm2[2]
+; AVX1-ONLY-NEXT: vmovdqa 1312(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 976(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm0[1],ymm3[0],ymm0[2],ymm3[2]
+; AVX1-ONLY-NEXT: vmovapd 1648(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm15[1],ymm1[0],ymm15[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovdqa 1760(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 1424(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm13[1],ymm1[0],ymm13[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovdqa 1536(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = mem[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 976(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm0[1],ymm4[0],ymm0[2],ymm4[2]
; AVX1-ONLY-NEXT: vmovdqa 1088(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm14[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm3[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 528(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm10[1],ymm1[0],ymm10[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovdqa 640(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = xmm6[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm14[1],ymm1[0],ymm14[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovdqa 640(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = xmm10[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm10[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[1],ymm2[0],ymm6[2],ymm2[2]
-; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm14 = mem[8,9,10,11,12,13,14,15],xmm10[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm14[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm10[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm10[1],ymm2[0],ymm10[2],ymm2[2]
+; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm10
+; AVX1-ONLY-NEXT: vmovapd 128(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm13[0,1,2],ymm10[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm2 = mem[0],xmm2[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm10[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm2[0,1,2],ymm6[3]
-; AVX1-ONLY-NEXT: vmovapd 256(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vmovapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm8 = xmm12[0],xmm8[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm8[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vmovapd 576(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm8[0,1,2],ymm6[3]
-; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm1 = xmm14[0],xmm1[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovapd 352(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm11[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT: vmovapd 256(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm7 = xmm15[0],xmm7[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovapd 576(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT: vmovapd 480(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vmovapd 800(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm1[0,1,2],ymm6[3]
-; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm5 = xmm13[0],xmm5[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vmovapd 800(%rdi), %ymm12
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm12[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 704(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovapd 1024(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm15[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 928(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm3 = xmm11[0],xmm3[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 1024(%rdi), %ymm14
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd 928(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovapd 1248(%rdi), %ymm6
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 1152(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = xmm8[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovapd 1152(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm4 = xmm9[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovapd 1472(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm12[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 1376(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm4 = xmm0[0],xmm9[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm10[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovaps 1376(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vblendps $12, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = xmm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vmovapd 1696(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm8[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT: vmovapd 1600(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm5[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm13[0],ymm7[0],ymm13[3],ymm7[2]
+; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm13 = mem[8,9,10,11,12,13,14,15],xmm13[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm13[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm11[0],ymm13[0],ymm11[3],ymm13[2]
+; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm15 = xmm15[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm15[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 544(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = xmm3[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 640(%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm3[0],ymm5[0],ymm3[3],ymm5[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm15[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vmovapd 1696(%rdi), %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm5[0,1,2],ymm3[3]
-; AVX1-ONLY-NEXT: vmovapd 1600(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = xmm4[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 192(%rdi), %ymm9
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm10[0],ymm9[0],ymm10[3],ymm9[2]
-; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm7[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 416(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm10[0],ymm2[3],ymm10[2]
-; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 544(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm14[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovapd 640(%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm2[0],ymm4[0],ymm2[3],ymm4[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm14[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 864(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm14[0],ymm1[3],ymm14[2]
-; AVX1-ONLY-NEXT: vmovdqa 768(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm13 = xmm13[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm13[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovapd 864(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm12 = ymm12[0],ymm4[0],ymm12[3],ymm4[2]
+; AVX1-ONLY-NEXT: vmovdqa 768(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm1 = xmm1[8,9,10,11,12,13,14,15],xmm15[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm12[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa 992(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm11 = xmm11[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovapd 1088(%rdi), %ymm2
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm15[0],ymm2[0],ymm15[3],ymm2[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm11[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 1088(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm12 = ymm14[0],ymm3[0],ymm14[3],ymm3[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm12[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 1312(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm6[0],ymm15[0],ymm6[3],ymm15[2]
+; AVX1-ONLY-NEXT: vmovapd 1312(%rdi), %ymm2
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm12 = ymm6[0],ymm2[0],ymm6[3],ymm2[2]
; AVX1-ONLY-NEXT: vmovdqa 1216(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = xmm8[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm8[0,1],ymm13[2,3]
-; AVX1-ONLY-NEXT: vmovdqa 1440(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm9 = xmm9[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm9[0,1],ymm12[2,3]
+; AVX1-ONLY-NEXT: vmovdqa 1440(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],xmm9[0,1,2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovapd 1536(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm12[0],ymm1[0],ymm12[3],ymm1[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm0[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vmovapd 1760(%rdi), %ymm12
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0],ymm12[0],ymm5[3],ymm12[2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm10[0],ymm1[0],ymm10[3],ymm1[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm0[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT: vmovapd 1760(%rdi), %ymm10
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm8[0],ymm10[0],ymm8[3],ymm10[2]
; AVX1-ONLY-NEXT: vmovdqa 1664(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm11[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm9 = mem[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = mem[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm12[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm12 = xmm12[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm7 = mem[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,1,2,3],xmm11[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = mem[0,1,2],ymm5[3]
; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm11 = xmm11[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm11[0,1],ymm9[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm10 = mem[0,1,2],ymm10[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,1,2,3],xmm7[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm10[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm5[2,3]
; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm4 = mem[0,1,2],ymm4[3]
-; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm10 = xmm10[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = mem[0,1,2],ymm14[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2,3],xmm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm3[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2,3],xmm15[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm3 = mem[0,1,2],ymm3[3]
+; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm5 = xmm5[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2],ymm2[3]
-; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm3 = xmm3[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2],ymm15[3]
-; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm6 = mem[0,1,2,3],xmm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,1,2,3],xmm6[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm6 = xmm6[0],mem[1]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm6 = mem[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} xmm5 = xmm5[0],mem[1]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vblendpd $7, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = mem[0,1,2],ymm10[3]
; AVX1-ONLY-NEXT: vpblendw $15, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2,3],xmm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 192(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 128(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 64(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 224(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 160(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 192(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 128(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 224(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 160(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rdx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 192(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 128(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 224(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 160(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 64(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 128(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 192(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 224(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 160(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 224(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 192(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 160(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 128(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 64(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 32(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm6, (%r9)
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, (%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%r9)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, (%r9)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovapd %ymm5, 224(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm8, 192(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm13, 160(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm8, 224(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, 192(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm14, 160(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
@@ -3281,9 +3281,9 @@
; AVX1-ONLY-NEXT: vmovapd %ymm2, 160(%rax)
; AVX1-ONLY-NEXT: vmovapd %ymm3, 128(%rax)
; AVX1-ONLY-NEXT: vmovapd %ymm4, 96(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm10, 64(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm11, 64(%rax)
; AVX1-ONLY-NEXT: vmovapd %ymm7, 32(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm9, (%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm12, (%rax)
; AVX1-ONLY-NEXT: addq $1720, %rsp # imm = 0x6B8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
@@ -3291,15 +3291,14 @@
; AVX2-ONLY-LABEL: load_i64_stride7_vf32:
; AVX2-ONLY: # %bb.0:
; AVX2-ONLY-NEXT: subq $1624, %rsp # imm = 0x658
-; AVX2-ONLY-NEXT: vmovdqa 1216(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 768(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %ymm15
+; AVX2-ONLY-NEXT: vmovdqa 1216(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 768(%rdi), %ymm15
+; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %xmm3
; AVX2-ONLY-NEXT: vmovdqa 272(%rdi), %xmm2
; AVX2-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3309,7 +3308,7 @@
; AVX2-ONLY-NEXT: vmovdqa 832(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovdqa 672(%rdi), %xmm4
; AVX2-ONLY-NEXT: vmovdqa 720(%rdi), %xmm2
; AVX2-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3319,32 +3318,33 @@
; AVX2-ONLY-NEXT: vmovdqa 1280(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 1120(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 1120(%rdi), %xmm6
; AVX2-ONLY-NEXT: vmovdqa 1168(%rdi), %xmm2
; AVX2-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm10[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm6[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1664(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vmovdqa 1664(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 1728(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovdqa 1568(%rdi), %xmm7
; AVX2-ONLY-NEXT: vmovdqa 1616(%rdi), %xmm2
; AVX2-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm7[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm5
; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 48(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = mem[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 48(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = mem[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %ymm9
@@ -3352,10 +3352,10 @@
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovdqa 496(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm6[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vmovdqa 496(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm10[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 992(%rdi), %ymm11
@@ -3364,37 +3364,36 @@
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm0[6,7]
; AVX2-ONLY-NEXT: vmovdqa 896(%rdi), %xmm13
-; AVX2-ONLY-NEXT: vmovdqa 944(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm13[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovdqa 944(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovdqa %xmm2, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm13[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 1440(%rdi), %ymm14
; AVX2-ONLY-NEXT: vmovdqa 1504(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm14[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 1344(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm14[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovdqa 1344(%rdi), %xmm8
; AVX2-ONLY-NEXT: vmovdqa 1392(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm12 = xmm2[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm3[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm1 = ymm15[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm15[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm12 = xmm8[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm3[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4,5,6,7],ymm1[24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20,21,22,23]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 736(%rdi), %xmm1
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm4[8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 832(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm4 = mem[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm4 = ymm15[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm15[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 1184(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm10[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm6[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqa 1280(%rdi), %ymm15
; AVX2-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm4 = mem[8,9,10,11,12,13,14,15],ymm15[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm15[16,17,18,19,20,21,22,23]
@@ -3402,36 +3401,38 @@
; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 1632(%rdi), %xmm3
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm3 = xmm7[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 1728(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vmovdqa 1728(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vpalignr $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = mem[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],mem[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm12
; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm4 = ymm5[8,9,10,11,12,13,14,15],ymm12[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm12[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm5 = ymm8[8,9,10,11,12,13,14,15],ymm3[0,1,2,3,4,5,6,7],ymm8[24,25,26,27,28,29,30,31],ymm3[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm7 = mem[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm7 = ymm9[8,9,10,11,12,13,14,15],ymm10[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm10[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 512(%rdi), %xmm5
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = xmm6[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1056(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm7 = ymm11[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 960(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm8 = xmm13[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm8[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1504(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm11 = ymm14[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
-; AVX2-ONLY-NEXT: vmovdqa 1408(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4,5,6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastq 352(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = mem[8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm5 = ymm9[8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4,5,6,7],ymm9[24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vmovdqa 512(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = xmm10[8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 1056(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm9 = ymm11[8,9,10,11,12,13,14,15],ymm6[0,1,2,3,4,5,6,7],ymm11[24,25,26,27,28,29,30,31],ymm6[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vmovdqa 960(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm10 = xmm13[8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovdqa 1504(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} ymm10 = ymm14[8,9,10,11,12,13,14,15],ymm9[0,1,2,3,4,5,6,7],ymm14[24,25,26,27,28,29,30,31],ymm9[16,17,18,19,20,21,22,23]
+; AVX2-ONLY-NEXT: vmovdqa 1408(%rdi), %xmm11
+; AVX2-ONLY-NEXT: vpalignr {{.*#+}} xmm0 = xmm8[8,9,10,11,12,13,14,15],xmm11[0,1,2,3,4,5,6,7]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpbroadcastq 352(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
; AVX2-ONLY-NEXT: vmovdqa 240(%rdi), %xmm2
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
@@ -3449,29 +3450,30 @@
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpbroadcastq 1696(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm12[1],ymm0[3],ymm12[3]
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm7[1],ymm0[3],ymm7[3]
; AVX2-ONLY-NEXT: vmovdqa 1584(%rdi), %xmm1
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpbroadcastq 1472(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = mem[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpbroadcastq 128(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = mem[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm12[1],ymm0[3],ymm12[3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = mem[0,1],xmm3[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpbroadcastq 576(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm10[1],ymm0[3],ymm10[3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = mem[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vpunpckhqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = mem[0,1],xmm4[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vpbroadcastq 1024(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = mem[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastq 1472(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = mem[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm6[1],ymm0[3],ymm6[3]
+; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = mem[0,1],xmm5[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm0
@@ -3657,7 +3659,7 @@
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm2[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX2-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 16-byte Folded Reload
@@ -3667,7 +3669,7 @@
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
-; AVX2-ONLY-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],mem[2,3]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vpbroadcastq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 16-byte Folded Reload
@@ -3708,14 +3710,14 @@
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%rdx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm3, 192(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm3, 128(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm3, 64(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 192(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm3, 224(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm3, 160(%rcx)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-8.ll
index d04f587..4392654 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i64-stride-8.ll
@@ -3199,55 +3199,55 @@
;
; AVX1-ONLY-LABEL: load_i64_stride8_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $2280, %rsp # imm = 0x8E8
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm1[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm2[0],xmm0[0]
+; AVX1-ONLY-NEXT: subq $2296, %rsp # imm = 0x8F8
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm9[0],xmm7[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm8[0],xmm6[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm7[0],xmm6[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm5[0],xmm4[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm9
; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm9[1],xmm7[1]
+; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm11[0],xmm10[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm9[0],xmm8[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm9[1],xmm8[1]
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm10[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm5[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm10[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm8[1],xmm6[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm4[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm3[0]
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -3260,18 +3260,6 @@
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -3284,6 +3272,18 @@
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1984(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 1920(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -3296,37 +3296,6 @@
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm10[0],xmm0[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm14[0],ymm0[0],ymm14[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm11[0],xmm0[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 1104(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1040(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm12[0],xmm0[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %ymm1
@@ -3334,21 +3303,52 @@
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX1-ONLY-NEXT: vmovaps 1616(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1552(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm7[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1552(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm13[0],ymm0[0],ymm13[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm11[0],xmm14[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm10[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps 1104(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1040(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm7[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm8[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1984(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1920(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 1872(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps 1872(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1808(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1808(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm8[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3375,31 +3375,31 @@
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm8[0],ymm3[0],ymm8[2],ymm3[2]
; AVX1-ONLY-NEXT: vmovaps 336(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm1[0],xmm2[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm11[1],xmm14[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm10 = xmm10[1],mem[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm14[1],mem[1],ymm14[3],mem[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = xmm11[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm7 = xmm7[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = xmm12[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = xmm7[1],mem[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
@@ -3515,13 +3515,13 @@
; AVX1-ONLY-NEXT: vmovaps 1888(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 1824(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 2016(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 1952(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm1
@@ -3686,14 +3686,14 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 240(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm5, 96(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm5, 112(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 160(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 176(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 96(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 112(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 32(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 48(%rdx)
@@ -3702,6 +3702,10 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 208(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, (%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 16(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 64(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 80(%rdx)
@@ -3709,10 +3713,6 @@
; AVX1-ONLY-NEXT: vmovaps %xmm5, 128(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 144(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm5, (%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm5, 16(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, 32(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
@@ -3722,14 +3722,14 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, 128(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, (%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, 224(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, 160(%r8)
@@ -3745,9 +3745,9 @@
; AVX1-ONLY-NEXT: vmovaps %ymm5, 64(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm5, (%r8)
-; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm5, 240(%r9)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 240(%r9)
+; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 224(%r9)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm5, 48(%r9)
@@ -3837,7 +3837,7 @@
; AVX1-ONLY-NEXT: vmovaps %ymm9, 32(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
-; AVX1-ONLY-NEXT: addq $2280, %rsp # imm = 0x8E8
+; AVX1-ONLY-NEXT: addq $2296, %rsp # imm = 0x8F8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -3845,12 +3845,12 @@
; AVX2-ONLY: # %bb.0:
; AVX2-ONLY-NEXT: subq $2408, %rsp # imm = 0x968
; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %xmm2
; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm1[0]
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %xmm4
; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %xmm6
@@ -3865,32 +3865,32 @@
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm10[0],xmm9[0]
; AVX2-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm6[1],xmm4[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm8[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm10[1],xmm9[1]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm6[1],xmm4[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm10[1],xmm9[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm8[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm1[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %xmm1
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm7[1],xmm5[1]
; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm2[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm8[1]
+; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm3[0],xmm2[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm8[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm1
@@ -3898,18 +3898,6 @@
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -3922,6 +3910,18 @@
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1984(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -3940,6 +3940,18 @@
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm1
@@ -3948,9 +3960,9 @@
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm8[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3960,28 +3972,16 @@
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm8[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm12[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %ymm15
+; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %ymm15
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm15[0],ymm14[0],ymm15[2],ymm14[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
@@ -3995,23 +3995,23 @@
; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm4[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %ymm12
; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %ymm11
; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %ymm10
; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %ymm8
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm8[0],ymm10[0],ymm8[2],ymm10[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm6[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm6
; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm4
; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm3
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm2[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm2
@@ -4038,15 +4038,15 @@
; AVX2-ONLY-NEXT: # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm13[2,3],ymm5[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm15[1],ymm14[1],ymm15[3],ymm14[3]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm13[2,3],ymm5[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm15[1],ymm14[1],ymm15[3],ymm14[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
@@ -4057,7 +4057,7 @@
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm6[1],ymm7[1],ymm6[3],ymm7[3]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm10[1],ymm8[3],ymm10[3]
@@ -4333,14 +4333,14 @@
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 240(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 96(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 112(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 160(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 176(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 96(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 112(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 32(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 48(%rdx)
@@ -4349,6 +4349,10 @@
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 208(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, (%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 16(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 64(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 80(%rdx)
@@ -4356,10 +4360,6 @@
; AVX2-ONLY-NEXT: vmovaps %xmm1, 128(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 144(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, (%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 16(%rdx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -4369,14 +4369,14 @@
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 192(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 128(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 64(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 192(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 160(%r8)
@@ -6970,75 +6970,147 @@
;
; AVX1-ONLY-LABEL: load_i64_stride8_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $4984, %rsp # imm = 0x1378
-; AVX1-ONLY-NEXT: vmovaps 2496(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 2432(%rdi), %xmm1
+; AVX1-ONLY-NEXT: subq $4968, %rsp # imm = 0x1368
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3008(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps 2944(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3520(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 3456(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm5
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2880(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 3392(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 3904(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 3840(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps 4032(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vmovaps 3968(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm11[0],xmm10[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm9[0],xmm8[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm9[1],xmm8[1]
+; AVX1-ONLY-NEXT: vmovaps 2240(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 2176(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3328(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovaps 2112(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vmovaps 2048(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm11[0],xmm10[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm11[1],xmm10[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %xmm6
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm8[0],xmm7[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm6[0],xmm9[0]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm8[1],xmm7[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm6[1],xmm9[1]
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2816(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm4
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm6[0]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm8[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm6[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm8[1]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2368(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm2
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2304(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1984(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1920(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1856(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1792(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 2752(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 2688(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 2624(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 2560(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3264(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 3200(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3136(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 3072(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3776(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 3712(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3648(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 3584(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 2496(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 2432(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 2368(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 2304(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3008(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 2944(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 2880(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 2816(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3520(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 3456(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3392(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 3328(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 4032(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 3968(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3904(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 3840(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
@@ -7067,182 +7139,122 @@
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3776(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 3712(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3648(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 3584(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1344(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3264(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 3200(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1984(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1920(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3136(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 3072(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1856(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovaps 1792(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2752(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 2688(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2624(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 2560(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1600(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1536(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 576(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 512(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2240(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 2176(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2112(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 2048(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 1024(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 336(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 848(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 784(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 1360(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps 1104(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1296(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 1040(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1984(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1920(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 1872(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps 1616(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1808(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 1552(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2496(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2432(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 2240(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 2384(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 2176(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps 2128(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2320(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 2064(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3008(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2944(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 2752(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 2896(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 2688(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps 2640(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2832(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 2576(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3520(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3456(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 3264(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 3408(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 3200(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps 3152(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3344(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 3088(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3776(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3712(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT: vmovaps 3664(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 3600(%rdi), %xmm2
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -7259,132 +7271,156 @@
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %ymm10
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm6
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm6[0],ymm10[0],ymm6[2],ymm10[2]
-; AVX1-ONLY-NEXT: vmovaps 80(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm3[0],xmm7[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 3520(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 704(%rdi), %ymm14
-; AVX1-ONLY-NEXT: vmovaps 640(%rdi), %ymm11
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm11[0],ymm14[0],ymm11[2],ymm14[2]
-; AVX1-ONLY-NEXT: vmovaps 592(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 528(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm4[0],xmm5[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1216(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1152(%rdi), %ymm15
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 1104(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps 1040(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm8[0],xmm9[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm12[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1728(%rdi), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1664(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 3456(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT: vmovaps 1616(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 3408(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 1552(%rdi), %xmm12
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm12[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 3344(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2240(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 3008(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 2944(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2176(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps 2896(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 2832(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 2496(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 2432(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT: vmovaps 2384(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 2320(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1984(%rdi), %ymm1
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 1920(%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 2128(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2064(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 1872(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vmovaps 1808(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm13[0],xmm14[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2752(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2688(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 1472(%rdi), %ymm12
+; AVX1-ONLY-NEXT: vmovaps 1408(%rdi), %ymm11
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
+; AVX1-ONLY-NEXT: vmovaps 1360(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vmovaps 1296(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm10[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 2640(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 2576(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 960(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovaps 896(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX1-ONLY-NEXT: vmovaps 848(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 784(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm5[0],xmm6[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3264(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3200(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm4
+; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %ymm3
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovaps 336(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 272(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm15 = xmm1[0],xmm2[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 3152(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3088(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3776(%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3712(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT: vmovaps 3664(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 3600(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm0[0],xmm1[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm6[1],ymm10[1],ymm6[3],ymm10[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm7[1]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15, %xmm15 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm15 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm6[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm12[1],ymm11[3],ymm12[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm9[1],xmm10[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm14[1],ymm11[3],ymm14[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm5[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm15[1],mem[1],ymm15[3],mem[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm8[1],xmm9[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm12[1],mem[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm13[1],xmm14[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -7399,14 +7435,6 @@
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -7415,14 +7443,6 @@
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -7431,30 +7451,6 @@
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovlps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,1],xmm1[2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = xmm1[1],mem[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
@@ -7474,7 +7470,7 @@
; AVX1-ONLY-NEXT: vmovaps 352(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 480(%rdi), %xmm0
@@ -7536,7 +7532,7 @@
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 1760(%rdi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 1696(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
@@ -7938,6 +7934,38 @@
; AVX1-ONLY-NEXT: # xmm8 = xmm15[1],mem[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 496(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 480(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 432(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 416(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 368(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 352(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 288(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 224(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 160(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 96(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 48(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 304(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 240(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 176(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 112(%rsi)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 464(%rsi)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 448(%rsi)
@@ -7970,69 +7998,13 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 16(%rsi)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 496(%rsi)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 224(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 480(%rsi)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 240(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 32(%rsi)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 160(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 416(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 352(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 288(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 224(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 160(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 96(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 48(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 432(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 368(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 304(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 240(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 176(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 112(%rsi)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 128(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 144(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 256(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 272(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 64(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 80(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, (%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 16(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 192(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 208(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 320(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 336(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 384(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 400(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 448(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 464(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 176(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 96(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
@@ -8042,29 +8014,69 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 48(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 160(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 480(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 176(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 224(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 240(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 288(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 304(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 352(%rdx)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 368(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 496(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 416(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 432(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 480(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 352(%rdx)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %xmm8, 496(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 368(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 288(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 304(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 448(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 464(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 384(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 400(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 320(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 336(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 192(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 208(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, (%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 16(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 64(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 80(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 128(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 144(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 256(%rdx)
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 272(%rdx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 160(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 224(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 288(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 352(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 416(%rcx)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 480(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 448(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
@@ -8082,52 +8094,36 @@
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, (%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 480(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 416(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 352(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 288(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 224(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 160(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 480(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 448(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 416(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 384(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 352(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 320(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 288(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 256(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 224(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 160(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 128(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 96(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%r8)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 448(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 384(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 320(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 256(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 192(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 128(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%r8)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm8, (%r8)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 496(%r9)
@@ -8187,7 +8183,7 @@
; AVX1-ONLY-NEXT: vmovaps %xmm8, 64(%r9)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 48(%r9)
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 32(%r9)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 16(%r9)
@@ -8232,7 +8228,7 @@
; AVX1-ONLY-NEXT: vmovaps %xmm8, 224(%rax)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 208(%rax)
-; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm8 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 192(%rax)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX1-ONLY-NEXT: vmovaps %xmm8, 176(%rax)
@@ -8316,81 +8312,147 @@
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
-; AVX1-ONLY-NEXT: addq $4984, %rsp # imm = 0x1378
+; AVX1-ONLY-NEXT: addq $4968, %rsp # imm = 0x1368
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-ONLY-LABEL: load_i64_stride8_vf64:
; AVX2-ONLY: # %bb.0:
; AVX2-ONLY-NEXT: subq $5560, %rsp # imm = 0x15B8
-; AVX2-ONLY-NEXT: vmovaps 2496(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 2432(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3008(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovaps 2944(%rdi), %xmm3
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm3
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm2[0]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3520(%rdi), %xmm4
-; AVX2-ONLY-NEXT: vmovaps 3456(%rdi), %xmm5
+; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %xmm5
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm5[0],xmm4[0]
; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2880(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vmovaps 3392(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovaps 3904(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vmovaps 3840(%rdi), %xmm9
-; AVX2-ONLY-NEXT: vmovaps 4032(%rdi), %xmm10
-; AVX2-ONLY-NEXT: vmovaps 3968(%rdi), %xmm11
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm11[0],xmm10[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm9[0],xmm8[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm9[1],xmm8[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3328(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vmovaps 2240(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vmovaps 2176(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm8[0],xmm7[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %xmm6
+; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %xmm9
+; AVX2-ONLY-NEXT: vmovaps 2112(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vmovaps 2048(%rdi), %xmm11
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm8[1],xmm7[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm11[0],xmm10[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm11[1],xmm10[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %xmm7
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm8[0],xmm7[0]
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm7[0],xmm9[0]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm8[1],xmm7[1]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm7[1],xmm9[1]
; AVX2-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2816(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %xmm4
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm6[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm4[1],xmm6[1]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2368(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %xmm2
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2304(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm2[0]
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1984(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1856(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1792(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 2752(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 2688(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 2624(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 2560(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 3264(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 3200(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3136(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 3072(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3776(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 3712(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3648(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 3584(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2496(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 2432(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2368(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 2304(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3008(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 2944(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2880(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 2816(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3520(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 3456(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3392(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 3328(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 4032(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 3968(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3904(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 3840(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
@@ -8419,92 +8481,26 @@
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3776(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 3712(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3648(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 3584(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3264(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 3200(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 1984(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3136(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 3072(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2752(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 2688(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2624(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 2560(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2240(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 2176(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2112(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 2048(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 1856(%rdi), %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1792(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
@@ -8515,61 +8511,121 @@
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1856(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1792(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1984(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2368(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2304(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2496(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2432(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2112(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2048(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2240(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2176(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2624(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2560(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2752(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 2688(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3136(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3072(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3264(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3200(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3648(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3584(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3776(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3712(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3904(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3840(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 4032(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3968(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3392(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3328(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3520(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 3456(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
@@ -8587,216 +8643,156 @@
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3392(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 2368(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3328(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 2304(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3520(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 2496(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3456(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 2432(%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm2[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm15[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3904(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 1856(%rdi), %ymm1
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3840(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 1792(%rdi), %ymm2
; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 4032(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3968(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX2-ONLY-NEXT: vmovaps 1984(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm11[0],ymm10[2],ymm11[2]
; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm15[0],ymm10[2],ymm15[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm14[0],ymm9[0],ymm14[2],ymm9[2]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2112(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm13
+; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm13[0],ymm6[0],ymm13[2],ymm6[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm12[0],ymm3[0],ymm12[2],ymm3[2]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2048(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2240(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovaps 2176(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm9[0],ymm7[2],ymm9[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2624(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2560(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 2752(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovaps 2688(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm15 = ymm15[1],mem[1],ymm15[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm15[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm12[1],ymm3[1],ymm12[3],ymm3[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3136(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vmovaps 3072(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovaps 3264(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovaps 3200(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm11[0],ymm14[0],ymm11[2],ymm14[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm13[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm13[1],ymm6[1],ymm13[3],ymm6[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 3648(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovaps 3584(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vmovaps 3776(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovaps 3712(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm13[0],ymm8[0],ymm13[2],ymm8[2]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm8[1],ymm7[3],ymm8[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm14[1],ymm9[1],ymm14[3],ymm9[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm11[1],ymm10[3],ymm11[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
+; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm12[1],mem[1],ymm12[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm12[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm15[1],ymm10[3],ymm15[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm10 = ymm10[1],mem[1],ymm10[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm10[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm10 = ymm10[1],mem[1],ymm10[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm10[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm9[1],ymm7[3],ymm9[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm7 = ymm7[1],mem[1],ymm7[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm7[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm6[1],ymm5[3],ymm6[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm5 = ymm5[1],mem[1],ymm5[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm5[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm11[1],ymm14[1],ymm11[3],ymm14[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm3 = ymm3[1],mem[1],ymm3[3],mem[3]
-; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm3[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm13[1],ymm8[1],ymm13[3],ymm8[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = ymm1[1],mem[1],ymm1[3],mem[3]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -9287,6 +9283,38 @@
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm12[1],ymm4[1],ymm12[3],ymm4[3]
; AVX2-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 496(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 480(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 432(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 416(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 368(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 352(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 288(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 224(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 160(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 96(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 48(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 304(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 240(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 176(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 112(%rsi)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 464(%rsi)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 448(%rsi)
@@ -9319,69 +9347,13 @@
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 16(%rsi)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 496(%rsi)
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 224(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 480(%rsi)
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 240(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 160(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 416(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 352(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 288(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 224(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 160(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 96(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 48(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 432(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 368(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 304(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 240(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 176(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 112(%rsi)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 128(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 144(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 256(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 272(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 64(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 80(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, (%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 16(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 192(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 208(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 320(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 336(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 384(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 400(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 448(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 464(%rdx)
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 176(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 96(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -9391,29 +9363,69 @@
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 48(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 160(%rdx)
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 480(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 176(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 224(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 240(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 288(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 304(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 352(%rdx)
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 368(%rdx)
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 496(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 416(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 432(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 480(%rdx)
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 352(%rdx)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %xmm1, 496(%rdx)
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 368(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 288(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 304(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 448(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 464(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 384(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 400(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 320(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 336(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 192(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 208(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, (%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 16(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 64(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 80(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 128(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 144(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 256(%rdx)
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %xmm1, 272(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 160(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 288(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 352(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 416(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 480(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 448(%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -9431,52 +9443,36 @@
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rcx)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 480(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 416(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 352(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 288(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 160(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 480(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 448(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 416(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 384(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 352(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 320(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 288(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 256(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 192(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 160(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 128(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm1, 64(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 32(%r8)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 448(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 384(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 320(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 256(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 192(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 128(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 64(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, (%r8)
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vmovaps %xmm1, 496(%r9)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-2.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-2.ll
index f7f6a5e..d9e15ec 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-2.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-2.ll
@@ -365,14 +365,14 @@
define void @load_i8_stride2_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1) nounwind {
; SSE-LABEL: load_i8_stride2_vf64:
; SSE: # %bb.0:
-; SSE-NEXT: movdqa 64(%rdi), %xmm0
-; SSE-NEXT: movdqa 80(%rdi), %xmm4
+; SSE-NEXT: movdqa (%rdi), %xmm0
+; SSE-NEXT: movdqa 16(%rdi), %xmm7
+; SSE-NEXT: movdqa 32(%rdi), %xmm2
+; SSE-NEXT: movdqa 48(%rdi), %xmm4
; SSE-NEXT: movdqa 96(%rdi), %xmm1
-; SSE-NEXT: movdqa 112(%rdi), %xmm7
-; SSE-NEXT: movdqa (%rdi), %xmm2
-; SSE-NEXT: movdqa 16(%rdi), %xmm9
-; SSE-NEXT: movdqa 32(%rdi), %xmm3
-; SSE-NEXT: movdqa 48(%rdi), %xmm11
+; SSE-NEXT: movdqa 112(%rdi), %xmm9
+; SSE-NEXT: movdqa 64(%rdi), %xmm3
+; SSE-NEXT: movdqa 80(%rdi), %xmm11
; SSE-NEXT: movdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255]
; SSE-NEXT: movdqa %xmm11, %xmm8
; SSE-NEXT: pand %xmm6, %xmm8
@@ -381,38 +381,38 @@
; SSE-NEXT: packuswb %xmm8, %xmm5
; SSE-NEXT: movdqa %xmm9, %xmm10
; SSE-NEXT: pand %xmm6, %xmm10
-; SSE-NEXT: movdqa %xmm2, %xmm8
+; SSE-NEXT: movdqa %xmm1, %xmm8
; SSE-NEXT: pand %xmm6, %xmm8
; SSE-NEXT: packuswb %xmm10, %xmm8
; SSE-NEXT: movdqa %xmm7, %xmm12
; SSE-NEXT: pand %xmm6, %xmm12
-; SSE-NEXT: movdqa %xmm1, %xmm10
+; SSE-NEXT: movdqa %xmm0, %xmm10
; SSE-NEXT: pand %xmm6, %xmm10
; SSE-NEXT: packuswb %xmm12, %xmm10
; SSE-NEXT: movdqa %xmm4, %xmm12
; SSE-NEXT: pand %xmm6, %xmm12
-; SSE-NEXT: pand %xmm0, %xmm6
+; SSE-NEXT: pand %xmm2, %xmm6
; SSE-NEXT: packuswb %xmm12, %xmm6
; SSE-NEXT: psrlw $8, %xmm11
; SSE-NEXT: psrlw $8, %xmm3
; SSE-NEXT: packuswb %xmm11, %xmm3
; SSE-NEXT: psrlw $8, %xmm9
-; SSE-NEXT: psrlw $8, %xmm2
-; SSE-NEXT: packuswb %xmm9, %xmm2
-; SSE-NEXT: psrlw $8, %xmm7
; SSE-NEXT: psrlw $8, %xmm1
-; SSE-NEXT: packuswb %xmm7, %xmm1
-; SSE-NEXT: psrlw $8, %xmm4
+; SSE-NEXT: packuswb %xmm9, %xmm1
+; SSE-NEXT: psrlw $8, %xmm7
; SSE-NEXT: psrlw $8, %xmm0
-; SSE-NEXT: packuswb %xmm4, %xmm0
-; SSE-NEXT: movdqa %xmm6, 32(%rsi)
-; SSE-NEXT: movdqa %xmm10, 48(%rsi)
-; SSE-NEXT: movdqa %xmm8, (%rsi)
-; SSE-NEXT: movdqa %xmm5, 16(%rsi)
-; SSE-NEXT: movdqa %xmm0, 32(%rdx)
+; SSE-NEXT: packuswb %xmm7, %xmm0
+; SSE-NEXT: psrlw $8, %xmm4
+; SSE-NEXT: psrlw $8, %xmm2
+; SSE-NEXT: packuswb %xmm4, %xmm2
+; SSE-NEXT: movdqa %xmm6, 16(%rsi)
+; SSE-NEXT: movdqa %xmm10, (%rsi)
+; SSE-NEXT: movdqa %xmm8, 48(%rsi)
+; SSE-NEXT: movdqa %xmm5, 32(%rsi)
+; SSE-NEXT: movdqa %xmm2, 16(%rdx)
+; SSE-NEXT: movdqa %xmm0, (%rdx)
; SSE-NEXT: movdqa %xmm1, 48(%rdx)
-; SSE-NEXT: movdqa %xmm2, (%rdx)
-; SSE-NEXT: movdqa %xmm3, 16(%rdx)
+; SSE-NEXT: movdqa %xmm3, 32(%rdx)
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i8_stride2_vf64:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
index fec4b1b..99db42b 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
@@ -2013,13 +2013,13 @@
; SSE-LABEL: load_i8_stride5_vf64:
; SSE: # %bb.0:
; SSE-NEXT: subq $568, %rsp # imm = 0x238
-; SSE-NEXT: movdqa 160(%rdi), %xmm9
-; SSE-NEXT: movdqa 176(%rdi), %xmm3
+; SSE-NEXT: movdqa (%rdi), %xmm9
+; SSE-NEXT: movdqa 16(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 208(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 192(%rdi), %xmm1
+; SSE-NEXT: movdqa 32(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 48(%rdi), %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{.*#+}} xmm13 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255]
; SSE-NEXT: movdqa %xmm13, %xmm0
; SSE-NEXT: pandn %xmm1, %xmm0
@@ -2076,7 +2076,7 @@
; SSE-NEXT: packuswb %xmm0, %xmm0
; SSE-NEXT: pand %xmm10, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa 224(%rdi), %xmm1
+; SSE-NEXT: movdqa 64(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, %xmm5
; SSE-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3],xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]
; SSE-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
@@ -2093,11 +2093,11 @@
; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: por %xmm0, %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 32(%rdi), %xmm1
+; SSE-NEXT: movdqa 272(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm13, %xmm0
; SSE-NEXT: pandn %xmm1, %xmm0
-; SSE-NEXT: movdqa 48(%rdi), %xmm1
+; SSE-NEXT: movdqa 288(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT: pand %xmm13, %xmm1
; SSE-NEXT: por %xmm0, %xmm1
@@ -2113,11 +2113,11 @@
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,2,3]
; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa 16(%rdi), %xmm0
+; SSE-NEXT: movdqa 256(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm15, %xmm2
; SSE-NEXT: pandn %xmm0, %xmm2
-; SSE-NEXT: movdqa (%rdi), %xmm3
+; SSE-NEXT: movdqa 240(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm11, %xmm0
; SSE-NEXT: pandn %xmm3, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2147,7 +2147,7 @@
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: pand %xmm10, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: movdqa 64(%rdi), %xmm1
+; SSE-NEXT: movdqa 304(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, %xmm3
; SSE-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2164,11 +2164,11 @@
; SSE-NEXT: pand %xmm14, %xmm2
; SSE-NEXT: por %xmm2, %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 272(%rdi), %xmm2
+; SSE-NEXT: movdqa 192(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm13, %xmm1
; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: movdqa 288(%rdi), %xmm2
+; SSE-NEXT: movdqa 208(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pand %xmm13, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
@@ -2184,11 +2184,11 @@
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,2,3]
; SSE-NEXT: movdqa %xmm10, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: movdqa 256(%rdi), %xmm1
+; SSE-NEXT: movdqa 176(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm15, %xmm7
; SSE-NEXT: pandn %xmm1, %xmm7
-; SSE-NEXT: movdqa 240(%rdi), %xmm4
+; SSE-NEXT: movdqa 160(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: pandn %xmm4, %xmm1
; SSE-NEXT: movdqa %xmm0, %xmm3
@@ -2217,7 +2217,7 @@
; SSE-NEXT: packuswb %xmm4, %xmm4
; SSE-NEXT: pand %xmm10, %xmm4
; SSE-NEXT: por %xmm2, %xmm4
-; SSE-NEXT: movdqa 304(%rdi), %xmm2
+; SSE-NEXT: movdqa 224(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, %xmm0
; SSE-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3109,36 +3109,36 @@
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 32(%rsi)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 32(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 32(%rdx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, (%rdx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 32(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 48(%rcx)
+; SSE-NEXT: movaps %xmm0, 32(%rcx)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movaps %xmm3, 48(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movaps %xmm3, (%rcx)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movaps %xmm3, 32(%rcx)
; SSE-NEXT: movdqa %xmm14, 16(%r8)
-; SSE-NEXT: movdqa %xmm6, 48(%r8)
-; SSE-NEXT: movdqa %xmm15, (%r8)
+; SSE-NEXT: movdqa %xmm6, 32(%r8)
+; SSE-NEXT: movdqa %xmm15, 48(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%r8)
+; SSE-NEXT: movaps %xmm0, (%r8)
; SSE-NEXT: movaps %xmm1, 16(%r9)
-; SSE-NEXT: movaps %xmm4, 48(%r9)
-; SSE-NEXT: movaps %xmm7, (%r9)
-; SSE-NEXT: movaps %xmm2, 32(%r9)
+; SSE-NEXT: movaps %xmm4, 32(%r9)
+; SSE-NEXT: movaps %xmm7, 48(%r9)
+; SSE-NEXT: movaps %xmm2, (%r9)
; SSE-NEXT: addq $568, %rsp # imm = 0x238
; SSE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
index 7054850..b017965 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
@@ -2290,17 +2290,17 @@
; SSE-LABEL: load_i8_stride6_vf64:
; SSE: # %bb.0:
; SSE-NEXT: subq $824, %rsp # imm = 0x338
-; SSE-NEXT: movdqa 64(%rdi), %xmm3
+; SSE-NEXT: movdqa 256(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 80(%rdi), %xmm4
+; SSE-NEXT: movdqa 272(%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps (%rdi), %xmm0
+; SSE-NEXT: movaps 192(%rdi), %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 16(%rdi), %xmm2
+; SSE-NEXT: movdqa 208(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 32(%rdi), %xmm1
+; SSE-NEXT: movdqa 240(%rdi), %xmm14
+; SSE-NEXT: movdqa 224(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 48(%rdi), %xmm14
; SSE-NEXT: movdqa {{.*#+}} xmm6 = [65535,65535,0,65535,65535,0,65535,65535]
; SSE-NEXT: movdqa %xmm6, %xmm0
; SSE-NEXT: pandn %xmm1, %xmm0
@@ -2424,11 +2424,11 @@
; SSE-NEXT: pand %xmm8, %xmm0
; SSE-NEXT: por %xmm0, %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 224(%rdi), %xmm2
+; SSE-NEXT: movdqa 32(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, (%rsp) # 16-byte Spill
; SSE-NEXT: movdqa %xmm6, %xmm0
; SSE-NEXT: pandn %xmm2, %xmm0
-; SSE-NEXT: movdqa 240(%rdi), %xmm12
+; SSE-NEXT: movdqa 48(%rdi), %xmm12
; SSE-NEXT: movdqa %xmm5, %xmm2
; SSE-NEXT: pandn %xmm12, %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2446,11 +2446,11 @@
; SSE-NEXT: packuswb %xmm2, %xmm0
; SSE-NEXT: movdqa %xmm11, %xmm2
; SSE-NEXT: pandn %xmm0, %xmm2
-; SSE-NEXT: movdqa 208(%rdi), %xmm3
+; SSE-NEXT: movdqa 16(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm5, %xmm0
; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: movdqa 192(%rdi), %xmm9
+; SSE-NEXT: movdqa (%rdi), %xmm9
; SSE-NEXT: movdqa %xmm9, %xmm4
; SSE-NEXT: pand %xmm5, %xmm4
; SSE-NEXT: por %xmm0, %xmm4
@@ -2465,10 +2465,10 @@
; SSE-NEXT: packuswb %xmm0, %xmm0
; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: por %xmm2, %xmm0
-; SSE-NEXT: movdqa 272(%rdi), %xmm14
+; SSE-NEXT: movdqa 80(%rdi), %xmm14
; SSE-NEXT: movdqa %xmm6, %xmm2
; SSE-NEXT: pandn %xmm14, %xmm2
-; SSE-NEXT: movdqa 256(%rdi), %xmm13
+; SSE-NEXT: movdqa 64(%rdi), %xmm13
; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pand %xmm6, %xmm13
; SSE-NEXT: por %xmm2, %xmm13
@@ -3550,46 +3550,46 @@
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rsi)
+; SSE-NEXT: movaps %xmm0, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rsi)
+; SSE-NEXT: movaps %xmm0, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rdx)
+; SSE-NEXT: movaps %xmm0, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 48(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: movaps %xmm0, 32(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%rcx)
+; SSE-NEXT: movaps %xmm0, (%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 48(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rcx)
+; SSE-NEXT: movaps %xmm0, 32(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 32(%r8)
+; SSE-NEXT: movaps %xmm0, (%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 48(%r8)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%r8)
+; SSE-NEXT: movaps %xmm0, 32(%r8)
; SSE-NEXT: movdqa %xmm7, 16(%r9)
-; SSE-NEXT: movdqa %xmm6, 32(%r9)
+; SSE-NEXT: movdqa %xmm6, (%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 48(%r9)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%r9)
+; SSE-NEXT: movaps %xmm0, 32(%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT: movdqa %xmm2, 16(%rax)
-; SSE-NEXT: movdqa %xmm13, 32(%rax)
+; SSE-NEXT: movdqa %xmm13, (%rax)
; SSE-NEXT: movdqa %xmm12, 48(%rax)
-; SSE-NEXT: movdqa %xmm8, (%rax)
+; SSE-NEXT: movdqa %xmm8, 32(%rax)
; SSE-NEXT: addq $824, %rsp # imm = 0x338
; SSE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
index 0cd7ba0..c1525cf 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
@@ -1808,16 +1808,16 @@
; SSE-LABEL: load_i8_stride7_vf32:
; SSE: # %bb.0:
; SSE-NEXT: subq $632, %rsp # imm = 0x278
-; SSE-NEXT: movdqa 208(%rdi), %xmm9
-; SSE-NEXT: movdqa 192(%rdi), %xmm6
-; SSE-NEXT: movdqa 176(%rdi), %xmm13
-; SSE-NEXT: movdqa 112(%rdi), %xmm4
+; SSE-NEXT: movdqa 96(%rdi), %xmm9
+; SSE-NEXT: movdqa 80(%rdi), %xmm6
+; SSE-NEXT: movdqa 64(%rdi), %xmm13
+; SSE-NEXT: movdqa (%rdi), %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 128(%rdi), %xmm3
+; SSE-NEXT: movdqa 16(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 160(%rdi), %xmm2
+; SSE-NEXT: movdqa 32(%rdi), %xmm1
+; SSE-NEXT: movdqa 48(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 144(%rdi), %xmm1
; SSE-NEXT: movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,65535,0,65535,65535]
; SSE-NEXT: movdqa %xmm11, %xmm0
; SSE-NEXT: pandn %xmm1, %xmm0
@@ -1900,11 +1900,11 @@
; SSE-NEXT: pand %xmm6, %xmm1
; SSE-NEXT: por %xmm1, %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 32(%rdi), %xmm2
+; SSE-NEXT: movdqa 144(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: movdqa 48(%rdi), %xmm2
+; SSE-NEXT: movdqa 160(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pand %xmm11, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
@@ -1918,11 +1918,11 @@
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
; SSE-NEXT: packuswb %xmm1, %xmm2
-; SSE-NEXT: movdqa 16(%rdi), %xmm3
+; SSE-NEXT: movdqa 128(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm15, %xmm1
; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: movdqa (%rdi), %xmm3
+; SSE-NEXT: movdqa 112(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pand %xmm15, %xmm3
; SSE-NEXT: por %xmm1, %xmm3
@@ -1941,11 +1941,11 @@
; SSE-NEXT: pand %xmm0, %xmm1
; SSE-NEXT: pandn %xmm2, %xmm0
; SSE-NEXT: por %xmm0, %xmm1
-; SSE-NEXT: movdqa 64(%rdi), %xmm14
+; SSE-NEXT: movdqa 176(%rdi), %xmm14
; SSE-NEXT: movdqa %xmm5, %xmm0
; SSE-NEXT: pandn %xmm14, %xmm0
; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 80(%rdi), %xmm11
+; SSE-NEXT: movdqa 192(%rdi), %xmm11
; SSE-NEXT: movdqa %xmm11, %xmm2
; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pand %xmm5, %xmm2
@@ -1958,7 +1958,7 @@
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
-; SSE-NEXT: movdqa 96(%rdi), %xmm2
+; SSE-NEXT: movdqa 208(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, %xmm3
; SSE-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm7[8],xmm3[9],xmm7[9],xmm3[10],xmm7[10],xmm3[11],xmm7[11],xmm3[12],xmm7[12],xmm3[13],xmm7[13],xmm3[14],xmm7[14],xmm3[15],xmm7[15]
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2843,29 +2843,29 @@
; SSE-NEXT: por %xmm11, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm7
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rsi)
+; SSE-NEXT: movaps %xmm0, 16(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movaps %xmm2, 16(%rsi)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rdx)
+; SSE-NEXT: movaps %xmm2, (%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%rcx)
+; SSE-NEXT: movaps %xmm0, (%rdx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, (%r8)
+; SSE-NEXT: movaps %xmm0, (%rcx)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 16(%r8)
-; SSE-NEXT: movdqa %xmm4, (%r9)
-; SSE-NEXT: movdqa %xmm3, 16(%r9)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, (%r8)
+; SSE-NEXT: movdqa %xmm4, 16(%r9)
+; SSE-NEXT: movdqa %xmm3, (%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movdqa %xmm5, (%rax)
-; SSE-NEXT: movdqa %xmm6, 16(%rax)
+; SSE-NEXT: movdqa %xmm5, 16(%rax)
+; SSE-NEXT: movdqa %xmm6, (%rax)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movdqa %xmm7, (%rax)
-; SSE-NEXT: movdqa %xmm8, 16(%rax)
+; SSE-NEXT: movdqa %xmm7, 16(%rax)
+; SSE-NEXT: movdqa %xmm8, (%rax)
; SSE-NEXT: addq $632, %rsp # imm = 0x278
; SSE-NEXT: retq
;
@@ -6703,717 +6703,721 @@
;
; AVX1-ONLY-LABEL: load_i8_stride7_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $744, %rsp # imm = 0x2E8
+; AVX1-ONLY-NEXT: subq $728, %rsp # imm = 0x2D8
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [128,128,6,13,0,0,0,128,128,128,6,13,0,0,0,128]
; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm9
-; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm8
-; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 400(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm13, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm9
+; AVX1-ONLY-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [128,128,128,5,12,0,0,0,128,128,128,5,12,0,0,0]
; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa 240(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm3, %xmm4
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [0,7,14,128,128,0,0,0,0,7,14,128,128,0,0,0]
; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa (%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm5
; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm5, %xmm5
; AVX1-ONLY-NEXT: vpor %xmm4, %xmm5, %xmm7
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [128,128,0,0,0,3,10,128,128,128,0,0,0,3,10,128]
; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa 32(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovdqa 256(%rdi), %xmm5
; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm6
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [8,15,0,0,0,128,128,1,8,15,0,0,0,128,128,1]
; AVX1-ONLY-NEXT: # xmm5 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa 272(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm8, %xmm8
; AVX1-ONLY-NEXT: vpor %xmm6, %xmm8, %xmm8
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm6 = <255,255,255,255,255,0,0,0,0,0,u,u,u,u,u,u>
; AVX1-ONLY-NEXT: vpblendvb %xmm6, %xmm7, %xmm8, %xmm7
; AVX1-ONLY-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 240(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vmovdqa 16(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 48(%rdi), %xmm7
; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm7, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa 224(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm7, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm8, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm9, %xmm3
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa 256(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vmovdqa 32(%rdi), %xmm3
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa 272(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm4, %xmm4
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm7, %xmm4
; AVX1-ONLY-NEXT: vpor %xmm3, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [8,15,128,128,0,0,0,1,8,15,128,128,0,0,0,1]
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm8 = [8,15,128,128,0,0,0,1,8,15,128,128,0,0,0,1]
+; AVX1-ONLY-NEXT: # xmm8 = mem[0,0]
; AVX1-ONLY-NEXT: vpblendvb %xmm6, %xmm2, %xmm3, %xmm2
; AVX1-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm11, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 384(%rdi), %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm3, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm5
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm2
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [128,3,10,128,128,3,10,128,128,3,10,128,128,3,10,128]
-; AVX1-ONLY-NEXT: vmovdqa 144(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm5, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm13
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm10 = [12,128,128,5,12,128,128,5,12,128,128,5,12,128,128,5]
-; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm5, %xmm8
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm15
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm8, %xmm8
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm14 = <u,u,u,0,0,0,0,255,255,255,255,255,u,u,u,u>
-; AVX1-ONLY-NEXT: vpblendvb %xmm14, %xmm2, %xmm8, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm4, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm11
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm6 = [12,128,128,5,12,128,128,5,12,128,128,5,12,128,128,5]
+; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm4, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm14
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm9, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm10 = <u,u,u,0,0,0,0,255,255,255,255,255,u,u,u,u>
+; AVX1-ONLY-NEXT: vpblendvb %xmm10, %xmm2, %xmm9, %xmm2
; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 400(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 176(%rdi), %xmm2
; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm2, %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm7
-; AVX1-ONLY-NEXT: vmovdqa 384(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa 160(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm4, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm9
; AVX1-ONLY-NEXT: vpor %xmm0, %xmm2, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa 368(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovdqa 144(%rdi), %xmm4
; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm4, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm8
-; AVX1-ONLY-NEXT: vmovdqa 352(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm15
+; AVX1-ONLY-NEXT: vmovdqa 128(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm12, %xmm3
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpblendvb %xmm14, %xmm0, %xmm2, %xmm0
+; AVX1-ONLY-NEXT: vpblendvb %xmm10, %xmm0, %xmm2, %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [9,128,128,128,0,0,0,2,9,128,128,128,0,0,0,2]
; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm11, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm5, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [128,0,7,14,0,0,0,128,128,0,7,14,0,0,0,128]
; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm9, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm6 = [128,4,11,128,128,4,11,128,128,4,11,128,128,4,11,128]
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm13, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa %xmm13, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm13, %xmm4
; AVX1-ONLY-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm4 = [13,128,128,6,13,128,128,6,13,128,128,6,13,128,128,6]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm15, %xmm13
-; AVX1-ONLY-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpor %xmm12, %xmm13, %xmm12
-; AVX1-ONLY-NEXT: vpblendvb %xmm14, %xmm3, %xmm12, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm13, %xmm6
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm6, %xmm3
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm5 = [128,4,11,128,128,4,11,128,128,4,11,128,128,4,11,128]
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm11, %xmm8
+; AVX1-ONLY-NEXT: vmovdqa %xmm11, %xmm6
+; AVX1-ONLY-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm13 = [13,128,128,6,13,128,128,6,13,128,128,6,13,128,128,6]
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm14, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpor %xmm8, %xmm11, %xmm8
+; AVX1-ONLY-NEXT: vpblendvb %xmm10, %xmm3, %xmm8, %xmm3
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm5, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm9, %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm7, %xmm2
; AVX1-ONLY-NEXT: vpor %xmm0, %xmm2, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm8, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm8, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm10, %xmm6
-; AVX1-ONLY-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm10, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm15, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm12, %xmm3
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpblendvb %xmm14, %xmm0, %xmm2, %xmm0
+; AVX1-ONLY-NEXT: vpblendvb %xmm10, %xmm0, %xmm2, %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [0,0,128,128,128,5,12,0,0,0,128,128,128,5,12,0]
; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm1, %xmm2
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [0,0,0,7,14,128,128,0,0,0,0,7,14,128,128,0]
-; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm15, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm2
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [0,0,0,7,14,128,128,0,0,0,0,7,14,128,128,0]
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm14, %xmm3
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm3
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [10,128,128,128,0,0,0,3,10,128,128,128,0,0,0,3]
; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm11, %xmm10
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [128,1,8,15,0,0,0,128,128,1,8,15,0,0,0,128]
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm9, %xmm12
-; AVX1-ONLY-NEXT: vpor %xmm10, %xmm12, %xmm12
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm1, %xmm8
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [128,1,8,15,0,0,0,128,128,1,8,15,0,0,0,128]
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vpor %xmm8, %xmm10, %xmm8
; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm10 = <u,u,255,255,255,255,255,0,0,0,0,0,u,u,u,u>
-; AVX1-ONLY-NEXT: vpblendvb %xmm10, %xmm3, %xmm12, %xmm3
+; AVX1-ONLY-NEXT: vpblendvb %xmm10, %xmm3, %xmm8, %xmm3
; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm8, %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm6, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm15, %xmm0
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm12, %xmm1
; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm5, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm7, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm9, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm7, %xmm2
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
; AVX1-ONLY-NEXT: vpblendvb %xmm10, %xmm0, %xmm1, %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm13 = [128,128,128,6,13,0,0,0,128,128,128,6,13,0,0,0]
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm11, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [128,128,128,6,13,0,0,0,128,128,128,6,13,0,0,0]
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm13, %xmm0
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [1,8,15,128,128,0,0,0,1,8,15,128,128,0,0,0]
; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm5, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm4, %xmm3
; AVX1-ONLY-NEXT: vpor %xmm0, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm4 = [9,128,128,2,9,128,128,2,9,128,128,2,9,128,128,2]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm14, %xmm0
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm10 = [128,4,11,128,128,4,11,128,128,4,11,128,128,4,11,128]
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm6 = [9,128,128,2,9,128,128,2,9,128,128,2,9,128,128,2]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm9, %xmm0
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm8 = [128,4,11,128,128,4,11,128,128,4,11,128,128,4,11,128]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm15, %xmm12
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm12, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm8 = <255,255,255,255,255,0,0,0,0,u,u,u,u,u,u,u>
-; AVX1-ONLY-NEXT: vpblendvb %xmm8, %xmm3, %xmm12, %xmm0
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm15, %xmm10
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm10, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm7 = <255,255,255,255,255,0,0,0,0,u,u,u,u,u,u,u>
+; AVX1-ONLY-NEXT: vpblendvb %xmm7, %xmm3, %xmm10, %xmm0
; AVX1-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm6, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm1, %xmm0
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm12, %xmm2
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm2, %xmm13
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm1, %xmm2
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm2, %xmm11
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm0, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm0, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm10, %xmm3
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpblendvb %xmm8, %xmm13, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpblendvb %xmm7, %xmm11, %xmm2, %xmm2
; AVX1-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = [2,9,128,128,128,0,0,0,2,9,128,128,128,0,0,0]
-; AVX1-ONLY-NEXT: # xmm9 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm5, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm8
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm14 = [2,9,128,128,128,0,0,0,2,9,128,128,128,0,0,0]
+; AVX1-ONLY-NEXT: # xmm14 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm4, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm4, %xmm7
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [128,128,0,7,14,0,0,0,128,128,0,7,14,0,0,0]
; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm11, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm4, %xmm2
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm4 = [10,128,128,3,10,128,128,3,10,128,128,3,10,128,128,3]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm14, %xmm10
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm5 = [128,5,12,128,128,5,12,128,128,5,12,128,128,5,12,128]
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm15, %xmm13
-; AVX1-ONLY-NEXT: vpor %xmm10, %xmm13, %xmm10
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm13 = <255,255,255,255,255,0,0,0,0,u,u,u,u,u,u,u>
-; AVX1-ONLY-NEXT: vpblendvb %xmm13, %xmm2, %xmm10, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm13, %xmm6
+; AVX1-ONLY-NEXT: vpor %xmm2, %xmm6, %xmm2
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm6 = [10,128,128,3,10,128,128,3,10,128,128,3,10,128,128,3]
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm9, %xmm8
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm4 = [128,5,12,128,128,5,12,128,128,5,12,128,128,5,12,128]
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm15, %xmm11
+; AVX1-ONLY-NEXT: vpor %xmm8, %xmm11, %xmm8
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm11 = <255,255,255,255,255,0,0,0,0,u,u,u,u,u,u,u>
+; AVX1-ONLY-NEXT: vpblendvb %xmm11, %xmm2, %xmm8, %xmm2
; AVX1-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm12, %xmm7
-; AVX1-ONLY-NEXT: vmovdqa %xmm12, %xmm9
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm6, %xmm2
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm2, %xmm6
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm1, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm7
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm0, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm12, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm1, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm14
+; AVX1-ONLY-NEXT: vpor %xmm5, %xmm2, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm0, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm10, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm10, %xmm4
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpblendvb %xmm13, %xmm6, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm13, %xmm6
+; AVX1-ONLY-NEXT: vpblendvb %xmm11, %xmm1, %xmm2, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm11, %xmm5
; AVX1-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [3,10,128,128,128,0,0,0,3,10,128,128,128,0,0,0]
; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm8, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm7, %xmm3
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [128,128,1,8,15,0,0,0,128,128,1,8,15,0,0,0]
; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm11, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm4, %xmm10
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm13, %xmm6
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm6, %xmm8
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [11,128,128,4,11,128,128,4,11,128,128,4,11,128,128,4]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm14, %xmm12
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm4 = [128,6,13,128,128,6,13,128,128,6,13,128,128,6,13,128]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm15, %xmm13
-; AVX1-ONLY-NEXT: vpor %xmm12, %xmm13, %xmm12
-; AVX1-ONLY-NEXT: vpblendvb %xmm6, %xmm10, %xmm12, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm9, %xmm10
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm6 = [128,6,13,128,128,6,13,128,128,6,13,128,128,6,13,128]
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm15, %xmm11
+; AVX1-ONLY-NEXT: vpor %xmm10, %xmm11, %xmm10
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm7
+; AVX1-ONLY-NEXT: vpblendvb %xmm5, %xmm8, %xmm10, %xmm5
; AVX1-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm9, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm5, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm12, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm14, %xmm2
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm7, %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm0, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm0, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm4, %xmm3
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpblendvb %xmm6, %xmm1, %xmm2, %xmm0
+; AVX1-ONLY-NEXT: vpblendvb %xmm7, %xmm1, %xmm2, %xmm0
; AVX1-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [0,0,128,128,128,6,13,0,0,0,128,128,128,6,13,0]
; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm13, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm10, %xmm1
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [0,0,1,8,15,128,128,0,0,0,1,8,15,128,128,0]
; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm15, %xmm3
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm3, %xmm1
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm10 = [128,2,9,128,128,2,9,128,128,2,9,128,128,2,9,128]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm6, %xmm3
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm12 = [11,128,128,4,11,128,128,4,11,128,128,4,11,128,128,4]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm11, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,255,255,255,255,255,0,0,0,0,u,u,u,u,u>
-; AVX1-ONLY-NEXT: vpblendvb %xmm3, %xmm1, %xmm4, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa (%rsp), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm8, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm9, %xmm1
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm5, %xmm1
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [128,2,9,128,128,2,9,128,128,2,9,128,128,2,9,128]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm7, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm7, %xmm6
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm11 = [11,128,128,4,11,128,128,4,11,128,128,4,11,128,128,4]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm9, %xmm8
+; AVX1-ONLY-NEXT: vpor %xmm6, %xmm8, %xmm6
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,255,255,255,255,255,0,0,0,0,u,u,u,u,u>
+; AVX1-ONLY-NEXT: vpblendvb %xmm8, %xmm1, %xmm6, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm4, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm5, %xmm1
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm13, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm14, %xmm2
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vpblendvb %xmm3, %xmm0, %xmm1, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm3, %xmm1
+; AVX1-ONLY-NEXT: vpblendvb %xmm8, %xmm0, %xmm1, %xmm0
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm15[u,u,2,9],zero,zero,zero,xmm15[u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vmovdqa %xmm15, %xmm14
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [0,0,2,9,128,128,128,0,0,0,2,9,128,128,128,0]
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm15, %xmm0
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm2 = [0,0,128,128,0,7,14,0,0,0,128,128,0,7,14,0]
; AVX1-ONLY-NEXT: # xmm2 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm13, %xmm10
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm10, %xmm12
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm10 = [128,3,10,128,128,3,10,128,128,3,10,128,128,3,10,128]
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm6, %xmm13
-; AVX1-ONLY-NEXT: vmovdqa %xmm6, %xmm15
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm10, %xmm3
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm6 = [128,3,10,128,128,3,10,128,128,3,10,128,128,3,10,128]
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm7, %xmm11
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm0 = [12,128,128,5,12,128,128,5,12,128,128,5,12,128,128,5]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm11, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, %xmm3
-; AVX1-ONLY-NEXT: vpor %xmm4, %xmm13, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm11
-; AVX1-ONLY-NEXT: vpblendvb %xmm1, %xmm12, %xmm4, %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm9[u,u,2,9],zero,zero,zero,xmm9[u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm8, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm9, %xmm12
+; AVX1-ONLY-NEXT: vpor %xmm11, %xmm12, %xmm11
+; AVX1-ONLY-NEXT: vpblendvb %xmm8, %xmm3, %xmm11, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm5, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm11
+; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm4, %xmm2
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm5, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm6
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm7, %xmm0
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm13, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm14, %xmm0
; AVX1-ONLY-NEXT: vpor %xmm2, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vpblendvb %xmm11, %xmm1, %xmm0, %xmm12
+; AVX1-ONLY-NEXT: vpblendvb %xmm8, %xmm1, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [0,0,3,10,128,128,128,0,0,0,3,10,128,128,128,0]
; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm14, %xmm0
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = [0,0,128,128,1,8,15,0,0,0,128,128,1,8,15,0]
-; AVX1-ONLY-NEXT: # xmm10 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm5, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm15, %xmm0
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [0,0,128,128,1,8,15,0,0,0,128,128,1,8,15,0]
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa %xmm10, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm10, %xmm2
; AVX1-ONLY-NEXT: vpor %xmm0, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm13 = [128,4,11,128,128,4,11,128,128,4,11,128,128,4,11,128]
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm15, %xmm4
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm12 = [128,4,11,128,128,4,11,128,128,4,11,128,128,4,11,128]
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm7, %xmm3
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm0 = [13,128,128,6,13,128,128,6,13,128,128,6,13,128,128,6]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm3, %xmm11
-; AVX1-ONLY-NEXT: vpor %xmm4, %xmm11, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,255,255,255,255,255,0,0,0,0,u,u,u,u,u>
-; AVX1-ONLY-NEXT: vpblendvb %xmm3, %xmm2, %xmm4, %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm9, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm8, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm1, %xmm4, %xmm1
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm6, %xmm4
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm7, %xmm0
-; AVX1-ONLY-NEXT: vpor %xmm4, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vpblendvb %xmm3, %xmm1, %xmm0, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [0,0,0,128,128,128,5,12,0,0,0,128,128,128,5,12]
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm6, %xmm10
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [0,0,0,0,7,14,128,128,0,0,0,0,7,14,128,128]
-; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm7, %xmm13
-; AVX1-ONLY-NEXT: vpor %xmm10, %xmm13, %xmm10
-; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm1 = <u,u,255,255,255,255,255,255,255,255,255,0,0,0,0,0>
-; AVX1-ONLY-NEXT: vpblendvb %xmm1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa 432(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm10, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa 416(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm13, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm4, %xmm0
-; AVX1-ONLY-NEXT: vpblendvb %xmm1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [0,0,0,128,128,128,6,13,0,0,0,128,128,128,6,13]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm9, %xmm10
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm10, %xmm3
+; AVX1-ONLY-NEXT: vpblendvb %xmm8, %xmm2, %xmm3, %xmm7
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm11, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm4, %xmm2
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm2, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm13, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm13, %xmm15
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm14, %xmm0
+; AVX1-ONLY-NEXT: vpor %xmm2, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vpblendvb %xmm8, %xmm1, %xmm0, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa 432(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vmovdqa 416(%rdi), %xmm12
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [0,0,0,128,128,128,5,12,0,0,0,128,128,128,5,12]
+; AVX1-ONLY-NEXT: # xmm6 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm11, %xmm8
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [0,0,0,0,7,14,128,128,0,0,0,0,7,14,128,128]
; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm6, %xmm4
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [0,0,0,1,8,15,128,128,0,0,0,1,8,15,128,128]
-; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm7, %xmm14
-; AVX1-ONLY-NEXT: vpor %xmm4, %xmm14, %xmm4
-; AVX1-ONLY-NEXT: vpblendvb %xmm1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm10, %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm13, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm4, %xmm0
-; AVX1-ONLY-NEXT: vpblendvb %xmm1, %xmm12, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm12, %xmm10
+; AVX1-ONLY-NEXT: vpor %xmm8, %xmm10, %xmm8
+; AVX1-ONLY-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,255,255,255,255,255,255,255,255,255,0,0,0,0,0>
+; AVX1-ONLY-NEXT: vpblendvb %xmm2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa 208(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm8, %xmm6
+; AVX1-ONLY-NEXT: vmovdqa 192(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm13, %xmm0
+; AVX1-ONLY-NEXT: vpor %xmm6, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vpblendvb %xmm2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [0,0,0,128,128,128,6,13,0,0,0,128,128,128,6,13]
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm11, %xmm6
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [0,0,0,1,8,15,128,128,0,0,0,1,8,15,128,128]
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm12, %xmm10
+; AVX1-ONLY-NEXT: vpor %xmm6, %xmm10, %xmm6
+; AVX1-ONLY-NEXT: vpblendvb %xmm2, (%rsp), %xmm6, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm8, %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm13, %xmm0
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vpblendvb %xmm2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm0 = [0,0,0,2,9,128,128,128,0,0,0,2,9,128,128,128]
; AVX1-ONLY-NEXT: # xmm0 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm7, %xmm4
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [0,0,0,128,128,0,7,14,0,0,0,128,128,0,7,14]
-; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm6, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa %xmm6, %xmm14
-; AVX1-ONLY-NEXT: vpor %xmm4, %xmm12, %xmm4
-; AVX1-ONLY-NEXT: vpblendvb %xmm1, %xmm2, %xmm4, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm13, %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm10, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm4, %xmm0
-; AVX1-ONLY-NEXT: vpblendvb %xmm1, %xmm3, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm15 = [2,9,0,0,2,9,0,0,2,9,0,0,2,9,0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm5, %xmm3
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm0 = [0,0,4,11,0,0,4,11,0,0,4,11,0,0,4,11]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm2, %xmm4
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [128,5,12,0,0,0,128,128,128,5,12,0,0,0,128,128]
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm2, %xmm6
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [14,128,128,0,0,0,0,7,14,128,128,0,0,0,0,7]
-; AVX1-ONLY-NEXT: # xmm5 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm2, %xmm9
-; AVX1-ONLY-NEXT: vpor %xmm6, %xmm9, %xmm6
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm9 = xmm6[0],xmm4[1,2],xmm6[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [0,0,0,3,10,128,128,128,0,0,0,3,10,128,128,128]
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm12, %xmm1
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm6 = [0,0,0,128,128,0,7,14,0,0,0,128,128,0,7,14]
; AVX1-ONLY-NEXT: # xmm6 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm7, %xmm11
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [0,0,0,128,128,1,8,15,0,0,0,128,128,1,8,15]
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm14, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpor %xmm11, %xmm12, %xmm11
-; AVX1-ONLY-NEXT: vpblendvb %xmm1, %xmm9, %xmm11, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm2, %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm8, %xmm8
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm2, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm2, %xmm5
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm5, %xmm3
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm3[0],xmm0[1,2],xmm3[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm13, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm10, %xmm9
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm10, %xmm4
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vpblendvb %xmm1, %xmm0, %xmm3, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm10 = [0,0,2,9,0,0,2,9,0,0,2,9,0,0,2,9]
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm14, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm11, %xmm9
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm9, %xmm1
+; AVX1-ONLY-NEXT: vpblendvb %xmm2, %xmm7, %xmm1, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm13, %xmm0
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm8, %xmm1
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-ONLY-NEXT: vpblendvb %xmm2, %xmm4, %xmm0, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm14 = [2,9,0,0,2,9,0,0,2,9,0,0,2,9,0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm5, %xmm1
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm0 = [0,0,4,11,0,0,4,11,0,0,4,11,0,0,4,11]
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm7, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, %xmm6
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,1,2,3,4,5],xmm3[6,7]
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm15 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
-; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm1, %xmm5
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm2 = [0,0,6,13,0,0,6,13,0,0,6,13,0,0,6,13]
-; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vpshufb %xmm2, %xmm14, %xmm8
-; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
-; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = xmm5[10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9]
-; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm8
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [0,0,0,1,2,3,8,15,0,0,0,1,2,3,8,15]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm3, %xmm4
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm5 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm1 = [128,5,12,0,0,0,128,128,128,5,12,0,0,0,128,128]
+; AVX1-ONLY-NEXT: # xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm3, %xmm6
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [14,128,128,0,0,0,0,7,14,128,128,0,0,0,0,7]
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm3, %xmm7
+; AVX1-ONLY-NEXT: vpor %xmm6, %xmm7, %xmm6
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0],xmm5[1,2],xmm6[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm5 = [0,0,0,3,10,128,128,128,0,0,0,3,10,128,128,128]
+; AVX1-ONLY-NEXT: # xmm5 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm12, %xmm9
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [0,0,0,128,128,1,8,15,0,0,0,128,128,1,8,15]
; AVX1-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm5, %xmm5
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm11 = [0,7,14,0,0,7,14,0,0,7,14,0,0,7,14,0]
-; AVX1-ONLY-NEXT: vmovdqa 112(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm11, %xmm10
+; AVX1-ONLY-NEXT: vpor %xmm9, %xmm10, %xmm9
+; AVX1-ONLY-NEXT: vpblendvb %xmm2, %xmm6, %xmm9, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm3, %xmm6
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm15, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm4, %xmm1
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm13, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm8, %xmm4
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm4, %xmm1
+; AVX1-ONLY-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm14 = [0,0,2,9,0,0,2,9,0,0,2,9,0,0,2,9]
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm11, %xmm0
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm1 = [0,0,4,11,0,0,4,11,0,0,4,11,0,0,4,11]
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm12, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm5
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2,3,4,5],xmm0[6,7]
+; AVX1-ONLY-NEXT: vmovdqa 304(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm1, %xmm12
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm5, %ymm5
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm1, %ymm5
-; AVX1-ONLY-NEXT: vorps %ymm5, %ymm12, %ymm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm12 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm12, %ymm4
-; AVX1-ONLY-NEXT: vandps %ymm5, %ymm12, %ymm5
-; AVX1-ONLY-NEXT: vorps %ymm4, %ymm5, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm9, %xmm4
-; AVX1-ONLY-NEXT: vpshufb %xmm0, %xmm13, %xmm5
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3,4,5],xmm4[6,7]
-; AVX1-ONLY-NEXT: vmovdqa 304(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm0, %xmm4
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, %xmm5
-; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm10, %xmm15
-; AVX1-ONLY-NEXT: vmovdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm15[0],xmm4[0],xmm15[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = xmm4[10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7,8,9]
-; AVX1-ONLY-NEXT: vmovdqa %xmm0, %xmm15
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm4, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm7, %xmm11
-; AVX1-ONLY-NEXT: vmovdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm1, %ymm0
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm11, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm12, %ymm1
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm12, %ymm0
-; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = xmm1[u,u,u,u,4,11,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm1 = [0,0,6,13,0,0,6,13,0,0,6,13,0,0,6,13]
+; AVX1-ONLY-NEXT: vmovdqa 288(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vpshufb %xmm1, %xmm7, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm6
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm4 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
+; AVX1-ONLY-NEXT: vmovdqa 320(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm4 = xmm4[10,11,12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm10 = xmm4[u,u,u,u,u,u,u,u,u,u,0,1,2,3,8,15]
+; AVX1-ONLY-NEXT: vmovdqa 336(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm15 = xmm2[0,7,14,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm10, %ymm10
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm2 = [65535,65535,65535,65535,65535,0,0,0,0,0,65535,65535,65535,65535,65535,65535]
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm15 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandnps %ymm10, %ymm2, %ymm10
+; AVX1-ONLY-NEXT: vorps %ymm10, %ymm15, %ymm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm15
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX1-ONLY-NEXT: vandnps %ymm15, %ymm4, %ymm15
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm10, %ymm10
+; AVX1-ONLY-NEXT: vorps %ymm15, %ymm10, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm4[u,u,u,u,u,u,3,10,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm6[u,u,u,u,u,u,5,12,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm11[0],xmm1[0],xmm11[1],xmm1[1],xmm11[2],xmm1[2],xmm11[3],xmm1[3]
-; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2,3,4,5],xmm1[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm14, %xmm8, %xmm10
+; AVX1-ONLY-NEXT: vpshufb %xmm5, %xmm13, %xmm14
+; AVX1-ONLY-NEXT: vmovdqa %xmm13, %xmm5
+; AVX1-ONLY-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1],xmm14[2],xmm10[2],xmm14[3],xmm10[3]
+; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = mem[0,1,2,3,4,5],xmm10[6,7]
+; AVX1-ONLY-NEXT: vmovdqa 80(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm15 = xmm0[u,u,u,u,4,11,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa 64(%rdi), %xmm9
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm9, %xmm6
+; AVX1-ONLY-NEXT: vmovdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm6[0],xmm15[0],xmm6[1],xmm15[1]
+; AVX1-ONLY-NEXT: vmovdqa 96(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = xmm6[10,11,12,13,14,15],xmm8[0,1,2,3,4,5,6,7,8,9]
+; AVX1-ONLY-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm6[u,u,u,u,u,u,u,u,u,u,0,1,2,3,8,15]
+; AVX1-ONLY-NEXT: vmovdqa 112(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm6[0,7,14,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, %xmm14
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm2, %ymm0
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm13, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm4, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm11[u,u,u,u,u,u,3,10,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm12[u,u,u,u,u,u,5,12,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm13[0],xmm2[0],xmm13[1],xmm2[1],xmm13[2],xmm2[2],xmm13[3],xmm2[3]
+; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = mem[0,1,2,3,4,5],xmm2[6,7]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm6[u,u,u,u,5,12,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm14[u,u,u,u,u,0,7,14,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm6[u,u,u,u,5,12,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm7[u,u,u,u,u,0,7,14,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX1-ONLY-NEXT: vxorps %xmm0, %xmm0, %xmm0
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6],xmm0[7]
-; AVX1-ONLY-NEXT: vpxor %xmm11, %xmm11, %xmm11
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm8[u,u,u,u,u,u,u,u,u],zero,zero,zero,zero,zero,xmm8[3,10]
-; AVX1-ONLY-NEXT: vpor %xmm0, %xmm2, %xmm0
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm14[1,8,15,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm2 = [0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandps %ymm2, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm0[7]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm1[u,u,u,u,u,u,u,u,u],zero,zero,zero,zero,zero,xmm1[3,10]
+; AVX1-ONLY-NEXT: vmovdqa %xmm1, %xmm2
+; AVX1-ONLY-NEXT: vpor %xmm0, %xmm3, %xmm0
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm15[1,8,15,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm12, %ymm1
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm12, %ymm0
+; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm3 = [0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm1, %ymm4, %ymm1
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm9[u,u,u,u,u,u,3,10,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm13[u,u,u,u,u,u,5,12,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm10[u,u,u,u,u,u,3,10,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm5[u,u,u,u,u,u,5,12,u,u,u,u,u,u,u,u]
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2,3,4,5],xmm0[6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm5[u,u,u,u,5,12,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm10[u,u,u,u,u,0,7,14,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6],xmm11[7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm15[u,u,u,u,u,u,u,u,u],zero,zero,zero,zero,zero,xmm15[3,10]
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm1, %xmm1
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm7[1,8,15,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandps %ymm2, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm9[u,u,u,u,u,0,7,14,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6],mem[7]
+; AVX1-ONLY-NEXT: vpxor %xmm9, %xmm9, %xmm9
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm8[u,u,u,u,u,u,u,u,u],zero,zero,zero,zero,zero,xmm8[3,10]
+; AVX1-ONLY-NEXT: vpor %xmm1, %xmm13, %xmm1
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm14[1,8,15,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vandnps {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandps %ymm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vorps %ymm1, %ymm13, %ymm1
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm12, %ymm0
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm12, %ymm1
+; AVX1-ONLY-NEXT: vandnps %ymm0, %ymm4, %ymm0
+; AVX1-ONLY-NEXT: vandps %ymm4, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm4[u,u,u,u,u,u,4,11,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[u,u,u,u,u,u,6,13,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm0 = xmm11[u,u,u,u,u,u,4,11,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm12[u,u,u,u,u,u,6,13,u,u,u,u,u,u,u,u]
; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3,4,5],xmm0[6,7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm1 = xmm6[u,u,u,u,u,u,u,u,u,u,u,u,6,13,u,u]
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [0,1,8,15,0,1,8,15,0,1,8,15,0,1,8,15]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm10, %xmm7
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm1 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6],xmm11[7]
-; AVX1-ONLY-NEXT: vpxor %xmm4, %xmm4, %xmm4
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [0,128,128,128,128,128,4,11,0,128,128,128,128,128,4,11]
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm8, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa %xmm8, %xmm11
+; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = mem[0,1,2,3,4,5],xmm0[6,7]
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [6,13,0,0,6,13,0,0,6,13,0,0,6,13,0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm6, %xmm1
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, %xmm8
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm4 = [0,1,8,15,0,1,8,15,0,1,8,15,0,1,8,15]
+; AVX1-ONLY-NEXT: vmovdqa %xmm7, %xmm6
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm7, %xmm11
+; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm1 = xmm11[2],xmm1[2],xmm11[3],xmm1[3]
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6],xmm9[7]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [0,128,128,128,128,128,4,11,0,128,128,128,128,128,4,11]
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, %xmm7
+; AVX1-ONLY-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm2, %xmm12
; AVX1-ONLY-NEXT: vpor %xmm1, %xmm12, %xmm1
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm8 = [2,9,0,0,2,9,0,0,2,9,0,0,2,9,0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm14, %xmm12
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm9 = [2,9,0,0,2,9,0,0,2,9,0,0,2,9,0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm15, %xmm12
+; AVX1-ONLY-NEXT: vmovdqa %xmm15, %xmm2
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm1, %ymm12
; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: vandnps %ymm12, %ymm1, %ymm12
-; AVX1-ONLY-NEXT: vorps %ymm0, %ymm12, %ymm12
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm12, %ymm13, %ymm12
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm13
; AVX1-ONLY-NEXT: vmovaps {{.*#+}} ymm0 = [65535,65535,65535,65535,65535,65535,65535,65535,65535,0,0,0,0,0,0,0]
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm0, %ymm13
; AVX1-ONLY-NEXT: vandps %ymm0, %ymm12, %ymm12
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm12, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm9[u,u,u,u,u,u,4,11,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm13[u,u,u,u,u,u,6,13,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm12[0],xmm2[0],xmm12[1],xmm2[1],xmm12[2],xmm2[2],xmm12[3],xmm2[3]
-; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = mem[0,1,2,3,4,5],xmm2[6,7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm5[u,u,u,u,u,u,u,u,u,u,u,u,6,13,u,u]
-; AVX1-ONLY-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm9, %xmm3
-; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm3[2],xmm12[2],xmm3[3],xmm12[3]
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm4[7]
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm15, %xmm7
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm4, %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vorps %ymm13, %ymm12, %ymm12
+; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm10[u,u,u,u,u,u,4,11,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm10[u,u,u,u,u,u,6,13,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3]
+; AVX1-ONLY-NEXT: vpblendw $63, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm12 = mem[0,1,2,3,4,5],xmm12[6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm5, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm15
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm4
+; AVX1-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm3 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; AVX1-ONLY-NEXT: vpxor %xmm13, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm13[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm10, %xmm4
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm14, %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm1, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm7, %ymm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm4, %ymm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm0, %ymm4
; AVX1-ONLY-NEXT: vandps %ymm0, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [0,2,9,128,128,128,0,0,0,2,9,128,128,128,0,0]
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm10, %xmm2
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [0,128,128,0,7,14,0,0,0,128,128,0,7,14,0,0]
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm6, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa %xmm6, %xmm14
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm12, %xmm2
-; AVX1-ONLY-NEXT: vpxor %xmm6, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6],xmm6[7]
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [0,2,9,128,128,128,0,0,0,2,9,128,128,128,0,0]
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm6, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, %xmm9
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [0,128,128,0,7,14,0,0,0,128,128,0,7,14,0,0]
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm8, %xmm12
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm12, %xmm3
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm13[7]
+; AVX1-ONLY-NEXT: vpxor %xmm8, %xmm8, %xmm8
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm12 = [0,128,128,128,128,128,5,12,0,128,128,128,128,128,5,12]
; AVX1-ONLY-NEXT: # xmm12 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, %xmm10
-; AVX1-ONLY-NEXT: vmovdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm11, %xmm13
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm13, %xmm13
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm8 = [3,10,0,0,3,10,0,0,3,10,0,0,3,10,0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm11, %xmm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm13, %ymm2
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm7, %xmm13
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm6 = [3,10,0,0,3,10,0,0,3,10,0,0,3,10,0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm2, %xmm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm13, %ymm3
; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm13, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm1, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm13, %ymm3
; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm13 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: vandnps %ymm13, %ymm0, %ymm13
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm13, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm13, %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm9, %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm5, %xmm3
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm3, %xmm2
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6],xmm6[7]
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm15, %xmm3
-; AVX1-ONLY-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm2, %xmm2
-; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm4, %xmm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa %xmm5, %xmm2
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm15, %xmm4
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm4, %xmm3
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm8[7]
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm10, %xmm4
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm14, %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vandps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm1, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm4, %ymm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vandnps %ymm4, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm9 = [2,9,0,0,2,9,0,0,2,9,0,0,2,9,0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm8, %xmm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = xmm4[4,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm7[0],xmm2[0],xmm7[1],xmm2[1],xmm7[2],xmm2[2],xmm7[3],xmm2[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm6[u,u,u,u],zero,zero,zero,xmm6[5,12,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm15 = [2,9,0,0,2,9,0,0,2,9,0,0,2,9,0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm8, %xmm3
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm10 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm5[u,u,u,u,0,7,14],zero,zero,xmm5[u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm3, %xmm12, %xmm3
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm3[u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm3[u,u]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm3 = [0,128,128,1,8,15,0,0,0,128,128,1,8,15,0,0]
-; AVX1-ONLY-NEXT: # xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm14, %xmm13
+; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm5, %xmm11
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm11[0],xmm3[0],xmm11[1],xmm3[1],xmm11[2],xmm3[2],xmm11[3],xmm3[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm7[u,u,u,u],zero,zero,zero,xmm7[5,12,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = xmm6[u,u,u,u,0,7,14],zero,zero,xmm6[u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm12, %xmm4
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm9[u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm9[u,u]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [0,128,128,1,8,15,0,0,0,128,128,1,8,15,0,0]
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm9, %xmm13
; AVX1-ONLY-NEXT: vpor %xmm12, %xmm13, %xmm12
; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,5,6],mem[7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = xmm10[u,u,u,u,u,u,u,u,u],zero,zero,zero,zero,zero,xmm10[6,13]
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm12, %xmm7
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm13 = [4,11,0,0,4,11,0,0,4,11,0,0,4,11,0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm11, %xmm12
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm7, %ymm7
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm1, %ymm7
-; AVX1-ONLY-NEXT: vorps %ymm7, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm7, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm7, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm11, %xmm2
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm9, %xmm7
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm7[0],xmm2[0],xmm7[1],xmm2[1],xmm7[2],xmm2[2],xmm7[3],xmm2[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = xmm10[u,u,u,u],zero,zero,zero,xmm10[5,12,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm9[u,u,u,u,u,u,u,u,u],zero,zero,zero,zero,zero,xmm9[6,13]
+; AVX1-ONLY-NEXT: vpor %xmm11, %xmm12, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm9, %xmm12
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm11, %ymm11
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm11, %ymm1, %ymm11
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm11, %ymm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vandnps %ymm11, %ymm0, %ymm11
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vorps %ymm3, %ymm11, %ymm3
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm14[u,u,u,u,0,7,14],zero,zero,xmm14[u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm12, %xmm7
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = xmm7[u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm7[u,u]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm12, %xmm3
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm3, %xmm3
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],mem[7]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = xmm15[u,u,u,u,u,u,u,u,u],zero,zero,zero,zero,zero,xmm15[6,13]
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm15, %xmm14, %xmm3
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm9, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa %xmm10, %xmm13
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm11[0],xmm3[0],xmm11[1],xmm3[1],xmm11[2],xmm3[2],xmm11[3],xmm3[3]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm10[u,u,u,u],zero,zero,zero,xmm10[5,12,u,u,u,u,u,u,u]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm15, %xmm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm1, %ymm3
-; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm15[u,u,u,u,0,7,14],zero,zero,xmm15[u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm11, %xmm12, %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm2[u,u,u,u,u,u,u,u,u,3,10],zero,zero,zero,xmm2[u,u]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm12, %xmm4
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm11, %xmm4
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2,3,4,5,6],mem[7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,u,u,u],zero,zero,zero,zero,zero,xmm4[6,13]
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm13, %xmm4, %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vandnps %ymm2, %ymm1, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm3, %ymm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm8[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [5,12,0,0,5,12,0,0,5,12,0,0,5,12,0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm4, %xmm7
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm7[0],xmm2[0],xmm7[1],xmm2[1],xmm7[2],xmm2[2],xmm7[3],xmm2[3]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = xmm6[u,u,u,u],zero,zero,zero,xmm6[6,13,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm5[u,u,u,u,1,8,15],zero,zero,xmm5[u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm13, %xmm7
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = xmm6[u,u,u,u,u,u,u,u,u],zero,zero,xmm6[2,9,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm5, %xmm4
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm7[u,u,u,u],zero,zero,zero,xmm7[6,13,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm6[u,u,u,u,1,8,15],zero,zero,xmm6[u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa %xmm6, %xmm8
+; AVX1-ONLY-NEXT: vpor %xmm11, %xmm13, %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm7[u,u,u,u,u,u,u,u,u],zero,zero,xmm7[2,9,u,u,u]
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm6 = [128,4,11,128,128,4,11,128,128,4,11,128,128,4,11,128]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm4[u,u,u,u,u,u,u,u,u,4,11],zero,zero,xmm4[u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm12, %xmm12
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm12[u,u,u,u,u,u,u,u,u,9,10,11,12],zero,zero,zero
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [0,128,128,128,128,0,7,14,0,128,128,128,128,0,7,14]
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm8, %xmm13
-; AVX1-ONLY-NEXT: vpor %xmm13, %xmm12, %xmm12
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm4, %xmm13
+; AVX1-ONLY-NEXT: vpor %xmm11, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm13[u,u,u,u,u,u,u,u,u,9,10,11,12],zero,zero,zero
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [0,128,128,128,128,0,7,14,0,128,128,128,128,0,7,14]
+; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm5, %xmm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm12, %ymm12
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm12, %ymm1, %ymm12
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm12, %ymm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm12, %ymm0, %ymm12
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm2, %ymm12, %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm11[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm9, %xmm13
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm10[u,u,u,u],zero,zero,zero,xmm10[6,13,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm14[u,u,u,u,1,8,15],zero,zero,xmm14[u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm2, %xmm13, %xmm2
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm12[0,1],xmm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm11
+; AVX1-ONLY-NEXT: vpor %xmm11, %xmm13, %xmm11
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm13[u,u,u,u,u,u,u,u,u],zero,zero,xmm13[2,9,u,u,u]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm14[u,u,u,u,u,u,u,u,u,4,11],zero,zero,xmm14[u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm12, %xmm11, %xmm11
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm11[u,u,u,u,u,u,u,u,u,9,10,11,12],zero,zero,zero
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm9, %xmm7
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm11, %xmm7
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm15, %xmm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm7, %ymm3
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm13, %xmm13
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm11, %ymm11
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm11, %ymm1, %ymm11
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm11, %ymm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vandnps %ymm11, %ymm0, %ymm11
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm2, %ymm11, %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm14[3,10,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm9, %xmm11
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm10[u,u,u,u],zero,zero,zero,xmm10[6,13,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm13 = xmm15[u,u,u,u,1,8,15],zero,zero,xmm15[u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vpor %xmm11, %xmm13, %xmm11
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm12[u,u,u,u,u,u,u,u,u],zero,zero,xmm12[2,9,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm6, %xmm10, %xmm12
+; AVX1-ONLY-NEXT: vpor %xmm11, %xmm12, %xmm11
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm9 = [0,9,10,11,12,128,128,128,0,9,10,11,12,128,128,128]
+; AVX1-ONLY-NEXT: # xmm9 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm11, %xmm11
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm6, %xmm4
+; AVX1-ONLY-NEXT: vpor %xmm4, %xmm11, %xmm4
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm13, %xmm3
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm3
; AVX1-ONLY-NEXT: vandps %ymm1, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm1, %ymm3
; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm2
@@ -7423,115 +7427,111 @@
; AVX1-ONLY-NEXT: vorps %ymm3, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[4,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [6,13,0,0,6,13,0,0,6,13,0,0,6,13,0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm7, %xmm7
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm7[0],xmm2[0],xmm7[1],xmm2[1],xmm7[2],xmm2[2],xmm7[3],xmm2[3]
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm7 = [128,0,0,0,2,9,128,128,128,0,0,0,2,9,128,128]
-; AVX1-ONLY-NEXT: # xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm9, %xmm12
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm11 = xmm9[u,u,u,u],zero,zero,xmm9[0,7,14,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vpor %xmm12, %xmm11, %xmm11
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm11[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm11 = [10,128,128,3,10,128,128,3,10,128,128,3,10,128,128,3]
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm6, %xmm10
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm2[4,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm4 = [6,13,0,0,6,13,0,0,6,13,0,0,6,13,0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm2, %xmm11
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm11[0],xmm3[0],xmm11[1],xmm3[1],xmm11[2],xmm3[2],xmm11[3],xmm3[3]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm12 = xmm8[u,u,u,u,2,9],zero,zero,zero,xmm8[u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm11 = [14,0,0,0,128,128,0,7,14,0,0,0,128,128,0,7]
+; AVX1-ONLY-NEXT: # xmm11 = mem[0,0]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm2, %xmm2
+; AVX1-ONLY-NEXT: vpor %xmm2, %xmm12, %xmm2
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm3 = [10,128,128,3,10,128,128,3,10,128,128,3,10,128,128,3]
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm7, %xmm8
; AVX1-ONLY-NEXT: vbroadcastss {{.*#+}} xmm12 = [128,5,12,128,128,5,12,128,128,5,12,128,128,5,12,128]
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm4, %xmm6
-; AVX1-ONLY-NEXT: vpor %xmm6, %xmm10, %xmm6
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = [0,9,10,11,12,128,128,128,0,9,10,11,12,128,128,128]
-; AVX1-ONLY-NEXT: # xmm4 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm6, %xmm6
-; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm10 = [0,128,128,128,128,1,8,15,0,128,128,128,128,1,8,15]
-; AVX1-ONLY-NEXT: # xmm10 = mem[0,0]
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm8, %xmm9
-; AVX1-ONLY-NEXT: vpor %xmm6, %xmm9, %xmm6
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm5, %xmm8
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm6, %ymm6
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm1, %ymm6
-; AVX1-ONLY-NEXT: vorps %ymm6, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: vandnps %ymm6, %ymm0, %ymm6
-; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vorps %ymm6, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm6 = xmm5[4,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm5, %xmm8
-; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm7, %xmm5, %xmm7
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm8 = xmm5[u,u,u,u],zero,zero,xmm5[0,7,14,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm7, %xmm7
; AVX1-ONLY-NEXT: vpor %xmm7, %xmm8, %xmm7
-; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm13, %xmm5
-; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm14, %xmm7
-; AVX1-ONLY-NEXT: vpor %xmm5, %xmm7, %xmm5
+; AVX1-ONLY-NEXT: vpshufb %xmm9, %xmm7, %xmm7
+; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm8 = [0,128,128,128,128,1,8,15,0,128,128,128,128,1,8,15]
+; AVX1-ONLY-NEXT: # xmm8 = mem[0,0]
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm5, %xmm9
+; AVX1-ONLY-NEXT: vpor %xmm7, %xmm9, %xmm7
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm5, %xmm5
-; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT: vpshufb %xmm10, %xmm4, %xmm7
-; AVX1-ONLY-NEXT: vpor %xmm7, %xmm5, %xmm5
-; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm15, %xmm3
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm5, %ymm3
-; AVX1-ONLY-NEXT: vandps %ymm1, %ymm6, %ymm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm7, %ymm5
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm1, %ymm5
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vandnps %ymm5, %ymm0, %ymm5
+; AVX1-ONLY-NEXT: vandps %ymm0, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vorps %ymm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm5 = xmm14[4,11,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm7, %xmm7
+; AVX1-ONLY-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3]
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm7 = xmm15[u,u,u,u,2,9],zero,zero,zero,xmm15[u,u,u,u,u,u,u]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm11, %xmm9, %xmm9
+; AVX1-ONLY-NEXT: vpor %xmm7, %xmm9, %xmm7
+; AVX1-ONLY-NEXT: vpblendw {{.*#+}} xmm5 = xmm5[0,1],xmm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX1-ONLY-NEXT: vpshufb %xmm3, %xmm7, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm12, %xmm10, %xmm7
+; AVX1-ONLY-NEXT: vpor %xmm3, %xmm7, %xmm3
+; AVX1-ONLY-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,u,u,9,10,11,12],zero,zero,zero
+; AVX1-ONLY-NEXT: vpshufb %xmm8, %xmm6, %xmm7
+; AVX1-ONLY-NEXT: vpor %xmm7, %xmm3, %xmm3
+; AVX1-ONLY-NEXT: vpshufb %xmm4, %xmm13, %xmm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vandps %ymm1, %ymm5, %ymm4
; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vorps %ymm1, %ymm4, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rsp), %ymm0, %ymm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: vandnps %ymm3, %ymm0, %ymm3
; AVX1-ONLY-NEXT: vandps %ymm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vorps %ymm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rsi)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rsi)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rdx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r8)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r8)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%r9)
-; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%r9)
+; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm1, (%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rax)
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, (%rax)
-; AVX1-ONLY-NEXT: addq $744, %rsp # imm = 0x2E8
+; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%rax)
+; AVX1-ONLY-NEXT: addq $728, %rsp # imm = 0x2D8
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-SLOW-LABEL: load_i8_stride7_vf64:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $744, %rsp # imm = 0x2E8
-; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm10
-; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm6
-; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = <255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0>
-; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm1, %ymm2, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm13
-; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm7
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm1, %ymm8
+; AVX2-SLOW-NEXT: subq $760, %rsp # imm = 0x2F8
+; AVX2-SLOW-NEXT: vmovdqa (%rdi), %ymm8
+; AVX2-SLOW-NEXT: vmovdqa 32(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovdqa 64(%rdi), %ymm7
+; AVX2-SLOW-NEXT: vmovdqa 320(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vmovdqa 224(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqa 256(%rdi), %ymm10
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0>
+; AVX2-SLOW-NEXT: vpblendvb %ymm2, %ymm1, %ymm10, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm14
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, %ymm12
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <128,128,128,5,12,128,128,1,8,15,u,u,u,u,u,u>
@@ -7540,8 +7540,8 @@
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm9 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
-; AVX2-SLOW-NEXT: vpblendvb %ymm9, %ymm5, %ymm4, %ymm3
-; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm14
+; AVX2-SLOW-NEXT: vpblendvb %ymm9, %ymm4, %ymm5, %ymm3
+; AVX2-SLOW-NEXT: vmovdqa %ymm9, %ymm15
; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm9
; AVX2-SLOW-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm11
@@ -7553,33 +7553,37 @@
; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm4, %ymm5
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
; AVX2-SLOW-NEXT: vpblendvb %ymm4, %ymm0, %ymm5, %ymm0
-; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm12, %ymm6, %ymm5
-; AVX2-SLOW-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm15
-; AVX2-SLOW-NEXT: vmovdqu %ymm6, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm8, %ymm6, %ymm5
+; AVX2-SLOW-NEXT: vmovdqa %ymm8, %ymm13
+; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm0
+; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm5, %xmm6
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm6, %xmm1
-; AVX2-SLOW-NEXT: vmovdqa 288(%rdi), %ymm6
+; AVX2-SLOW-NEXT: vmovdqa 96(%rdi), %ymm6
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm2
; AVX2-SLOW-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm10, %ymm6, %ymm2
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm6, %ymm7, %ymm2
+; AVX2-SLOW-NEXT: vmovdqa %ymm6, %ymm8
; AVX2-SLOW-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm5
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm5[2],ymm2[3,4],ymm5[5],ymm2[6,7,8,9],ymm5[10],ymm2[11,12],ymm5[13],ymm2[14,15]
; AVX2-SLOW-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-SLOW-NEXT: vpblendvb %ymm4, %ymm1, %ymm2, %ymm14
-; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm11, %ymm9, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm4, %ymm1, %ymm2, %ymm1
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm9, %ymm11, %ymm1
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7,8,9],ymm2[10],ymm1[11,12,13],ymm2[14],ymm1[15]
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm8, %ymm7, %ymm4
-; AVX2-SLOW-NEXT: vmovdqa %ymm1, %ymm3
+; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm12, %ymm10, %ymm4
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, %ymm6
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm1 = <128,128,128,6,13,128,128,2,9,u,u,u,u,u,u,u>
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm5
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm7 = <1,8,15,128,128,4,11,128,128,u,u,u,u,u,u,u>
-; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm10 = <1,8,15,128,128,4,11,128,128,u,u,u,u,u,u,u>
+; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm4, %xmm4
; AVX2-SLOW-NEXT: vpor %xmm5, %xmm4, %xmm4
; AVX2-SLOW-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,8,15,0,0,0,0,0,0,0,7,14,5,12,3,10,1,8,15,0,0,0,0,0,0,0,7,14,5,12,3,10]
; AVX2-SLOW-NEXT: # ymm5 = mem[0,1,0,1]
@@ -7587,98 +7591,100 @@
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm9 = <0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,u,u,u,u,u,u,u,u,u,u,u,u,u>
; AVX2-SLOW-NEXT: vpblendvb %ymm9, %ymm2, %ymm4, %ymm2
; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm6, %ymm10, %ymm2
+; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm7, %ymm8, %ymm2
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm4
; AVX2-SLOW-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4,5],ymm4[6],ymm2[7,8,9],ymm4[10],ymm2[11,12,13],ymm4[14],ymm2[15]
; AVX2-SLOW-NEXT: vpshufb %ymm5, %ymm2, %ymm2
-; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm12, %ymm15, %ymm4
+; AVX2-SLOW-NEXT: vpblendvb %ymm6, %ymm13, %ymm0, %ymm4
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm1
-; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vpshufb %xmm10, %xmm4, %xmm4
; AVX2-SLOW-NEXT: vpor %xmm1, %xmm4, %xmm1
-; AVX2-SLOW-NEXT: vpblendvb %ymm9, %ymm2, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm9, %ymm2, %ymm1, %ymm11
+; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm1
+; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm13
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm0 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm13, %ymm4
+; AVX2-SLOW-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %ymm1, %ymm7
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm3
-; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm4
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm10 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
-; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm3, %ymm4, %ymm2
-; AVX2-SLOW-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm2, %xmm7
+; AVX2-SLOW-NEXT: vmovdqa %ymm0, %ymm10
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm9 = <u,u,u,128,128,3,10,128,128,128,6,13,u,u,u,u>
-; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm7, %xmm7
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm11 = <u,u,u,5,12,128,128,1,8,15,128,128,u,u,u,u>
-; AVX2-SLOW-NEXT: vpshufb %xmm11, %xmm2, %xmm2
-; AVX2-SLOW-NEXT: vpor %xmm7, %xmm2, %xmm2
-; AVX2-SLOW-NEXT: vpbroadcastw {{.*#+}} xmm13 = [2,9,2,9,2,9,2,9,2,9,2,9,2,9,2,9]
-; AVX2-SLOW-NEXT: vmovdqa 208(%rdi), %xmm1
-; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm1, %xmm7
-; AVX2-SLOW-NEXT: vmovdqa %xmm1, %xmm6
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm5, %xmm5
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,u,5,12,128,128,1,8,15,128,128,u,u,u,u>
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vpor %xmm5, %xmm4, %xmm4
+; AVX2-SLOW-NEXT: vpbroadcastw {{.*#+}} xmm2 = [2,9,2,9,2,9,2,9,2,9,2,9,2,9,2,9]
+; AVX2-SLOW-NEXT: vmovdqa 432(%rdi), %xmm1
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm1, %xmm5
; AVX2-SLOW-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpbroadcastw {{.*#+}} xmm1 = [4,11,4,11,4,11,4,11,4,11,4,11,4,11,4,11]
-; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %xmm8
-; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm8, %xmm15
-; AVX2-SLOW-NEXT: vmovdqa %xmm8, %xmm12
-; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm15[0],xmm7[0],xmm15[1],xmm7[1],xmm15[2],xmm7[2],xmm15[3],xmm7[3]
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm7[7]
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm15 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm0, %ymm2, %ymm0
+; AVX2-SLOW-NEXT: vpbroadcastw {{.*#+}} xmm0 = [4,11,4,11,4,11,4,11,4,11,4,11,4,11,4,11]
+; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %xmm3
+; AVX2-SLOW-NEXT: vpshufb %xmm0, %xmm3, %xmm14
+; AVX2-SLOW-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %xmm0, %xmm8
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm14 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX2-SLOW-NEXT: vpblendvb %ymm14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa 384(%rdi), %ymm2
-; AVX2-SLOW-NEXT: vmovdqa 352(%rdi), %ymm5
-; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm2, %ymm5, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm10
-; AVX2-SLOW-NEXT: vmovdqa %ymm2, %ymm7
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm8
-; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm8, %xmm8
-; AVX2-SLOW-NEXT: vpshufb %xmm11, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vpor %xmm0, %xmm8, %xmm0
-; AVX2-SLOW-NEXT: vmovdqa 432(%rdi), %xmm5
-; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm5, %xmm8
-; AVX2-SLOW-NEXT: vmovdqa 416(%rdi), %xmm2
-; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm2, %xmm9
-; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; AVX2-SLOW-NEXT: vmovdqa 160(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovdqa 128(%rdi), %ymm5
+; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm4, %ymm5, %ymm0
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm15
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm15, %xmm9
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vpor %xmm0, %xmm9, %xmm0
+; AVX2-SLOW-NEXT: vmovdqa 208(%rdi), %xmm10
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm10, %xmm9
+; AVX2-SLOW-NEXT: vmovdqa %xmm10, %xmm2
+; AVX2-SLOW-NEXT: vmovdqa 192(%rdi), %xmm10
+; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm10, %xmm12
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1],xmm12[2],xmm9[2],xmm12[3],xmm9[3]
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
+; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm9[7]
+; AVX2-SLOW-NEXT: vpblendvb %ymm14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
+; AVX2-SLOW-NEXT: vpblendvb %ymm6, %ymm7, %ymm13, %ymm0
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,6,13,128,128,2,9,128,128,128,u,u,u,u>
+; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm0, %xmm9
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,u,128,128,4,11,128,128,0,7,14,u,u,u,u>
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vpor %xmm0, %xmm9, %xmm0
+; AVX2-SLOW-NEXT: vpbroadcastw {{.*#+}} xmm9 = [3,10,3,10,3,10,3,10,3,10,3,10,3,10,3,10]
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm1, %xmm13
+; AVX2-SLOW-NEXT: vpbroadcastw {{.*#+}} xmm15 = [5,12,5,12,5,12,5,12,5,12,5,12,5,12,5,12]
+; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm3, %xmm8
+; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm13[0],xmm8[1],xmm13[1],xmm8[2],xmm13[2],xmm8[3],xmm13[3]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm14, %ymm0, %ymm0
+; AVX2-SLOW-NEXT: vpblendvb %ymm14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm3, %ymm4, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,6,13,128,128,2,9,128,128,128,u,u,u,u>
-; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm0, %xmm8
+; AVX2-SLOW-NEXT: vpblendvb %ymm6, %ymm4, %ymm5, %ymm0
+; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm0, %xmm3
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm9 = <u,u,u,128,128,4,11,128,128,0,7,14,u,u,u,u>
-; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vpor %xmm0, %xmm8, %xmm0
-; AVX2-SLOW-NEXT: vpbroadcastw {{.*#+}} xmm8 = [3,10,3,10,3,10,3,10,3,10,3,10,3,10,3,10]
-; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm6, %xmm11
-; AVX2-SLOW-NEXT: vpbroadcastw {{.*#+}} xmm13 = [5,12,5,12,5,12,5,12,5,12,5,12,5,12,5,12]
-; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm12, %xmm14
-; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm14[0],xmm11[0],xmm14[1],xmm11[1],xmm14[2],xmm11[2],xmm14[3],xmm11[3]
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
-; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm11[7]
-; AVX2-SLOW-NEXT: vpblendvb %ymm15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendvb %ymm1, %ymm7, %ymm10, %ymm0
-; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm0, %xmm3
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm0, %xmm0
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vpor %xmm3, %xmm0, %xmm0
-; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm5, %xmm3
-; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm2, %xmm6
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm2, %xmm3
+; AVX2-SLOW-NEXT: vmovdqa %xmm2, %xmm13
+; AVX2-SLOW-NEXT: vpshufb %xmm15, %xmm10, %xmm6
; AVX2-SLOW-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-SLOW-NEXT: vpblendvb %ymm15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: vpblendvb %ymm14, %ymm11, %ymm0, %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm4 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
-; AVX2-SLOW-NEXT: vpblendvb %ymm4, %ymm10, %ymm7, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa %ymm10, %ymm14
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendvb %ymm2, %ymm11, %ymm7, %ymm0
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,128,128,128,6,13,128,128,2,9,u,u,u,u,u>
; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm1, %xmm1
@@ -7686,38 +7692,34 @@
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm1 = [0,0,0,128,128,128,5,12,0,0,0,128,128,128,5,12]
-; AVX2-SLOW-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm5, %xmm8
+; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm15, %xmm8
; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm9 = [0,0,0,0,7,14,128,128,0,0,0,0,7,14,128,128]
-; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm2, %xmm11
-; AVX2-SLOW-NEXT: vmovdqa %xmm2, %xmm13
-; AVX2-SLOW-NEXT: vpor %xmm8, %xmm11, %xmm8
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm11
+; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm14, %xmm12
+; AVX2-SLOW-NEXT: vpor %xmm8, %xmm12, %xmm8
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm12
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX2-SLOW-NEXT: vbroadcasti128 {{.*#+}} ymm10 = [0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; AVX2-SLOW-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm11, %ymm8, %ymm8
+; AVX2-SLOW-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; AVX2-SLOW-NEXT: # ymm0 = mem[0,1,0,1]
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm12, %ymm8, %ymm8
; AVX2-SLOW-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendvb %ymm4, %ymm2, %ymm15, %ymm8
-; AVX2-SLOW-NEXT: vextracti128 $1, %ymm8, %xmm11
-; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm11, %xmm3
+; AVX2-SLOW-NEXT: vpblendvb %ymm2, %ymm5, %ymm4, %ymm8
+; AVX2-SLOW-NEXT: vextracti128 $1, %ymm8, %xmm12
+; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm12, %xmm3
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm8, %xmm6
; AVX2-SLOW-NEXT: vpor %xmm3, %xmm6, %xmm3
-; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm0, %xmm1
-; AVX2-SLOW-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm12, %xmm6
+; AVX2-SLOW-NEXT: vmovdqa %xmm13, %xmm2
+; AVX2-SLOW-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vpshufb %xmm1, %xmm13, %xmm1
+; AVX2-SLOW-NEXT: vpshufb %xmm9, %xmm10, %xmm6
; AVX2-SLOW-NEXT: vpor %xmm1, %xmm6, %xmm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm3, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm3, %ymm1, %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
-; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm14, %ymm7, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa %ymm14, %ymm4
-; AVX2-SLOW-NEXT: vmovdqa %ymm7, %ymm14
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm11, %ymm7, %ymm1
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,2,9,128,128,128,5,12,128,128,u,u,u,u,u>
; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm1, %xmm6
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm1
@@ -7725,70 +7727,70 @@
; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vpor %xmm6, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm6 = [0,0,0,128,128,128,6,13,0,0,0,128,128,128,6,13]
-; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm5, %xmm9
-; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm11 = [0,0,0,1,8,15,128,128,0,0,0,1,8,15,128,128]
-; AVX2-SLOW-NEXT: vmovdqa %xmm13, %xmm5
-; AVX2-SLOW-NEXT: vpshufb %xmm11, %xmm13, %xmm13
+; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm15, %xmm9
+; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm12 = [0,0,0,1,8,15,128,128,0,0,0,1,8,15,128,128]
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm14, %xmm13
; AVX2-SLOW-NEXT: vpor %xmm9, %xmm13, %xmm9
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm1, %ymm9, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm9, %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm2, %ymm15, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa %ymm15, %ymm2
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm5, %ymm4, %ymm1
; AVX2-SLOW-NEXT: vpshufb %xmm7, %xmm1, %xmm3
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vpor %xmm3, %xmm1, %xmm1
-; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm0, %xmm3
-; AVX2-SLOW-NEXT: vpshufb %xmm11, %xmm12, %xmm6
+; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm2, %xmm3
+; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm10, %xmm6
+; AVX2-SLOW-NEXT: vmovdqa %xmm10, %xmm2
; AVX2-SLOW-NEXT: vpor %xmm3, %xmm6, %xmm3
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm1, %ymm3, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm3, %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255>
-; AVX2-SLOW-NEXT: vmovdqa %ymm14, %ymm7
-; AVX2-SLOW-NEXT: vmovdqa %ymm4, %ymm0
-; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm4, %ymm14, %ymm1
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,3,10,128,128,128,6,13,128,128,u,u,u,u,u>
-; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm1, %xmm6
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm11, %ymm7, %ymm1
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, %ymm15
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,3,10,128,128,128,6,13,128,128,u,u,u,u,u>
+; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm1, %xmm6
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,128,128,1,8,15,128,128,4,11,u,u,u,u,u>
; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vpor %xmm6, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm6 = [0,0,0,2,9,128,128,128,0,0,0,2,9,128,128,128]
-; AVX2-SLOW-NEXT: vmovdqa %xmm5, %xmm9
-; AVX2-SLOW-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm5, %xmm11
+; AVX2-SLOW-NEXT: vmovdqa %xmm14, %xmm9
+; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm14, %xmm12
; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm13 = [0,0,0,128,128,0,7,14,0,0,0,128,128,0,7,14]
-; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm4, %xmm14
-; AVX2-SLOW-NEXT: vpor %xmm11, %xmm14, %xmm11
+; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm10, %xmm14
+; AVX2-SLOW-NEXT: vpor %xmm12, %xmm14, %xmm12
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-SLOW-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
-; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm1, %ymm11, %ymm1
+; AVX2-SLOW-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm12, %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm15, %ymm2, %ymm1
-; AVX2-SLOW-NEXT: vpshufb %xmm12, %xmm1, %xmm3
+; AVX2-SLOW-NEXT: vpblendvb %ymm15, %ymm5, %ymm4, %ymm1
+; AVX2-SLOW-NEXT: vmovdqa %ymm5, %ymm15
+; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm1, %xmm3
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vpor %xmm3, %xmm1, %xmm1
+; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm2, %xmm3
+; AVX2-SLOW-NEXT: vmovdqa %xmm2, %xmm5
+; AVX2-SLOW-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm14, %xmm3
-; AVX2-SLOW-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm5, %xmm6
+; AVX2-SLOW-NEXT: vpshufb %xmm13, %xmm14, %xmm6
; AVX2-SLOW-NEXT: vpor %xmm3, %xmm6, %xmm3
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm1, %ymm3, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm3, %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm12 = <255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0>
-; AVX2-SLOW-NEXT: vpblendvb %ymm12, %ymm0, %ymm7, %ymm2
-; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm13 = <255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u>
-; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm7, %ymm0, %ymm1
+; AVX2-SLOW-NEXT: vmovdqa %ymm11, %ymm3
+; AVX2-SLOW-NEXT: vpblendvb %ymm12, %ymm11, %ymm7, %ymm1
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm11 = <255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u>
+; AVX2-SLOW-NEXT: vpblendvb %ymm11, %ymm7, %ymm3, %ymm1
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,128,128,2,9,128,128,128,5,12,u,u,u,u,u>
; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm2, %xmm2
@@ -7798,50 +7800,49 @@
; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [0,0,0,3,10,128,128,128,0,0,0,3,10,128,128,128]
; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm9, %xmm7
; AVX2-SLOW-NEXT: vpbroadcastq {{.*#+}} xmm8 = [0,0,0,128,128,1,8,15,0,0,0,128,128,1,8,15]
-; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm4, %xmm11
-; AVX2-SLOW-NEXT: vpor %xmm7, %xmm11, %xmm7
+; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm10, %xmm13
+; AVX2-SLOW-NEXT: vpor %xmm7, %xmm13, %xmm7
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm1, %ymm7, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm7, %ymm1
; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendvb %ymm12, %ymm15, %ymm1, %ymm0
-; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa %ymm13, %ymm12
-; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm1, %ymm15, %ymm1
+; AVX2-SLOW-NEXT: vpblendvb %ymm12, %ymm15, %ymm4, %ymm7
+; AVX2-SLOW-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa %ymm11, %ymm12
+; AVX2-SLOW-NEXT: vpblendvb %ymm11, %ymm4, %ymm15, %ymm1
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm1, %xmm4
; AVX2-SLOW-NEXT: vpshufb %xmm3, %xmm4, %xmm3
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm1, %xmm1
; AVX2-SLOW-NEXT: vpor %xmm3, %xmm1, %xmm1
-; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm14, %xmm2
-; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm5, %xmm3
+; AVX2-SLOW-NEXT: vpshufb %xmm2, %xmm5, %xmm2
+; AVX2-SLOW-NEXT: vpshufb %xmm8, %xmm14, %xmm3
; AVX2-SLOW-NEXT: vpor %xmm2, %xmm3, %xmm2
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-SLOW-NEXT: vpblendvb %ymm10, %ymm1, %ymm2, %ymm0
+; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm0
; AVX2-SLOW-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
-; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm2, %ymm1, %ymm0
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm13 = <u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255>
-; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm2, %ymm1, %ymm15
-; AVX2-SLOW-NEXT: vpblendvb %ymm12, %ymm1, %ymm2, %ymm11
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm4 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
-; AVX2-SLOW-NEXT: vpblendvb %ymm4, %ymm1, %ymm2, %ymm3
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
-; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
-; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vpblendvb %ymm3, %ymm2, %ymm1, %ymm3
-; AVX2-SLOW-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm1, %ymm2, %ymm3
-; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm1, %ymm2, %ymm7
-; AVX2-SLOW-NEXT: vpblendvb %ymm12, %ymm2, %ymm1, %ymm14
-; AVX2-SLOW-NEXT: vpblendvb %ymm4, %ymm2, %ymm1, %ymm1
-; AVX2-SLOW-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
+; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm1, %ymm10, %ymm0
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm13 = <u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255>
+; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm1, %ymm10, %ymm15
+; AVX2-SLOW-NEXT: vpblendvb %ymm11, %ymm10, %ymm1, %ymm11
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm4 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
+; AVX2-SLOW-NEXT: vpblendvb %ymm4, %ymm10, %ymm1, %ymm2
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
+; AVX2-SLOW-NEXT: vpblendvb %ymm2, %ymm10, %ymm1, %ymm1
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-SLOW-NEXT: vpblendvb %ymm2, %ymm1, %ymm10, %ymm2
+; AVX2-SLOW-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm10, %ymm1, %ymm3
+; AVX2-SLOW-NEXT: vpblendvb %ymm13, %ymm10, %ymm1, %ymm7
+; AVX2-SLOW-NEXT: vpblendvb %ymm12, %ymm1, %ymm10, %ymm14
+; AVX2-SLOW-NEXT: vpblendvb %ymm4, %ymm1, %ymm10, %ymm1
+; AVX2-SLOW-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm6 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
@@ -7940,7 +7941,7 @@
; AVX2-SLOW-NEXT: # ymm8 = mem[0,1,0,1]
; AVX2-SLOW-NEXT: vpshufb %ymm8, %ymm7, %ymm7
; AVX2-SLOW-NEXT: vpblendvb %ymm0, %ymm4, %ymm7, %ymm4
-; AVX2-SLOW-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vextracti128 $1, %ymm9, %xmm7
; AVX2-SLOW-NEXT: vpshufb %xmm5, %xmm7, %xmm5
; AVX2-SLOW-NEXT: vpshufb %xmm6, %xmm9, %xmm6
@@ -8035,42 +8036,42 @@
; AVX2-SLOW-NEXT: # ymm10 = ymm0[0],mem[1,2,3,4,5,6,7],ymm0[8],mem[9,10,11,12,13,14,15]
; AVX2-SLOW-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm10, 32(%rsi)
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm10, (%rsi)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovaps %ymm10, 32(%rdx)
+; AVX2-SLOW-NEXT: vmovaps %ymm10, 32(%rsi)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm10, (%rdx)
-; AVX2-SLOW-NEXT: vmovdqa %ymm6, 32(%rcx)
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-SLOW-NEXT: vmovaps %ymm10, 32(%rdx)
; AVX2-SLOW-NEXT: vmovdqa %ymm8, (%rcx)
-; AVX2-SLOW-NEXT: vmovdqa %ymm2, 32(%r8)
+; AVX2-SLOW-NEXT: vmovdqa %ymm6, 32(%rcx)
; AVX2-SLOW-NEXT: vmovdqa %ymm9, (%r8)
-; AVX2-SLOW-NEXT: vmovdqa %ymm3, 32(%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm2, 32(%r8)
; AVX2-SLOW-NEXT: vmovdqa %ymm1, (%r9)
+; AVX2-SLOW-NEXT: vmovdqa %ymm3, 32(%r9)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovdqa %ymm4, 32(%rax)
; AVX2-SLOW-NEXT: vmovdqa %ymm5, (%rax)
+; AVX2-SLOW-NEXT: vmovdqa %ymm4, 32(%rax)
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-SLOW-NEXT: vmovdqa %ymm7, 32(%rax)
; AVX2-SLOW-NEXT: vmovdqa %ymm0, (%rax)
-; AVX2-SLOW-NEXT: addq $744, %rsp # imm = 0x2E8
+; AVX2-SLOW-NEXT: vmovdqa %ymm7, 32(%rax)
+; AVX2-SLOW-NEXT: addq $760, %rsp # imm = 0x2F8
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
; AVX2-FAST-LABEL: load_i8_stride7_vf64:
; AVX2-FAST: # %bb.0:
-; AVX2-FAST-NEXT: subq $712, %rsp # imm = 0x2C8
-; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm7
-; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm6
-; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm8
-; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm4
-; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm5
+; AVX2-FAST-NEXT: subq $728, %rsp # imm = 0x2D8
+; AVX2-FAST-NEXT: vmovdqa (%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovdqa 32(%rdi), %ymm8
+; AVX2-FAST-NEXT: vmovdqa 64(%rdi), %ymm7
+; AVX2-FAST-NEXT: vmovdqa 320(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm5
+; AVX2-FAST-NEXT: vmovdqa 224(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqa 256(%rdi), %ymm2
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0>
; AVX2-FAST-NEXT: vpblendvb %ymm3, %ymm1, %ymm2, %ymm0
-; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm15
+; AVX2-FAST-NEXT: vmovdqa %ymm3, %ymm14
; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm9
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm10
@@ -8081,8 +8082,8 @@
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = <0,7,14,128,128,3,10,128,128,128,u,u,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
; AVX2-FAST-NEXT: vpor %xmm3, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm5, %ymm4, %ymm3
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
+; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm4, %ymm5, %ymm3
; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm11
; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm13
@@ -8095,24 +8096,24 @@
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm12 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
; AVX2-FAST-NEXT: vpblendvb %ymm12, %ymm0, %ymm4, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm6, %ymm8, %ymm4
-; AVX2-FAST-NEXT: vmovdqu %ymm8, (%rsp) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm6, %ymm8, %ymm4
+; AVX2-FAST-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm5
; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm6
; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm6, %xmm1
-; AVX2-FAST-NEXT: vmovdqa 288(%rdi), %ymm6
+; AVX2-FAST-NEXT: vmovdqa 96(%rdi), %ymm6
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm4, %xmm2
; AVX2-FAST-NEXT: vpor %xmm1, %xmm2, %xmm1
; AVX2-FAST-NEXT: vmovdqa %ymm7, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm7, %ymm6, %ymm2
+; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm6, %ymm7, %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm4
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7,8,9],ymm4[10],ymm2[11,12],ymm4[13],ymm2[14,15]
; AVX2-FAST-NEXT: vpshufb %ymm3, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpblendvb %ymm12, %ymm1, %ymm2, %ymm12
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm13, %ymm11, %ymm2
+; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm11, %ymm13, %ymm2
; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4,5],ymm3[6],ymm2[7,8,9],ymm3[10],ymm2[11,12,13],ymm3[14],ymm2[15]
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
@@ -8128,9 +8129,8 @@
; AVX2-FAST-NEXT: # ymm4 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm2
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,u,u,u,u,u,u,u,u,u,u,u,u,u>
-; AVX2-FAST-NEXT: vpblendvb %ymm1, %ymm2, %ymm3, %ymm2
-; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm6, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vpblendvb %ymm1, %ymm2, %ymm3, %ymm13
+; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm0, %ymm6, %ymm2
; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4,5],ymm3[6],ymm2[7,8,9],ymm3[10],ymm2[11,12,13],ymm3[14],ymm2[15]
; AVX2-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm2
@@ -8140,48 +8140,48 @@
; AVX2-FAST-NEXT: vpshufb %xmm7, %xmm3, %xmm3
; AVX2-FAST-NEXT: vpor %xmm4, %xmm3, %xmm3
; AVX2-FAST-NEXT: vpblendvb %ymm1, %ymm2, %ymm3, %ymm7
-; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm1
+; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm3
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm4, %ymm2
-; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm15
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm3, %ymm4
+; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm2
+; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm14
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,u,128,128,3,10,128,128,128,6,13,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm3, %xmm3
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm5, %xmm5
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm9 = <u,u,u,5,12,128,128,1,8,15,128,128,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm2, %xmm2
-; AVX2-FAST-NEXT: vpor %xmm3, %xmm2, %xmm2
-; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %ymm0
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm4, %xmm4
+; AVX2-FAST-NEXT: vpor %xmm5, %xmm4, %xmm4
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %ymm0
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = [1,2,0,2,1,2,4,6]
-; AVX2-FAST-NEXT: vpermd %ymm0, %ymm10, %ymm3
+; AVX2-FAST-NEXT: vpermd %ymm0, %ymm10, %ymm5
; AVX2-FAST-NEXT: vmovdqa %ymm0, %ymm6
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm11 = [0,7,10,13,0,7,10,13,0,7,10,13,0,7,10,13,0,7,10,13,0,7,10,13,0,7,10,13,0,7,10,13]
-; AVX2-FAST-NEXT: vpshufb %ymm11, %ymm3, %ymm3
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX2-FAST-NEXT: vpblendvb %ymm3, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpshufb %ymm11, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX2-FAST-NEXT: vpblendvb %ymm1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa 384(%rdi), %ymm2
-; AVX2-FAST-NEXT: vmovdqa 352(%rdi), %ymm13
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm2, %ymm13, %ymm0
-; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm5
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm15
-; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm15, %xmm8
+; AVX2-FAST-NEXT: vmovdqa 160(%rdi), %ymm4
+; AVX2-FAST-NEXT: vmovdqa 128(%rdi), %ymm5
+; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm4, %ymm5, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm14
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm14, %xmm8
; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm0, %xmm0
; AVX2-FAST-NEXT: vpor %xmm0, %xmm8, %xmm0
-; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %ymm9
+; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %ymm9
; AVX2-FAST-NEXT: vpermd %ymm9, %ymm10, %ymm8
-; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm15
+; AVX2-FAST-NEXT: vmovdqa %ymm9, %ymm14
; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufb %ymm11, %ymm8, %ymm8
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-FAST-NEXT: vpblendvb %ymm3, %ymm12, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpblendvb %ymm1, %ymm12, %ymm0, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm1, %ymm4, %ymm0
-; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm2
+; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm2, %ymm3, %ymm0
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,u,6,13,128,128,2,9,128,128,128,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm0, %xmm9
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm0
@@ -8194,23 +8194,23 @@
; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm12 = [1,4,11,14,1,4,11,14,1,4,11,14,1,4,11,14,1,4,11,14,1,4,11,14,1,4,11,14,1,4,11,14]
; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm11, %ymm11
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm11[7]
-; AVX2-FAST-NEXT: vpblendvb %ymm3, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpblendvb %ymm1, %ymm13, %ymm0, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm5, %ymm13, %ymm0
+; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm4, %ymm5, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm0, %xmm6
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm0
; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm0, %xmm0
; AVX2-FAST-NEXT: vpor %xmm6, %xmm0, %xmm0
-; AVX2-FAST-NEXT: vpermd %ymm15, %ymm9, %ymm6
+; AVX2-FAST-NEXT: vpermd %ymm14, %ymm9, %ymm6
; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm6, %ymm6
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm6[7]
-; AVX2-FAST-NEXT: vpblendvb %ymm3, %ymm7, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpblendvb %ymm1, %ymm7, %ymm0, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
-; AVX2-FAST-NEXT: vmovdqa %ymm5, %ymm15
-; AVX2-FAST-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm13, %ymm5, %ymm0
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
+; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm7
+; AVX2-FAST-NEXT: vpblendvb %ymm13, %ymm3, %ymm2, %ymm0
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm6 = <u,u,128,128,128,6,13,128,128,2,9,u,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm1, %xmm1
@@ -8218,22 +8218,21 @@
; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm0, %xmm0
; AVX2-FAST-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm9 = [0,0,0,128,128,128,5,12,0,0,0,128,128,128,5,12]
-; AVX2-FAST-NEXT: vmovdqa 432(%rdi), %xmm5
-; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm5, %xmm10
+; AVX2-FAST-NEXT: vmovdqa 432(%rdi), %xmm1
+; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm1, %xmm10
+; AVX2-FAST-NEXT: vmovdqa %xmm1, %xmm14
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm11 = [0,0,0,0,7,14,128,128,0,0,0,0,7,14,128,128]
-; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %xmm7
-; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm7, %xmm12
+; AVX2-FAST-NEXT: vmovdqa 416(%rdi), %xmm15
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm15, %xmm12
; AVX2-FAST-NEXT: vpor %xmm10, %xmm12, %xmm10
; AVX2-FAST-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm12
; AVX2-FAST-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; AVX2-FAST-NEXT: # ymm0 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm12, %ymm10, %ymm1
+; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; AVX2-FAST-NEXT: # ymm5 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm12, %ymm10, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm4, %ymm3
-; AVX2-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa %ymm2, %ymm4
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm3, %ymm2, %ymm10
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendvb %ymm13, %ymm0, %ymm4, %ymm10
; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm12
; AVX2-FAST-NEXT: vpshufb %xmm6, %xmm12, %xmm6
; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm10, %xmm8
@@ -8241,15 +8240,17 @@
; AVX2-FAST-NEXT: vmovdqa 208(%rdi), %xmm1
; AVX2-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-NEXT: vpshufb %xmm9, %xmm1, %xmm9
-; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %xmm6
-; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm6, %xmm11
+; AVX2-FAST-NEXT: vmovdqa 192(%rdi), %xmm1
+; AVX2-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm1, %xmm11
; AVX2-FAST-NEXT: vpor %xmm9, %xmm11, %xmm9
; AVX2-FAST-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
; AVX2-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm10, %ymm9, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm10, %ymm9, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
-; AVX2-FAST-NEXT: vpblendvb %ymm2, %ymm13, %ymm15, %ymm9
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
+; AVX2-FAST-NEXT: vpblendvb %ymm1, %ymm3, %ymm2, %ymm9
+; AVX2-FAST-NEXT: vmovdqa %ymm1, %ymm8
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm10 = <u,u,2,9,128,128,128,5,12,128,128,u,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm9, %xmm11
; AVX2-FAST-NEXT: vextracti128 $1, %ymm9, %xmm9
@@ -8257,65 +8258,67 @@
; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm9, %xmm9
; AVX2-FAST-NEXT: vpor %xmm11, %xmm9, %xmm9
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm11 = [0,0,0,128,128,128,6,13,0,0,0,128,128,128,6,13]
-; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm5, %xmm15
-; AVX2-FAST-NEXT: vmovdqa %xmm5, %xmm8
+; AVX2-FAST-NEXT: vmovdqa %xmm14, %xmm13
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm14, %xmm14
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm1 = [0,0,0,1,8,15,128,128,0,0,0,1,8,15,128,128]
-; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm7, %xmm14
-; AVX2-FAST-NEXT: vpor %xmm15, %xmm14, %xmm14
+; AVX2-FAST-NEXT: vmovdqa %xmm15, %xmm4
+; AVX2-FAST-NEXT: vmovdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm15, %xmm15
+; AVX2-FAST-NEXT: vpor %xmm14, %xmm15, %xmm14
; AVX2-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-FAST-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm9, %ymm14, %ymm9
+; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm9, %ymm14, %ymm9
; AVX2-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm2, %ymm3, %ymm4, %ymm9
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendvb %ymm8, %ymm0, %ymm6, %ymm9
; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm9, %xmm10
; AVX2-FAST-NEXT: vextracti128 $1, %ymm9, %xmm9
; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm9, %xmm9
; AVX2-FAST-NEXT: vpor %xmm10, %xmm9, %xmm9
-; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm3, %xmm10
-; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm6, %xmm1
+; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm8, %xmm10
+; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-FAST-NEXT: vpshufb %xmm1, %xmm0, %xmm1
; AVX2-FAST-NEXT: vpor %xmm1, %xmm10, %xmm1
; AVX2-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm9, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm9, %ymm1, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = <u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255>
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendvb %ymm9, %ymm13, %ymm2, %ymm1
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <u,u,3,10,128,128,128,6,13,128,128,u,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm1, %xmm11
+; AVX2-FAST-NEXT: vpblendvb %ymm9, %ymm3, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = <u,u,3,10,128,128,128,6,13,128,128,u,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm1, %xmm11
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,128,128,1,8,15,128,128,4,11,u,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm1, %xmm1
; AVX2-FAST-NEXT: vpor %xmm1, %xmm11, %xmm1
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm11 = [0,0,0,2,9,128,128,128,0,0,0,2,9,128,128,128]
-; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm7, %xmm14
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm4, %xmm14
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm15 = [0,0,0,128,128,0,7,14,0,0,0,128,128,0,7,14]
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm8, %xmm10
+; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm13, %xmm10
; AVX2-FAST-NEXT: vpor %xmm14, %xmm10, %xmm10
; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm10, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm1, %ymm10, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendvb %ymm9, %ymm14, %ymm4, %ymm1
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm1, %xmm9
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendvb %ymm9, %ymm4, %ymm6, %ymm1
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm1, %xmm9
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm1, %xmm1
; AVX2-FAST-NEXT: vpor %xmm1, %xmm9, %xmm1
-; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm6, %xmm9
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm3, %xmm10
-; AVX2-FAST-NEXT: vmovdqa %xmm3, %xmm5
+; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm0, %xmm9
+; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm8, %xmm10
; AVX2-FAST-NEXT: vpor %xmm9, %xmm10, %xmm9
; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm9, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm1, %ymm9, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm9 = <255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0>
-; AVX2-FAST-NEXT: vpblendvb %ymm9, %ymm13, %ymm2, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm9, %ymm3, %ymm7, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u>
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm2, %ymm13, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm7, %ymm3, %ymm1
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,128,128,2,9,128,128,128,5,12,u,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm2, %xmm2
@@ -8323,37 +8326,39 @@
; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm1, %xmm1
; AVX2-FAST-NEXT: vpor %xmm2, %xmm1, %xmm1
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm2 = [0,0,0,3,10,128,128,128,0,0,0,3,10,128,128,128]
+; AVX2-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm7, %xmm7
; AVX2-FAST-NEXT: vpbroadcastq {{.*#+}} xmm12 = [0,0,0,128,128,1,8,15,0,0,0,128,128,1,8,15]
-; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm8, %xmm11
+; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm13, %xmm11
; AVX2-FAST-NEXT: vpor %xmm7, %xmm11, %xmm7
; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm7, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm1, %ymm7, %ymm1
; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm9, %ymm14, %ymm4, %ymm7
-; AVX2-FAST-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm4, %ymm14, %ymm1
+; AVX2-FAST-NEXT: vmovdqa %ymm6, %ymm1
+; AVX2-FAST-NEXT: vpblendvb %ymm9, %ymm4, %ymm6, %ymm6
+; AVX2-FAST-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm1, %ymm4, %ymm1
; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm4
; AVX2-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm3
; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm1, %xmm1
; AVX2-FAST-NEXT: vpor %xmm3, %xmm1, %xmm1
-; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm6, %xmm2
-; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm5, %xmm3
+; AVX2-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm2
+; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm8, %xmm3
; AVX2-FAST-NEXT: vpor %xmm2, %xmm3, %xmm2
; AVX2-FAST-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm0
+; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm1, %ymm2, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm1 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
-; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm0, %ymm1, %ymm3
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255>
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm0, %ymm1, %ymm12
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
+; AVX2-FAST-NEXT: vpblendvb %ymm6, %ymm0, %ymm1, %ymm3
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = <u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255>
+; AVX2-FAST-NEXT: vpblendvb %ymm13, %ymm0, %ymm1, %ymm12
; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm1, %ymm0, %ymm8
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm10 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm1, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm1, %ymm0, %ymm2
; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
; AVX2-FAST-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0
@@ -8361,73 +8366,73 @@
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm2
-; AVX2-FAST-NEXT: vmovdqu %ymm2, (%rsp) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm0, %ymm1, %ymm9
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm0, %ymm1, %ymm4
+; AVX2-FAST-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendvb %ymm6, %ymm0, %ymm1, %ymm9
+; AVX2-FAST-NEXT: vpblendvb %ymm13, %ymm0, %ymm1, %ymm5
; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm1, %ymm0, %ymm11
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm1, %ymm0, %ymm0
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm1, %ymm0, %ymm0
; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
-; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm2, %ymm0, %ymm6
-; AVX2-FAST-NEXT: vpblendvb %ymm5, %ymm2, %ymm0, %ymm1
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm2, %ymm0, %ymm5
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm0, %ymm2, %ymm13
-; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm0, %ymm2, %ymm0
-; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm10, %ymm13, %ymm0
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
-; AVX2-FAST-NEXT: vpblendvb %ymm2, %ymm10, %ymm13, %ymm2
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm10, %ymm13, %ymm7
-; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm13, %ymm10, %ymm14
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
+; AVX2-FAST-NEXT: vpblendvb %ymm4, %ymm1, %ymm0, %ymm10
+; AVX2-FAST-NEXT: vpblendvb %ymm6, %ymm1, %ymm0, %ymm2
+; AVX2-FAST-NEXT: vpblendvb %ymm13, %ymm1, %ymm0, %ymm6
+; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm0, %ymm1, %ymm14
; AVX2-FAST-NEXT: vmovdqu %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm14 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
-; AVX2-FAST-NEXT: vpblendvb %ymm14, %ymm13, %ymm10, %ymm10
-; AVX2-FAST-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vpblendvb %ymm7, %ymm0, %ymm1, %ymm0
+; AVX2-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-NEXT: vpblendvb %ymm4, %ymm1, %ymm14, %ymm0
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
+; AVX2-FAST-NEXT: vpblendvb %ymm4, %ymm1, %ymm14, %ymm4
+; AVX2-FAST-NEXT: vpblendvb %ymm13, %ymm1, %ymm14, %ymm7
+; AVX2-FAST-NEXT: vpblendvb %ymm15, %ymm14, %ymm1, %ymm13
+; AVX2-FAST-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
+; AVX2-FAST-NEXT: vpblendvb %ymm13, %ymm14, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = <2,9,128,128,128,5,12,128,128,u,u,u,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm3, %xmm10
+; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm3, %xmm1
; AVX2-FAST-NEXT: vextracti128 $1, %ymm3, %xmm3
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm14 = <128,128,0,7,14,128,128,3,10,u,u,u,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm3, %xmm3
-; AVX2-FAST-NEXT: vpor %xmm3, %xmm10, %xmm3
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm10
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0,1,2],ymm10[3],ymm6[4,5],ymm10[6],ymm6[7,8,9,10],ymm10[11],ymm6[12,13],ymm10[14],ymm6[15]
+; AVX2-FAST-NEXT: vpor %xmm1, %xmm3, %xmm1
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm10, %xmm3
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm10[0,1,2],ymm3[3],ymm10[4,5],ymm3[6],ymm10[7,8,9,10],ymm3[11],ymm10[12,13],ymm3[14],ymm10[15]
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm13 = [2,9,0,0,0,0,0,0,0,1,8,15,6,13,4,11,2,9,0,0,0,0,0,0,0,1,8,15,6,13,4,11]
; AVX2-FAST-NEXT: # ymm13 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm6, %ymm6
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm3, %ymm3
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm3, %ymm6, %ymm6
-; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm9, %xmm3
+; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm1, %ymm3, %ymm3
+; AVX2-FAST-NEXT: vpshufb %xmm15, %xmm9, %xmm1
; AVX2-FAST-NEXT: vextracti128 $1, %ymm9, %xmm9
; AVX2-FAST-NEXT: vpshufb %xmm14, %xmm9, %xmm9
-; AVX2-FAST-NEXT: vpor %xmm3, %xmm9, %xmm3
+; AVX2-FAST-NEXT: vpor %xmm1, %xmm9, %xmm1
; AVX2-FAST-NEXT: vextracti128 $1, %ymm0, %xmm9
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3],ymm0[4,5],ymm9[6],ymm0[7,8,9,10],ymm9[11],ymm0[12,13],ymm9[14],ymm0[15]
; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm0, %ymm0
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm3, %ymm0, %ymm3
+; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm1, %ymm0, %ymm15
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = <3,10,128,128,128,6,13,128,128,u,u,u,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm12, %xmm9
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm12, %xmm12
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm13 = <128,128,1,8,15,128,128,4,11,u,u,u,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm12, %xmm12
-; AVX2-FAST-NEXT: vpor %xmm9, %xmm12, %xmm9
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm1, %xmm12
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm12[0],ymm1[1,2],ymm12[3],ymm1[4,5,6],ymm12[7,8],ymm1[9,10],ymm12[11],ymm1[12,13,14],ymm12[15]
-; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm12 = [3,10,0,0,0,0,0,0,0,2,9,0,7,14,5,12,3,10,0,0,0,0,0,0,0,2,9,0,7,14,5,12]
-; AVX2-FAST-NEXT: # ymm12 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm1, %ymm1
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm9, %ymm1, %ymm9
-; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm4, %xmm0
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm1
-; AVX2-FAST-NEXT: vpshufb %xmm13, %xmm1, %xmm1
+; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm12, %xmm1
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm12, %xmm9
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm12 = <128,128,1,8,15,128,128,4,11,u,u,u,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm9, %xmm9
+; AVX2-FAST-NEXT: vpor %xmm1, %xmm9, %xmm1
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm9
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm9[0],ymm2[1,2],ymm9[3],ymm2[4,5,6],ymm9[7,8],ymm2[9,10],ymm9[11],ymm2[12,13,14],ymm9[15]
+; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm13 = [3,10,0,0,0,0,0,0,0,2,9,0,7,14,5,12,3,10,0,0,0,0,0,0,0,2,9,0,7,14,5,12]
+; AVX2-FAST-NEXT: # ymm13 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm2, %ymm2
+; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm1, %ymm2, %ymm9
+; AVX2-FAST-NEXT: vpshufb %xmm0, %xmm5, %xmm0
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm1
+; AVX2-FAST-NEXT: vpshufb %xmm12, %xmm1, %xmm1
; AVX2-FAST-NEXT: vpor %xmm0, %xmm1, %xmm0
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3],ymm2[4,5,6],ymm1[7,8],ymm2[9,10],ymm1[11],ymm2[12,13,14],ymm1[15]
-; AVX2-FAST-NEXT: vpshufb %ymm12, %ymm1, %ymm1
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm4, %xmm1
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm4[1,2],ymm1[3],ymm4[4,5,6],ymm1[7,8],ymm4[9,10],ymm1[11],ymm4[12,13,14],ymm1[15]
+; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm1, %ymm1
; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm0, %ymm1, %ymm12
; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm0
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = <128,128,2,9,128,128,128,5,12,u,u,u,u,u,u,u>
@@ -8435,8 +8440,8 @@
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = <4,11,128,128,0,7,14,128,128,u,u,u,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm8, %xmm1
; AVX2-FAST-NEXT: vpor %xmm0, %xmm1, %xmm0
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm5, %xmm1
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm5[1,2,3],ymm1[4],ymm5[5,6],ymm1[7,8],ymm5[9,10,11],ymm1[12],ymm5[13,14],ymm1[15]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm1
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0],ymm6[1,2,3],ymm1[4],ymm6[5,6],ymm1[7,8],ymm6[9,10,11],ymm1[12],ymm6[13,14],ymm1[15]
; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [4,11,0,0,0,0,0,0,0,3,10,1,8,15,6,13,4,11,0,0,0,0,0,0,0,3,10,1,8,15,6,13]
; AVX2-FAST-NEXT: # ymm5 = mem[0,1,0,1]
; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm1, %ymm1
@@ -8449,80 +8454,80 @@
; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0],ymm7[1,2,3],ymm2[4],ymm7[5,6],ymm2[7,8],ymm7[9,10,11],ymm2[12],ymm7[13,14],ymm2[15]
; AVX2-FAST-NEXT: vpshufb %ymm5, %ymm2, %ymm2
; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm0, %ymm2, %ymm0
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm7, %xmm2
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm2
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = <128,128,3,10,128,128,128,6,13,u,u,u,u,u,u,u>
; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm2, %xmm2
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <5,12,128,128,1,8,15,128,128,u,u,u,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm7, %xmm7
-; AVX2-FAST-NEXT: vpor %xmm2, %xmm7, %xmm2
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm7
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2,3],ymm7[4],ymm8[5,6,7,8],ymm7[9],ymm8[10,11],ymm7[12],ymm8[13,14,15]
-; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm8 = [5,12,0,0,0,0,0,0,0,4,11,2,9,0,7,14,5,12,0,0,0,0,0,0,0,4,11,2,9,0,7,14]
-; AVX2-FAST-NEXT: # ymm8 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm7, %ymm7
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm2, %ymm7, %ymm2
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm11, %xmm7
-; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm7, %xmm4
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm11, %xmm5
-; AVX2-FAST-NEXT: vpor %xmm4, %xmm5, %xmm4
+; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm6, %xmm6
+; AVX2-FAST-NEXT: vpor %xmm2, %xmm6, %xmm2
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm7, %xmm5
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm7[0],ymm5[1],ymm7[2,3],ymm5[4],ymm7[5,6,7,8],ymm5[9],ymm7[10,11],ymm5[12],ymm7[13,14,15]
-; AVX2-FAST-NEXT: vpshufb %ymm8, %ymm5, %ymm5
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm7, %xmm6
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm7[0],ymm6[1],ymm7[2,3],ymm6[4],ymm7[5,6,7,8],ymm6[9],ymm7[10,11],ymm6[12],ymm7[13,14,15]
+; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm7 = [5,12,0,0,0,0,0,0,0,4,11,2,9,0,7,14,5,12,0,0,0,0,0,0,0,4,11,2,9,0,7,14]
+; AVX2-FAST-NEXT: # ymm7 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm6, %ymm6
+; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm2, %ymm6, %ymm2
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm6
+; AVX2-FAST-NEXT: vpshufb %xmm4, %xmm6, %xmm4
+; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm8, %xmm5
+; AVX2-FAST-NEXT: vpor %xmm4, %xmm5, %xmm4
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm6, %xmm5
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm5 = ymm6[0],ymm5[1],ymm6[2,3],ymm5[4],ymm6[5,6,7,8],ymm5[9],ymm6[10,11],ymm5[12],ymm6[13,14,15]
+; AVX2-FAST-NEXT: vpshufb %ymm7, %ymm5, %ymm5
; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm4, %ymm5, %ymm4
; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = <6,13,128,128,2,9,128,128,128,u,u,u,u,u,u,u>
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm8, %xmm7
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm8
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm11 = <128,128,4,11,128,128,0,7,14,u,u,u,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm8, %xmm8
-; AVX2-FAST-NEXT: vpor %xmm7, %xmm8, %xmm7
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm13, %xmm8
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm13[0],ymm8[1],ymm13[2,3,4],ymm8[5],ymm13[6,7,8],ymm8[9],ymm13[10,11,12],ymm8[13],ymm13[14,15]
-; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm13 = [6,13,0,0,0,0,0,0,0,5,12,3,10,1,8,15,6,13,0,0,0,0,0,0,0,5,12,3,10,1,8,15]
-; AVX2-FAST-NEXT: # ymm13 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm8, %ymm8
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm7, %ymm8, %ymm7
-; AVX2-FAST-NEXT: vmovdqu (%rsp), %ymm8 # 32-byte Reload
-; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm8, %xmm5
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm8
-; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm8, %xmm8
-; AVX2-FAST-NEXT: vpor %xmm5, %xmm8, %xmm5
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm7, %xmm6
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm7, %xmm7
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = <128,128,4,11,128,128,0,7,14,u,u,u,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm7, %xmm7
+; AVX2-FAST-NEXT: vpor %xmm6, %xmm7, %xmm6
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm11, %xmm8
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm11[0],ymm8[1],ymm11[2,3,4],ymm8[5],ymm11[6,7,8],ymm8[9],ymm11[10,11,12],ymm8[13],ymm11[14,15]
-; AVX2-FAST-NEXT: vpshufb %ymm13, %ymm8, %ymm8
-; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm5, %ymm8, %ymm5
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm11, %xmm7
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm11[0],ymm7[1],ymm11[2,3,4],ymm7[5],ymm11[6,7,8],ymm7[9],ymm11[10,11,12],ymm7[13],ymm11[14,15]
+; AVX2-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm11 = [6,13,0,0,0,0,0,0,0,5,12,3,10,1,8,15,6,13,0,0,0,0,0,0,0,5,12,3,10,1,8,15]
+; AVX2-FAST-NEXT: # ymm11 = mem[0,1,0,1]
+; AVX2-FAST-NEXT: vpshufb %ymm11, %ymm7, %ymm7
+; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm6, %ymm7, %ymm6
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-NEXT: vpshufb %xmm5, %xmm7, %xmm5
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm7, %xmm7
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm7, %xmm7
+; AVX2-FAST-NEXT: vpor %xmm5, %xmm7, %xmm5
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm8, %xmm7
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2,3,4],ymm7[5],ymm8[6,7,8],ymm7[9],ymm8[10,11,12],ymm7[13],ymm8[14,15]
+; AVX2-FAST-NEXT: vpshufb %ymm11, %ymm7, %ymm7
+; AVX2-FAST-NEXT: vpblendvb %ymm10, %ymm5, %ymm7, %ymm5
+; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm11, %xmm7
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,128,128,128,5,12,128,128,1,8,15,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm7, %xmm7
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm10 = <u,u,0,7,14,128,128,3,10,128,128,128,u,u,u,u>
+; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm11, %xmm11
+; AVX2-FAST-NEXT: vpor %xmm7, %xmm11, %xmm7
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm11 = [1,3,1,2,1,3,5,6]
+; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm13 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm14 = [2,5,8,15,2,5,8,15,2,5,8,15,2,5,8,15,2,5,8,15,2,5,8,15,2,5,8,15,2,5,8,15]
+; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm13, %ymm13
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm3[0],ymm7[1,2,3,4,5,6,7],ymm3[8],ymm7[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm13, %xmm8
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm10 = <u,u,128,128,128,5,12,128,128,1,8,15,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm8, %xmm8
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} xmm11 = <u,u,0,7,14,128,128,3,10,128,128,128,u,u,u,u>
-; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm13, %xmm13
-; AVX2-FAST-NEXT: vpor %xmm8, %xmm13, %xmm8
-; AVX2-FAST-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX2-FAST-NEXT: vmovdqa {{.*#+}} ymm13 = [1,3,1,2,1,3,5,6]
-; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm14 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpbroadcastd {{.*#+}} ymm15 = [2,5,8,15,2,5,8,15,2,5,8,15,2,5,8,15,2,5,8,15,2,5,8,15,2,5,8,15,2,5,8,15]
-; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm14, %ymm14
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm14[7]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm6[0],ymm8[1,2,3,4,5,6,7],ymm6[8],ymm8[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-NEXT: vextracti128 $1, %ymm14, %xmm8
-; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm8, %xmm8
-; AVX2-FAST-NEXT: vpshufb %xmm11, %xmm14, %xmm10
-; AVX2-FAST-NEXT: vpor %xmm8, %xmm10, %xmm8
-; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: vpshufb %ymm15, %ymm10, %ymm10
-; AVX2-FAST-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm10[7]
-; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm3[0],ymm8[1,2,3,4,5,6,7],ymm3[8],ymm8[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-FAST-NEXT: vextracti128 $1, %ymm13, %xmm7
+; AVX2-FAST-NEXT: vpshufb %xmm8, %xmm7, %xmm7
+; AVX2-FAST-NEXT: vpshufb %xmm10, %xmm13, %xmm8
+; AVX2-FAST-NEXT: vpor %xmm7, %xmm8, %xmm7
+; AVX2-FAST-NEXT: vpermd {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm8 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: vpshufb %ymm14, %ymm8, %ymm8
+; AVX2-FAST-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5,6],ymm8[7]
+; AVX2-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm15[0],ymm7[1,2,3,4,5,6,7],ymm15[8],ymm7[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm15[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-NEXT: vpblendw $254, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm8 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm8 = ymm9[0],mem[1,2,3,4,5,6,7],ymm9[8],mem[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
@@ -8541,52 +8546,51 @@
; AVX2-FAST-NEXT: vpblendw $254, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm10 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm10 = ymm4[0],mem[1,2,3,4,5,6,7],ymm4[8],mem[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-FAST-NEXT: vpblendw $254, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm10 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm10 = ymm7[0],mem[1,2,3,4,5,6,7],ymm7[8],mem[9,10,11,12,13,14,15]
-; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-FAST-NEXT: vpblendw $254, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm10 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm10 = ymm6[0],mem[1,2,3,4,5,6,7],ymm6[8],mem[9,10,11,12,13,14,15]
+; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-NEXT: vpblendw $254, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm10 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm10 = ymm5[0],mem[1,2,3,4,5,6,7],ymm5[8],mem[9,10,11,12,13,14,15]
; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm10, 32(%rsi)
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm10, (%rsi)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-NEXT: vmovaps %ymm10, 32(%rdx)
+; AVX2-FAST-NEXT: vmovaps %ymm10, 32(%rsi)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FAST-NEXT: vmovaps %ymm10, (%rdx)
-; AVX2-FAST-NEXT: vmovdqa %ymm6, 32(%rcx)
-; AVX2-FAST-NEXT: vmovdqa %ymm3, (%rcx)
-; AVX2-FAST-NEXT: vmovdqa %ymm8, 32(%r8)
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovaps %ymm10, 32(%rdx)
+; AVX2-FAST-NEXT: vmovdqa %ymm7, (%rcx)
+; AVX2-FAST-NEXT: vmovdqa %ymm3, 32(%rcx)
; AVX2-FAST-NEXT: vmovdqa %ymm9, (%r8)
-; AVX2-FAST-NEXT: vmovdqa %ymm1, 32(%r9)
+; AVX2-FAST-NEXT: vmovdqa %ymm8, 32(%r8)
; AVX2-FAST-NEXT: vmovdqa %ymm0, (%r9)
+; AVX2-FAST-NEXT: vmovdqa %ymm1, 32(%r9)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovdqa %ymm2, 32(%rax)
; AVX2-FAST-NEXT: vmovdqa %ymm4, (%rax)
+; AVX2-FAST-NEXT: vmovdqa %ymm2, 32(%rax)
; AVX2-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-NEXT: vmovdqa %ymm7, 32(%rax)
; AVX2-FAST-NEXT: vmovdqa %ymm5, (%rax)
-; AVX2-FAST-NEXT: addq $712, %rsp # imm = 0x2C8
+; AVX2-FAST-NEXT: vmovdqa %ymm6, 32(%rax)
+; AVX2-FAST-NEXT: addq $728, %rsp # imm = 0x2D8
; AVX2-FAST-NEXT: vzeroupper
; AVX2-FAST-NEXT: retq
;
; AVX2-FAST-PERLANE-LABEL: load_i8_stride7_vf64:
; AVX2-FAST-PERLANE: # %bb.0:
-; AVX2-FAST-PERLANE-NEXT: subq $744, %rsp # imm = 0x2E8
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm10
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm3 = <255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm1, %ymm2, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm13
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, %ymm8
+; AVX2-FAST-PERLANE-NEXT: subq $760, %rsp # imm = 0x2F8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 64(%rdi), %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 320(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 224(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 256(%rdi), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = <255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0>
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm2, %ymm1, %ymm10, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, %ymm12
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = <128,128,128,5,12,128,128,1,8,15,u,u,u,u,u,u>
@@ -8595,8 +8599,8 @@
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpor %xmm3, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm9 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm9, %ymm5, %ymm4, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm9, %ymm4, %ymm5, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, %ymm15
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm9
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm11
@@ -8608,33 +8612,37 @@
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm4, %ymm5
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0]
; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm4, %ymm0, %ymm5, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm12, %ymm6, %ymm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm8, %ymm6, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm8, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm5, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm6, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 288(%rdi), %ymm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 96(%rdi), %ymm6
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm5, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpor %xmm1, %xmm2, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm10, %ymm6, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm15, %ymm6, %ymm7, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, %ymm8
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm5
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm5[2],ymm2[3,4],ymm5[5],ymm2[6,7,8,9],ymm5[10],ymm2[11,12],ymm5[13],ymm2[14,15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm4, %ymm1, %ymm2, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm11, %ymm9, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm4, %ymm1, %ymm2, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm9, %ymm11, %ymm1
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1],ymm2[2],ymm1[3,4,5],ymm2[6],ymm1[7,8,9],ymm2[10],ymm1[11,12,13],ymm2[14],ymm1[15]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm1 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm8, %ymm7, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm12, %ymm10, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, %ymm6
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm1 = <128,128,128,6,13,128,128,2,9,u,u,u,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm5, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm7 = <1,8,15,128,128,4,11,128,128,u,u,u,u,u,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm10 = <1,8,15,128,128,4,11,128,128,u,u,u,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm10, %xmm4, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpor %xmm5, %xmm4, %xmm4
; AVX2-FAST-PERLANE-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,8,15,0,0,0,0,0,0,0,7,14,5,12,3,10,1,8,15,0,0,0,0,0,0,0,7,14,5,12,3,10]
; AVX2-FAST-PERLANE-NEXT: # ymm5 = mem[0,1,0,1]
@@ -8642,98 +8650,100 @@
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm9 = <0,0,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,u,u,u,u,u,u,u,u,u,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm9, %ymm2, %ymm4, %ymm2
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm6, %ymm10, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm7, %ymm8, %ymm2
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4,5],ymm4[6],ymm2[7,8,9],ymm4[10],ymm2[11,12,13],ymm4[14],ymm2[15]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm5, %ymm2, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm12, %ymm15, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm13, %ymm0, %ymm4
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm5, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm10, %xmm4, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpor %xmm1, %xmm4, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm9, %ymm2, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm9, %ymm2, %ymm1, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm13
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm0 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm1, %ymm13, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, %ymm7
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm10 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm3, %ymm4, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm2, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm4, %xmm5
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm9 = <u,u,u,128,128,3,10,128,128,128,6,13,u,u,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm7, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm11 = <u,u,u,5,12,128,128,1,8,15,128,128,u,u,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm2, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpor %xmm7, %xmm2, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastw {{.*#+}} xmm13 = [2,9,2,9,2,9,2,9,2,9,2,9,2,9,2,9]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 208(%rdi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm1, %xmm7
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm5, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,u,5,12,128,128,1,8,15,128,128,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpor %xmm5, %xmm4, %xmm4
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastw {{.*#+}} xmm2 = [2,9,2,9,2,9,2,9,2,9,2,9,2,9,2,9]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 432(%rdi), %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm1, %xmm5
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastw {{.*#+}} xmm1 = [4,11,4,11,4,11,4,11,4,11,4,11,4,11,4,11]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %xmm8
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm8, %xmm15
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm8, %xmm12
-; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm15[0],xmm7[0],xmm15[1],xmm7[1],xmm15[2],xmm7[2],xmm15[3],xmm7[3]
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm7[7]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm15 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm15, %ymm0, %ymm2, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastw {{.*#+}} xmm0 = [4,11,4,11,4,11,4,11,4,11,4,11,4,11,4,11]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm0, %xmm3, %xmm14
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm0, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm14 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0,0,0,0,0,0]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 384(%rdi), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 352(%rdi), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm2, %ymm5, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm8
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm8, %xmm8
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpor %xmm0, %xmm8, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 432(%rdi), %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm5, %xmm8
-; AVX2-FAST-PERLANE-NEXT: vmovdqa 416(%rdi), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm2, %xmm9
-; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 160(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 128(%rdi), %ymm5
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm4, %ymm5, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm15
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm15, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpor %xmm0, %xmm9, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 208(%rdi), %xmm10
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm10, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm10, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqa 192(%rdi), %xmm10
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm10, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm12[0],xmm9[0],xmm12[1],xmm9[1],xmm12[2],xmm9[2],xmm12[3],xmm9[3]
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm9[7]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm6 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm7, %ymm13, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,u,6,13,128,128,2,9,128,128,128,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm0, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,u,128,128,4,11,128,128,0,7,14,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpor %xmm0, %xmm9, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastw {{.*#+}} xmm9 = [3,10,3,10,3,10,3,10,3,10,3,10,3,10,3,10]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm1, %xmm13
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastw {{.*#+}} xmm15 = [5,12,5,12,5,12,5,12,5,12,5,12,5,12,5,12]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm3, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm13[0],xmm8[1],xmm13[1],xmm8[2],xmm13[2],xmm8[3],xmm13[3]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm15, %ymm14, %ymm0, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm1 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm3, %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,u,6,13,128,128,2,9,128,128,128,u,u,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm0, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm6, %ymm4, %ymm5, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm0, %xmm3
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm9 = <u,u,u,128,128,4,11,128,128,0,7,14,u,u,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpor %xmm0, %xmm8, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastw {{.*#+}} xmm8 = [3,10,3,10,3,10,3,10,3,10,3,10,3,10,3,10]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm6, %xmm11
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastw {{.*#+}} xmm13 = [5,12,5,12,5,12,5,12,5,12,5,12,5,12,5,12]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm12, %xmm14
-; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm14[0],xmm11[0],xmm14[1],xmm11[1],xmm14[2],xmm11[2],xmm14[3],xmm11[3]
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm11[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm1, %ymm7, %ymm10, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm0, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpor %xmm3, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm5, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm2, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm2, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, %xmm13
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm15, %xmm10, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm15, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm14, %ymm11, %ymm0, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm4 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm4, %ymm10, %ymm7, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm10, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm2, %ymm11, %ymm7, %ymm0
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,128,128,128,6,13,128,128,2,9,u,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm1
@@ -8741,38 +8751,34 @@
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm1 = [0,0,0,128,128,128,5,12,0,0,0,128,128,128,5,12]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm5, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm15, %xmm8
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm9 = [0,0,0,0,7,14,128,128,0,0,0,0,7,14,128,128]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm2, %xmm11
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, %xmm13
-; AVX2-FAST-PERLANE-NEXT: vpor %xmm8, %xmm11, %xmm8
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm14, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vpor %xmm8, %xmm12, %xmm8
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm12
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm8, %ymm0, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vbroadcasti128 {{.*#+}} ymm10 = [0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; AVX2-FAST-PERLANE-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm11, %ymm8, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[0,1,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm12, %ymm8, %ymm8
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm4, %ymm2, %ymm15, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm8, %xmm11
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm11, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm2, %ymm5, %ymm4, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm8, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm12, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm8, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpor %xmm3, %xmm6, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm0, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm12, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm13, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm1, %xmm13, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm9, %xmm10, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpor %xmm1, %xmm6, %xmm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm3, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm3, %ymm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm14, %ymm7, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm14, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm11, %ymm7, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm7 = <u,u,2,9,128,128,128,5,12,128,128,u,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm1, %xmm6
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm1
@@ -8780,70 +8786,70 @@
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpor %xmm6, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm6 = [0,0,0,128,128,128,6,13,0,0,0,128,128,128,6,13]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm5, %xmm9
-; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm11 = [0,0,0,1,8,15,128,128,0,0,0,1,8,15,128,128]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm13, %xmm5
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm13, %xmm13
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm15, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm12 = [0,0,0,1,8,15,128,128,0,0,0,1,8,15,128,128]
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm14, %xmm13
; AVX2-FAST-PERLANE-NEXT: vpor %xmm9, %xmm13, %xmm9
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm1, %ymm9, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm1, %ymm9, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm2, %ymm15, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm15, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm5, %ymm4, %ymm1
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm7, %xmm1, %xmm3
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpor %xmm3, %xmm1, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm0, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm11, %xmm12, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm2, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm10, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm10, %xmm2
; AVX2-FAST-PERLANE-NEXT: vpor %xmm3, %xmm6, %xmm3
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm1, %ymm3, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm1, %ymm3, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255>
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm14, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm4, %ymm14, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm12 = <u,u,3,10,128,128,128,6,13,128,128,u,u,u,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm1, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm11, %ymm7, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,3,10,128,128,128,6,13,128,128,u,u,u,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm6
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm8 = <u,u,128,128,1,8,15,128,128,4,11,u,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpor %xmm6, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm6 = [0,0,0,2,9,128,128,128,0,0,0,2,9,128,128,128]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm5, %xmm9
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm5, %xmm11
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm14, %xmm9
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm14, %xmm12
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm13 = [0,0,0,128,128,0,7,14,0,0,0,128,128,0,7,14]
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm4, %xmm14
-; AVX2-FAST-PERLANE-NEXT: vpor %xmm11, %xmm14, %xmm11
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm10, %xmm14
+; AVX2-FAST-PERLANE-NEXT: vpor %xmm12, %xmm14, %xmm12
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm1, %ymm11, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm1, %ymm12, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm15, %ymm2, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm12, %xmm1, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm15, %ymm5, %ymm4, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm1, %xmm3
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpor %xmm3, %xmm1, %xmm1
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm2, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm14, %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm5, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm13, %xmm14, %xmm6
; AVX2-FAST-PERLANE-NEXT: vpor %xmm3, %xmm6, %xmm3
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm1, %ymm3, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm1, %ymm3, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm12 = <255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm12, %ymm0, %ymm7, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm13 = <255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm7, %ymm0, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm12, %ymm11, %ymm7, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm11 = <255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm11, %ymm7, %ymm3, %ymm1
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm3 = <u,u,128,128,2,9,128,128,128,5,12,u,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm2, %xmm2
@@ -8853,50 +8859,49 @@
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm2 = [0,0,0,3,10,128,128,128,0,0,0,3,10,128,128,128]
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm9, %xmm7
; AVX2-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} xmm8 = [0,0,0,128,128,1,8,15,0,0,0,128,128,1,8,15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm4, %xmm11
-; AVX2-FAST-PERLANE-NEXT: vpor %xmm7, %xmm11, %xmm7
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm10, %xmm13
+; AVX2-FAST-PERLANE-NEXT: vpor %xmm7, %xmm13, %xmm7
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm1, %ymm7, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm1, %ymm7, %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm12, %ymm15, %ymm1, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm13, %ymm12
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm1, %ymm15, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm12, %ymm15, %ymm4, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, %ymm12
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm11, %ymm4, %ymm15, %ymm1
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm1, %xmm4
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm3, %xmm4, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vpor %xmm3, %xmm1, %xmm1
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm14, %xmm2
-; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm5, %xmm3
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm2, %xmm5, %xmm2
+; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm8, %xmm14, %xmm3
; AVX2-FAST-PERLANE-NEXT: vpor %xmm2, %xmm3, %xmm2
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm10, %ymm1, %ymm2, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm2, %ymm1, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm13 = <u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm2, %ymm1, %ymm15
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm12, %ymm1, %ymm2, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm4 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm4, %ymm1, %ymm2, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm1, %ymm2, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm3, %ymm2, %ymm1, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm1, %ymm2, %ymm3
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm1, %ymm2, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm12, %ymm2, %ymm1, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm4, %ymm2, %ymm1, %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm5 = <u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255>
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm1, %ymm10, %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm13 = <u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255>
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm1, %ymm10, %ymm15
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm11, %ymm10, %ymm1, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm4 = <255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u>
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm4, %ymm10, %ymm1, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255,u,u,0,0,255,255,u,u,0,0,u,u,255,255>
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm2, %ymm10, %ymm1, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm2, %ymm1, %ymm10, %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm10, %ymm1, %ymm3
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm10, %ymm1, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm12, %ymm1, %ymm10, %ymm14
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm4, %ymm1, %ymm10, %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm6 = <255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u,0,0,u,u,255,255,0,0,u,u,255,255,u,u>
@@ -8995,7 +9000,7 @@
; AVX2-FAST-PERLANE-NEXT: # ymm8 = mem[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT: vpshufb %ymm8, %ymm7, %ymm7
; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm0, %ymm4, %ymm7, %ymm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsp), %ymm9 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vextracti128 $1, %ymm9, %xmm7
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm5, %xmm7, %xmm5
; AVX2-FAST-PERLANE-NEXT: vpshufb %xmm6, %xmm9, %xmm6
@@ -9090,26 +9095,26 @@
; AVX2-FAST-PERLANE-NEXT: # ymm10 = ymm0[0],mem[1,2,3,4,5,6,7],ymm0[8],mem[9,10,11,12,13,14,15]
; AVX2-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, 32(%rsi)
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, (%rsi)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, 32(%rdx)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, 32(%rsi)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, (%rdx)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, 32(%rcx)
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm10, 32(%rdx)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm8, (%rcx)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, 32(%r8)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, 32(%rcx)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, (%r8)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, 32(%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm2, 32(%r8)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm1, (%r9)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm3, 32(%r9)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, 32(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm5, (%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm4, 32(%rax)
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, 32(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm0, (%rax)
-; AVX2-FAST-PERLANE-NEXT: addq $744, %rsp # imm = 0x2E8
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, 32(%rax)
+; AVX2-FAST-PERLANE-NEXT: addq $760, %rsp # imm = 0x2F8
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
index 9c6b03d..1afeca0 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
@@ -5588,78 +5588,78 @@
define void @load_i8_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
; SSE-LABEL: load_i8_stride8_vf64:
; SSE: # %bb.0:
-; SSE-NEXT: subq $2024, %rsp # imm = 0x7E8
+; SSE-NEXT: subq $2040, %rsp # imm = 0x7F8
; SSE-NEXT: movdqa 64(%rdi), %xmm6
; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 80(%rdi), %xmm8
+; SSE-NEXT: movdqa 80(%rdi), %xmm7
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 96(%rdi), %xmm8
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 96(%rdi), %xmm11
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 128(%rdi), %xmm5
; SSE-NEXT: movdqa 144(%rdi), %xmm10
-; SSE-NEXT: movdqa 160(%rdi), %xmm7
+; SSE-NEXT: movdqa 160(%rdi), %xmm4
; SSE-NEXT: movdqa 176(%rdi), %xmm13
; SSE-NEXT: movdqa 192(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 208(%rdi), %xmm15
; SSE-NEXT: movdqa 224(%rdi), %xmm9
; SSE-NEXT: movdqa 240(%rdi), %xmm12
-; SSE-NEXT: movdqa {{.*#+}} xmm4 = [255,0,0,0,255,0,0,0]
+; SSE-NEXT: movdqa {{.*#+}} xmm14 = [255,0,0,0,255,0,0,0]
; SSE-NEXT: movdqa %xmm12, %xmm0
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa %xmm9, %xmm1
-; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm1
; SSE-NEXT: packuswb %xmm0, %xmm1
; SSE-NEXT: packuswb %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: pand %xmm4, %xmm1
-; SSE-NEXT: pand %xmm4, %xmm2
+; SSE-NEXT: pand %xmm14, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm2
; SSE-NEXT: packuswb %xmm1, %xmm2
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: packuswb %xmm0, %xmm2
; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: pand %xmm4, %xmm0
-; SSE-NEXT: movdqa %xmm7, %xmm1
-; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm0
+; SSE-NEXT: movdqa %xmm4, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm1
; SSE-NEXT: packuswb %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm10, %xmm0
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa %xmm5, %xmm3
-; SSE-NEXT: pand %xmm4, %xmm3
+; SSE-NEXT: pand %xmm14, %xmm3
; SSE-NEXT: packuswb %xmm0, %xmm3
-; SSE-NEXT: movdqa 112(%rdi), %xmm14
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 112(%rdi), %xmm11
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: packuswb %xmm1, %xmm0
; SSE-NEXT: packuswb %xmm3, %xmm3
; SSE-NEXT: packuswb %xmm0, %xmm3
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,3],xmm2[0,3]
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm14, %xmm0
-; SSE-NEXT: pand %xmm4, %xmm0
-; SSE-NEXT: movdqa %xmm11, %xmm1
-; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm11, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
+; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm1
; SSE-NEXT: packuswb %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm8, %xmm0
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa %xmm6, %xmm2
-; SSE-NEXT: pand %xmm4, %xmm2
+; SSE-NEXT: pand %xmm14, %xmm2
; SSE-NEXT: packuswb %xmm0, %xmm2
; SSE-NEXT: packuswb %xmm1, %xmm0
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: packuswb %xmm0, %xmm2
-; SSE-NEXT: movdqa 48(%rdi), %xmm14
-; SSE-NEXT: movdqa %xmm14, %xmm0
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: movdqa 48(%rdi), %xmm11
+; SSE-NEXT: movdqa %xmm11, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa 32(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm1
; SSE-NEXT: packuswb %xmm0, %xmm1
-; SSE-NEXT: movdqa 16(%rdi), %xmm8
-; SSE-NEXT: movdqa %xmm8, %xmm0
-; SSE-NEXT: pand %xmm4, %xmm0
-; SSE-NEXT: movdqa (%rdi), %xmm11
-; SSE-NEXT: movdqa %xmm11, %xmm3
-; SSE-NEXT: pand %xmm4, %xmm3
+; SSE-NEXT: movdqa 16(%rdi), %xmm7
+; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
+; SSE-NEXT: movdqa (%rdi), %xmm8
+; SSE-NEXT: movdqa %xmm8, %xmm3
+; SSE-NEXT: pand %xmm14, %xmm3
; SSE-NEXT: packuswb %xmm0, %xmm3
; SSE-NEXT: packuswb %xmm1, %xmm0
; SSE-NEXT: packuswb %xmm3, %xmm3
@@ -5668,34 +5668,34 @@
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 496(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa 480(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm1
; SSE-NEXT: packuswb %xmm0, %xmm1
; SSE-NEXT: movdqa 464(%rdi), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa 448(%rdi), %xmm2
-; SSE-NEXT: movdqa %xmm2, (%rsp) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm2
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pand %xmm14, %xmm2
; SSE-NEXT: packuswb %xmm0, %xmm2
; SSE-NEXT: packuswb %xmm1, %xmm0
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: packuswb %xmm0, %xmm2
; SSE-NEXT: movdqa 432(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa 416(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm1
; SSE-NEXT: packuswb %xmm0, %xmm1
; SSE-NEXT: movdqa 400(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa 384(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm3
+; SSE-NEXT: pand %xmm14, %xmm3
; SSE-NEXT: packuswb %xmm0, %xmm3
; SSE-NEXT: packuswb %xmm1, %xmm0
; SSE-NEXT: packuswb %xmm3, %xmm3
@@ -5704,40 +5704,39 @@
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 368(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa 352(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm1
; SSE-NEXT: packuswb %xmm0, %xmm1
; SSE-NEXT: movdqa 336(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa 320(%rdi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm2
+; SSE-NEXT: pand %xmm14, %xmm2
; SSE-NEXT: packuswb %xmm0, %xmm2
; SSE-NEXT: packuswb %xmm1, %xmm0
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: packuswb %xmm0, %xmm2
; SSE-NEXT: movdqa 304(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa 288(%rdi), %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: pand %xmm14, %xmm1
; SSE-NEXT: packuswb %xmm0, %xmm1
; SSE-NEXT: movdqa 272(%rdi), %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm4, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: movdqa 256(%rdi), %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pand %xmm3, %xmm4
-; SSE-NEXT: packuswb %xmm0, %xmm4
+; SSE-NEXT: pand %xmm3, %xmm14
+; SSE-NEXT: packuswb %xmm0, %xmm14
; SSE-NEXT: packuswb %xmm1, %xmm0
-; SSE-NEXT: packuswb %xmm4, %xmm4
-; SSE-NEXT: packuswb %xmm0, %xmm4
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,3],xmm2[0,3]
-; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: packuswb %xmm14, %xmm14
+; SSE-NEXT: packuswb %xmm0, %xmm14
+; SSE-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,3],xmm2[0,3]
; SSE-NEXT: pxor %xmm6, %xmm6
; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -5780,15 +5779,15 @@
; SSE-NEXT: movdqa {{.*#+}} xmm13 = [65535,65535,65535,0,65535,65535,65535,65535]
; SSE-NEXT: movdqa %xmm13, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm7, %xmm1
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3],xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3]
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,1,1,3]
+; SSE-NEXT: punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3],xmm4[4],xmm6[4],xmm4[5],xmm6[5],xmm4[6],xmm6[6],xmm4[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,1,1,3]
; SSE-NEXT: packuswb %xmm1, %xmm1
; SSE-NEXT: pand %xmm13, %xmm1
; SSE-NEXT: por %xmm2, %xmm1
@@ -5848,39 +5847,39 @@
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm8, %xmm0
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm7, %xmm0
; SSE-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3],xmm8[4],xmm6[4],xmm8[5],xmm6[5],xmm8[6],xmm6[6],xmm8[7],xmm6[7]
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
-; SSE-NEXT: packuswb %xmm8, %xmm8
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3],xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
+; SSE-NEXT: packuswb %xmm7, %xmm7
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm10, %xmm1
-; SSE-NEXT: pandn %xmm8, %xmm1
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm2
+; SSE-NEXT: pandn %xmm7, %xmm1
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm8, %xmm2
; SSE-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm6[0],xmm11[1],xmm6[1],xmm11[2],xmm6[2],xmm11[3],xmm6[3],xmm11[4],xmm6[4],xmm11[5],xmm6[5],xmm11[6],xmm6[6],xmm11[7],xmm6[7]
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3]
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; SSE-NEXT: punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3],xmm8[4],xmm6[4],xmm8[5],xmm6[5],xmm8[6],xmm6[6],xmm8[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1],xmm8[2],xmm2[2],xmm8[3],xmm2[3]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
; SSE-NEXT: packuswb %xmm0, %xmm0
; SSE-NEXT: pand %xmm10, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1],xmm14[2],xmm6[2],xmm14[3],xmm6[3],xmm14[4],xmm6[4],xmm14[5],xmm6[5],xmm14[6],xmm6[6],xmm14[7],xmm6[7]
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1],xmm14[2],xmm1[2],xmm14[3],xmm1[3]
-; SSE-NEXT: packuswb %xmm14, %xmm14
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[0,0,2,3]
+; SSE-NEXT: punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm6[0],xmm11[1],xmm6[1],xmm11[2],xmm6[2],xmm11[3],xmm6[3],xmm11[4],xmm6[4],xmm11[5],xmm6[5],xmm11[6],xmm6[6],xmm11[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1],xmm11[2],xmm1[2],xmm11[3],xmm1[3]
+; SSE-NEXT: packuswb %xmm11, %xmm11
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,0,2,3]
; SSE-NEXT: movdqa %xmm13, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
@@ -5948,17 +5947,17 @@
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm14, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm4, %xmm0
; SSE-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1],xmm14[2],xmm6[2],xmm14[3],xmm6[3],xmm14[4],xmm6[4],xmm14[5],xmm6[5],xmm14[6],xmm6[6],xmm14[7],xmm6[7]
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
-; SSE-NEXT: packuswb %xmm14, %xmm14
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3],xmm4[4],xmm6[4],xmm4[5],xmm6[5],xmm4[6],xmm6[6],xmm4[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; SSE-NEXT: packuswb %xmm4, %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm10, %xmm1
-; SSE-NEXT: pandn %xmm14, %xmm1
+; SSE-NEXT: pandn %xmm4, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
@@ -5971,15 +5970,15 @@
; SSE-NEXT: packuswb %xmm0, %xmm0
; SSE-NEXT: pand %xmm10, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm8, %xmm1
; SSE-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1],xmm14[2],xmm6[2],xmm14[3],xmm6[3],xmm14[4],xmm6[4],xmm14[5],xmm6[5],xmm14[6],xmm6[6],xmm14[7],xmm6[7]
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1],xmm14[2],xmm1[2],xmm14[3],xmm1[3]
-; SSE-NEXT: packuswb %xmm14, %xmm14
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm14[0,0,2,3]
+; SSE-NEXT: punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3],xmm8[4],xmm6[4],xmm8[5],xmm6[5],xmm8[6],xmm6[6],xmm8[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1],xmm8[2],xmm1[2],xmm8[3],xmm1[3]
+; SSE-NEXT: packuswb %xmm8, %xmm8
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[0,0,2,3]
; SSE-NEXT: movdqa %xmm13, %xmm2
; SSE-NEXT: pandn %xmm1, %xmm2
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
@@ -6020,7 +6019,7 @@
; SSE-NEXT: packuswb %xmm1, %xmm1
; SSE-NEXT: pand %xmm12, %xmm1
; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: movdqa (%rsp), %xmm2 # 16-byte Reload
; SSE-NEXT: movdqa %xmm2, %xmm3
; SSE-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm6[8],xmm3[9],xmm6[9],xmm3[10],xmm6[10],xmm3[11],xmm6[11],xmm3[12],xmm6[12],xmm3[13],xmm6[13],xmm3[14],xmm6[14],xmm3[15],xmm6[15]
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -6031,13 +6030,13 @@
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm9, %xmm2
; SSE-NEXT: pandn %xmm3, %xmm2
-; SSE-NEXT: movdqa (%rsp), %xmm3 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm3, %xmm4
-; SSE-NEXT: punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm3, %xmm5
+; SSE-NEXT: punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3],xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[1,1,1,1]
; SSE-NEXT: packuswb %xmm3, %xmm3
@@ -6047,16 +6046,16 @@
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm8, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm4, %xmm0
; SSE-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3],xmm8[4],xmm6[4],xmm8[5],xmm6[5],xmm8[6],xmm6[6],xmm8[7],xmm6[7]
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
-; SSE-NEXT: packuswb %xmm8, %xmm8
+; SSE-NEXT: punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3],xmm4[4],xmm6[4],xmm4[5],xmm6[5],xmm4[6],xmm6[6],xmm4[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; SSE-NEXT: packuswb %xmm4, %xmm4
; SSE-NEXT: movdqa %xmm10, %xmm1
-; SSE-NEXT: pandn %xmm8, %xmm1
+; SSE-NEXT: pandn %xmm4, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
@@ -6065,10 +6064,10 @@
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[1,1,1,1]
-; SSE-NEXT: packuswb %xmm4, %xmm4
-; SSE-NEXT: pand %xmm10, %xmm4
-; SSE-NEXT: por %xmm1, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; SSE-NEXT: packuswb %xmm0, %xmm0
+; SSE-NEXT: pand %xmm10, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: movdqa %xmm3, %xmm1
; SSE-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
@@ -6093,7 +6092,7 @@
; SSE-NEXT: pand %xmm13, %xmm1
; SSE-NEXT: por %xmm5, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movdqa %xmm1, %xmm2
; SSE-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
@@ -6101,9 +6100,9 @@
; SSE-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT: packuswb %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,1,2,2]
+; SSE-NEXT: packuswb %xmm1, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,2]
; SSE-NEXT: movdqa %xmm12, %xmm5
; SSE-NEXT: pandn %xmm1, %xmm5
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
@@ -6118,105 +6117,35 @@
; SSE-NEXT: packuswb %xmm1, %xmm1
; SSE-NEXT: pand %xmm12, %xmm1
; SSE-NEXT: por %xmm5, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm5, %xmm2
; SSE-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
-; SSE-NEXT: movdqa %xmm2, %xmm15
+; SSE-NEXT: movdqa %xmm2, %xmm7
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm0, %xmm2
-; SSE-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3],xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm11, %xmm0
-; SSE-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm11, %xmm2
+; SSE-NEXT: punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm6[0],xmm11[1],xmm6[1],xmm11[2],xmm6[2],xmm11[3],xmm6[3],xmm11[4],xmm6[4],xmm11[5],xmm6[5],xmm11[6],xmm6[6],xmm11[7],xmm6[7]
; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm2, %xmm5
-; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm15[0],xmm5[1],xmm15[1],xmm5[2],xmm15[2],xmm5[3],xmm15[3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]
; SSE-NEXT: packuswb %xmm5, %xmm6
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm9, %xmm5
; SSE-NEXT: pandn %xmm6, %xmm5
-; SSE-NEXT: punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3]
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm11[1,1,1,1]
; SSE-NEXT: packuswb %xmm15, %xmm15
; SSE-NEXT: pand %xmm9, %xmm15
; SSE-NEXT: por %xmm5, %xmm15
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm15[2,2,2,2]
; SSE-NEXT: punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm1[2],xmm5[3],xmm1[3]
-; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; SSE-NEXT: movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
; SSE-NEXT: movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm7
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[0,2,2,3]
-; SSE-NEXT: movdqa %xmm7, %xmm4
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT: packuswb %xmm0, %xmm0
-; SSE-NEXT: movdqa %xmm10, %xmm5
-; SSE-NEXT: pandn %xmm0, %xmm5
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT: packuswb %xmm0, %xmm0
-; SSE-NEXT: pand %xmm10, %xmm0
-; SSE-NEXT: por %xmm5, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,7,5]
-; SSE-NEXT: packuswb %xmm5, %xmm5
-; SSE-NEXT: movdqa %xmm13, %xmm15
-; SSE-NEXT: pandn %xmm5, %xmm15
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,5,6,7]
-; SSE-NEXT: packuswb %xmm5, %xmm5
-; SSE-NEXT: pand %xmm13, %xmm5
-; SSE-NEXT: por %xmm15, %xmm5
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,7,5]
-; SSE-NEXT: packuswb %xmm5, %xmm5
-; SSE-NEXT: movdqa %xmm12, %xmm15
-; SSE-NEXT: pandn %xmm5, %xmm15
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,1,2,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,5,6,7]
-; SSE-NEXT: packuswb %xmm5, %xmm5
-; SSE-NEXT: pand %xmm12, %xmm5
-; SSE-NEXT: por %xmm15, %xmm5
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm1[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm15[0,1,1,3,4,5,6,7]
-; SSE-NEXT: packuswb %xmm15, %xmm15
-; SSE-NEXT: movdqa %xmm9, %xmm7
-; SSE-NEXT: pandn %xmm15, %xmm7
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pand %xmm2, %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm1[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm15[1,3,2,3,4,5,6,7]
-; SSE-NEXT: packuswb %xmm15, %xmm15
-; SSE-NEXT: pand %xmm9, %xmm15
-; SSE-NEXT: por %xmm7, %xmm15
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm15[2,2,2,2]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm5[2],xmm7[3],xmm5[3]
-; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
-; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -6239,8 +6168,8 @@
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,7,5]
; SSE-NEXT: packuswb %xmm5, %xmm5
-; SSE-NEXT: movdqa %xmm13, %xmm7
-; SSE-NEXT: pandn %xmm5, %xmm7
+; SSE-NEXT: movdqa %xmm13, %xmm15
+; SSE-NEXT: pandn %xmm5, %xmm15
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -6248,7 +6177,7 @@
; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,5,6,7]
; SSE-NEXT: packuswb %xmm5, %xmm5
; SSE-NEXT: pand %xmm13, %xmm5
-; SSE-NEXT: por %xmm7, %xmm5
+; SSE-NEXT: por %xmm15, %xmm5
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -6257,8 +6186,8 @@
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm0[0,1,2,0]
; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,7,5]
; SSE-NEXT: packuswb %xmm5, %xmm5
-; SSE-NEXT: movdqa %xmm12, %xmm7
-; SSE-NEXT: pandn %xmm5, %xmm7
+; SSE-NEXT: movdqa %xmm12, %xmm15
+; SSE-NEXT: pandn %xmm5, %xmm15
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -6266,24 +6195,24 @@
; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,5,6,7]
; SSE-NEXT: packuswb %xmm5, %xmm5
; SSE-NEXT: pand %xmm12, %xmm5
-; SSE-NEXT: por %xmm7, %xmm5
+; SSE-NEXT: por %xmm15, %xmm5
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
-; SSE-NEXT: packuswb %xmm7, %xmm7
-; SSE-NEXT: movdqa %xmm9, %xmm15
-; SSE-NEXT: pandn %xmm7, %xmm15
+; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm15[0,1,1,3,4,5,6,7]
+; SSE-NEXT: packuswb %xmm15, %xmm15
+; SSE-NEXT: movdqa %xmm9, %xmm7
+; SSE-NEXT: pandn %xmm15, %xmm7
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm0[0,2,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
-; SSE-NEXT: packuswb %xmm7, %xmm7
-; SSE-NEXT: pand %xmm9, %xmm7
-; SSE-NEXT: por %xmm15, %xmm7
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[2,2,2,2]
+; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm15 = xmm15[1,3,2,3,4,5,6,7]
+; SSE-NEXT: packuswb %xmm15, %xmm15
+; SSE-NEXT: pand %xmm9, %xmm15
+; SSE-NEXT: por %xmm7, %xmm15
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm15[2,2,2,2]
; SSE-NEXT: punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm5[2],xmm7[3],xmm5[3]
; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -6345,10 +6274,79 @@
; SSE-NEXT: packuswb %xmm7, %xmm7
; SSE-NEXT: movdqa %xmm9, %xmm15
; SSE-NEXT: pandn %xmm7, %xmm15
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
+; SSE-NEXT: packuswb %xmm7, %xmm7
+; SSE-NEXT: pand %xmm9, %xmm7
+; SSE-NEXT: por %xmm15, %xmm7
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[2,2,2,2]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm5[2],xmm7[3],xmm5[3]
+; SSE-NEXT: movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
+; SSE-NEXT: movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
+; SSE-NEXT: packuswb %xmm0, %xmm0
+; SSE-NEXT: movdqa %xmm10, %xmm5
+; SSE-NEXT: pandn %xmm0, %xmm5
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
+; SSE-NEXT: packuswb %xmm0, %xmm0
+; SSE-NEXT: pand %xmm10, %xmm0
+; SSE-NEXT: por %xmm5, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,7,5]
+; SSE-NEXT: packuswb %xmm5, %xmm5
+; SSE-NEXT: movdqa %xmm13, %xmm7
+; SSE-NEXT: pandn %xmm5, %xmm7
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT: packuswb %xmm5, %xmm5
+; SSE-NEXT: pand %xmm13, %xmm5
+; SSE-NEXT: por %xmm7, %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm5[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,7,5]
+; SSE-NEXT: packuswb %xmm5, %xmm5
+; SSE-NEXT: movdqa %xmm12, %xmm7
+; SSE-NEXT: pandn %xmm5, %xmm7
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[0,1,2,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT: packuswb %xmm5, %xmm5
+; SSE-NEXT: pand %xmm12, %xmm5
+; SSE-NEXT: por %xmm7, %xmm5
; SSE-NEXT: movdqa (%rsp), %xmm1 # 16-byte Reload
; SSE-NEXT: pand %xmm2, %xmm1
; SSE-NEXT: movdqa %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm1[0,2,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
+; SSE-NEXT: packuswb %xmm7, %xmm7
+; SSE-NEXT: movdqa %xmm9, %xmm15
+; SSE-NEXT: pandn %xmm7, %xmm15
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pand %xmm2, %xmm6
+; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm6[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
; SSE-NEXT: packuswb %xmm7, %xmm7
; SSE-NEXT: pand %xmm9, %xmm7
@@ -6513,7 +6511,7 @@
; SSE-NEXT: pand %xmm10, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm13, %xmm1
-; SSE-NEXT: pandn %xmm14, %xmm1
+; SSE-NEXT: pandn %xmm8, %xmm1
; SSE-NEXT: pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; SSE-NEXT: # xmm5 = mem[2,2,3,3]
; SSE-NEXT: packuswb %xmm5, %xmm5
@@ -6541,7 +6539,7 @@
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm8[1,1,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[1,1,2,3]
; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
@@ -6565,10 +6563,12 @@
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: pand %xmm12, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,1,3,3]
+; SSE-NEXT: pshufd $244, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,3,3]
; SSE-NEXT: movdqa %xmm9, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
+; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,3,3,3]
; SSE-NEXT: packuswb %xmm1, %xmm1
; SSE-NEXT: pand %xmm9, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
@@ -6576,7 +6576,8 @@
; SSE-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; SSE-NEXT: movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
; SSE-NEXT: movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm4[3,1,2,3]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT: # xmm7 = mem[3,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[0,1,2,0,4,5,6,7]
; SSE-NEXT: packuswb %xmm0, %xmm0
; SSE-NEXT: movdqa %xmm10, %xmm1
@@ -6636,9 +6637,10 @@
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT: # xmm14 = mem[3,1,2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm14[0,1,2,0,4,5,6,7]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[3,1,2,3]
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,0,4,5,6,7]
; SSE-NEXT: packuswb %xmm0, %xmm0
; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
@@ -6740,16 +6742,15 @@
; SSE-NEXT: packuswb %xmm1, %xmm1
; SSE-NEXT: pand %xmm12, %xmm1
; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: pshufd $231, (%rsp), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[3,1,2,3]
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm2, (%rsp) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,0,4,5,6,7]
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: movdqa %xmm9, %xmm3
; SSE-NEXT: pandn %xmm2, %xmm3
-; SSE-NEXT: pshufd $231, (%rsp), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3]
-; SSE-NEXT: movdqa %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm6[3,1,2,3]
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: pand %xmm9, %xmm2
@@ -6809,10 +6810,9 @@
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: movdqa %xmm9, %xmm3
; SSE-NEXT: pandn %xmm2, %xmm3
-; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3]
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[2,0,2,3,4,5,6,7]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = mem[3,1,2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm8[2,0,2,3,4,5,6,7]
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: pand %xmm9, %xmm2
; SSE-NEXT: por %xmm3, %xmm2
@@ -6820,6 +6820,74 @@
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: packuswb %xmm0, %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm10, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
+; SSE-NEXT: packuswb %xmm0, %xmm0
+; SSE-NEXT: pand %xmm10, %xmm0
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: packuswb %xmm1, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,2,3]
+; SSE-NEXT: movdqa %xmm13, %xmm2
+; SSE-NEXT: pandn %xmm1, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
+; SSE-NEXT: packuswb %xmm1, %xmm1
+; SSE-NEXT: pand %xmm13, %xmm1
+; SSE-NEXT: por %xmm2, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: packuswb %xmm1, %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,2]
+; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: pandn %xmm1, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
+; SSE-NEXT: packuswb %xmm1, %xmm1
+; SSE-NEXT: pand %xmm12, %xmm1
+; SSE-NEXT: por %xmm2, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: packuswb %xmm2, %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm9, %xmm2
+; SSE-NEXT: pandn %xmm4, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[1,1,1,1]
+; SSE-NEXT: packuswb %xmm3, %xmm3
+; SSE-NEXT: pand %xmm9, %xmm3
+; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[2,2,2,2]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT: movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
@@ -6944,9 +7012,9 @@
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; SSE-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: packuswb %xmm2, %xmm8
+; SSE-NEXT: packuswb %xmm2, %xmm4
; SSE-NEXT: movdqa %xmm9, %xmm2
-; SSE-NEXT: pandn %xmm8, %xmm2
+; SSE-NEXT: pandn %xmm4, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[1,1,1,1]
; SSE-NEXT: packuswb %xmm3, %xmm3
; SSE-NEXT: pand %xmm9, %xmm3
@@ -7007,88 +7075,23 @@
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT: # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT: # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: packuswb %xmm2, %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm9, %xmm2
; SSE-NEXT: pandn %xmm3, %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,1,1]
-; SSE-NEXT: packuswb %xmm3, %xmm3
-; SSE-NEXT: pand %xmm9, %xmm3
-; SSE-NEXT: por %xmm2, %xmm3
-; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm3[2,2,2,2]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm1[2],xmm6[3],xmm1[3]
-; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm0[0],xmm6[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: packuswb %xmm2, %xmm2
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm10, %xmm1
-; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT: packuswb %xmm0, %xmm0
-; SSE-NEXT: pand %xmm10, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: packuswb %xmm1, %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,2,3]
-; SSE-NEXT: movdqa %xmm13, %xmm2
-; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
-; SSE-NEXT: packuswb %xmm1, %xmm1
-; SSE-NEXT: pand %xmm13, %xmm1
-; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: packuswb %xmm1, %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,2]
-; SSE-NEXT: movdqa %xmm12, %xmm2
-; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT: # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
-; SSE-NEXT: packuswb %xmm1, %xmm1
-; SSE-NEXT: pand %xmm12, %xmm1
-; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: packuswb %xmm2, %xmm4
-; SSE-NEXT: movdqa %xmm9, %xmm2
-; SSE-NEXT: pandn %xmm4, %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[1,1,1,1]
; SSE-NEXT: packuswb %xmm3, %xmm3
; SSE-NEXT: pand %xmm9, %xmm3
; SSE-NEXT: por %xmm2, %xmm3
; SSE-NEXT: pshufd {{.*#+}} xmm15 = xmm3[2,2,2,2]
; SSE-NEXT: punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm1[2],xmm15[3],xmm1[3]
; SSE-NEXT: movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm14[0,1,3,1,4,5,6,7]
+; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT: # xmm0 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: packuswb %xmm0, %xmm0
; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
@@ -7128,9 +7131,9 @@
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: pand %xmm9, %xmm2
; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm2[2,2,2,2]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm1[2],xmm14[3],xmm1[3]
-; SSE-NEXT: movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm1[2],xmm6[3],xmm1[3]
+; SSE-NEXT: movsd {{.*#+}} xmm6 = xmm0[0],xmm6[1]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm7[0,1,3,1,4,5,6,7]
; SSE-NEXT: packuswb %xmm0, %xmm0
; SSE-NEXT: movdqa %xmm10, %xmm1
@@ -7212,8 +7215,7 @@
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: movdqa %xmm9, %xmm3
; SSE-NEXT: pandn %xmm2, %xmm3
-; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm8[3,1,2,3,4,5,6,7]
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: pand %xmm9, %xmm2
; SSE-NEXT: por %xmm3, %xmm2
@@ -7252,12 +7254,12 @@
; SSE-NEXT: packuswb %xmm1, %xmm1
; SSE-NEXT: pand %xmm12, %xmm1
; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: pshuflw $116, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: pshuflw $116, (%rsp), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[0,1,3,1,4,5,6,7]
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: movdqa %xmm9, %xmm3
; SSE-NEXT: pandn %xmm2, %xmm3
-; SSE-NEXT: pshuflw $231, (%rsp), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; SSE-NEXT: # xmm2 = mem[3,1,2,3,4,5,6,7]
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: pand %xmm9, %xmm2
@@ -7290,7 +7292,8 @@
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: pand %xmm12, %xmm2
; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[0,1,3,3]
+; SSE-NEXT: pshufd $244, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[0,1,3,3]
; SSE-NEXT: movdqa %xmm9, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
; SSE-NEXT: pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
@@ -7407,8 +7410,7 @@
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
; SSE-NEXT: punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
; SSE-NEXT: movsd {{.*#+}} xmm0 = xmm10[0],xmm0[1]
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, 32(%rsi)
+; SSE-NEXT: movaps %xmm14, 32(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: movaps %xmm4, 48(%rsi)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
@@ -7448,23 +7450,24 @@
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: movaps %xmm4, 16(%r9)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movapd %xmm15, 32(%rax)
-; SSE-NEXT: movapd %xmm6, 48(%rax)
+; SSE-NEXT: movapd %xmm15, 48(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT: movaps %xmm4, (%rax)
+; SSE-NEXT: movaps %xmm4, 32(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: movaps %xmm4, 16(%rax)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT: movaps %xmm4, (%rax)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT: movapd %xmm5, 48(%rax)
; SSE-NEXT: movapd %xmm7, 32(%rax)
; SSE-NEXT: movapd %xmm11, 16(%rax)
-; SSE-NEXT: movapd %xmm14, (%rax)
+; SSE-NEXT: movapd %xmm6, (%rax)
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT: movapd %xmm0, 48(%rax)
; SSE-NEXT: movapd %xmm1, 32(%rax)
; SSE-NEXT: movapd %xmm2, 16(%rax)
; SSE-NEXT: movapd %xmm3, (%rax)
-; SSE-NEXT: addq $2024, %rsp # imm = 0x7E8
+; SSE-NEXT: addq $2040, %rsp # imm = 0x7F8
; SSE-NEXT: retq
;
; AVX1-ONLY-LABEL: load_i8_stride8_vf64:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
index f9d8d0b..da36f16 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
@@ -11754,156 +11754,153 @@
;
; AVX512F-ONLY-FAST-LABEL: store_i16_stride7_vf64:
; AVX512F-ONLY-FAST: # %bb.0:
-; AVX512F-ONLY-FAST-NEXT: subq $2264, %rsp # imm = 0x8D8
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rsi), %ymm9
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rcx), %ymm2
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rdx), %ymm8
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm2, %ymm23
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u>
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm8, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vporq %ymm0, %ymm2, %ymm16
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128,128,128,128,128]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm0, %ymm9, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u,16,17,18,19>
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm1, %ymm18
-; AVX512F-ONLY-FAST-NEXT: vporq %ymm5, %ymm6, %ymm17
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128]
+; AVX512F-ONLY-FAST-NEXT: subq $2200, %rsp # imm = 0x898
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rsi), %ymm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rdi), %ymm6
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rcx), %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rdx), %ymm15
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm3, %ymm25
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm7, %ymm15, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vporq %ymm0, %ymm3, %ymm17
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128,128,128,128,128]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm1, %ymm20
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u,16,17,18,19>
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm0, %ymm6, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm6, %ymm23
+; AVX512F-ONLY-FAST-NEXT: vporq %ymm4, %ymm5, %ymm16
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = [128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128]
; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%r9), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm6, %ymm1, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm5, %ymm1, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm5, %ymm6
; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%r8), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u>
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vporq %ymm5, %ymm6, %ymm24
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm8, %ymm1, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vporq %ymm4, %ymm5, %ymm24
; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rcx), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm4
; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rdx), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm5, %ymm6, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm7, %ymm1, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpor %ymm4, %ymm5, %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rsi), %ymm1
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm0, %ymm1, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm1, %ymm30
; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rdi), %ymm14
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm14, %ymm10
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm5, %ymm10, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm0, %ymm14, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpor %ymm4, %ymm5, %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa (%r9), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm7, %ymm1, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm7, %ymm11
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm6, %ymm1, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm6, %ymm10
; AVX512F-ONLY-FAST-NEXT: vmovdqa (%r8), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm10
-; AVX512F-ONLY-FAST-NEXT: vporq %ymm5, %ymm10, %ymm19
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm8, %ymm1, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vporq %ymm4, %ymm5, %ymm31
; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rcx), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm4
; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rdx), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm10
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm5, %ymm10, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm7, %ymm1, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpor %ymm4, %ymm5, %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rsi), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm0, %ymm1, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm4
; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rdi), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm10
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm5, %ymm10, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm0, %ymm1, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpor %ymm4, %ymm5, %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rcx), %ymm13
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm4, %ymm13, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm13, %ymm2
; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rdx), %ymm12
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm12, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm4, %ymm3, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm7, %ymm12, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vpor %ymm2, %ymm4, %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rsi), %ymm10
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm0, %ymm10, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rsi), %ymm11
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm11, %ymm2
; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rdi), %ymm7
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm7, %ymm2
-; AVX512F-ONLY-FAST-NEXT: vpor %ymm0, %ymm2, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm0, %ymm7, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpor %ymm2, %ymm0, %ymm0
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%r8), %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm15, %ymm6, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%r9), %ymm4
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm11, %ymm4, %ymm15
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm11, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vporq %ymm15, %ymm0, %ymm22
-; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
-; AVX512F-ONLY-FAST-NEXT: # ymm1 = mem[0,1,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm1, %ymm31
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm15 = ymm6[3,3,3,3,7,7,7,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm15[2],ymm0[3,4],ymm15[5],ymm0[6,7,8,9],ymm15[10],ymm0[11,12],ymm15[13],ymm0[14,15]
-; AVX512F-ONLY-FAST-NEXT: vprold $16, %ymm4, %ymm15
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm6[1,2,2,3,5,6,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm11[0,1],ymm15[2],ymm11[3,4],ymm15[5],ymm11[6,7,8,9],ymm15[10],ymm11[11,12],ymm15[13],ymm11[14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm8, %ymm6, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%r9), %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm10, %ymm5, %ymm9
+; AVX512F-ONLY-FAST-NEXT: vporq %ymm9, %ymm0, %ymm22
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm9 = ymm6[3,3,3,3,7,7,7,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm9[2],ymm0[3,4],ymm9[5],ymm0[6,7,8,9],ymm9[10],ymm0[11,12],ymm9[13],ymm0[14,15]
+; AVX512F-ONLY-FAST-NEXT: vprold $16, %ymm5, %ymm9
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm6[1,2,2,3,5,6,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4],ymm9[5],ymm8[6,7,8,9],ymm9[10],ymm8[11,12],ymm9[13],ymm8[14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [2,1,3,2,10,10,10,11]
-; AVX512F-ONLY-FAST-NEXT: vpermi2q %zmm0, %zmm11, %zmm1
+; AVX512F-ONLY-FAST-NEXT: vpermi2q %zmm0, %zmm8, %zmm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u>
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm1, %ymm9, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm1, %ymm29
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm18, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm18[2,2,2,2,6,6,6,6]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0],ymm0[1],ymm11[2,3],ymm0[4],ymm11[5,6,7,8],ymm0[9],ymm11[10,11],ymm0[12],ymm11[13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u>
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm20, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm2, %ymm29
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm23[2,2,2,2,6,6,6,6]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm0[1],ymm8[2,3],ymm0[4],ymm8[5,6,7,8],ymm0[9],ymm8[10,11],ymm0[12],ymm8[13,14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29>
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm9, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm3, %ymm25
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm18[3,3,3,3,7,7,7,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0,1,2],ymm0[3],ymm11[4,5],ymm0[6],ymm11[7,8,9,10],ymm0[11],ymm11[12,13],ymm0[14],ymm11[15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29>
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm2, %ymm20
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm23[3,3,3,3,7,7,7,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0,1,2],ymm0[3],ymm8[4,5],ymm0[6],ymm8[7,8,9,10],ymm0[11],ymm8[12,13],ymm0[14],ymm8[15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27,22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27]
; AVX512F-ONLY-FAST-NEXT: # ymm2 = mem[0,1,0,1]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm23, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm2, %ymm20
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm8[2,2,2,2,6,6,6,6]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm11[2],ymm0[3,4],ymm11[5],ymm0[6,7,8,9],ymm11[10],ymm0[11,12],ymm11[13],ymm0[14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm25, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm2, %ymm21
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm15[2,2,2,2,6,6,6,6]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm8[2],ymm0[3,4],ymm8[5],ymm0[6,7,8,9],ymm8[10],ymm0[11,12],ymm8[13],ymm0[14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u>
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm2, %ymm21
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm8[3,3,3,3,7,7,7,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0],ymm0[1],ymm11[2,3],ymm0[4],ymm11[5,6,7,8],ymm0[9],ymm11[10,11],ymm0[12],ymm11[13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm2, %ymm25
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm15[3,3,3,3,7,7,7,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm0[1],ymm8[2,3],ymm0[4],ymm8[5,6,7,8],ymm0[9],ymm8[10,11],ymm0[12],ymm8[13,14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm3[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm8[0,1,1,3,4,5,5,7]
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm4[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm15[0,1,1,3,4,5,5,7]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0,1],ymm0[2],ymm8[3,4],ymm0[5],ymm8[6,7,8,9],ymm0[10],ymm8[11,12],ymm0[13],ymm8[14,15]
; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,1,3,2]
; AVX512F-ONLY-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
; AVX512F-ONLY-FAST-NEXT: # ymm2 = mem[0,1,0,1]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm9, %ymm8
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm8
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm2, %ymm18
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm1[1,1,1,1,5,5,5,5]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm8[0,1],ymm11[2],ymm8[3,4],ymm11[5],ymm8[6,7,8,9],ymm11[10],ymm8[11,12],ymm11[13],ymm8[14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm9 = ymm23[1,1,1,1,5,5,5,5]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4],ymm9[5],ymm8[6,7,8,9],ymm9[10],ymm8[11,12],ymm9[13],ymm8[14,15]
; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm8[2,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm16, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm17, %zmm8
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm17, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm16, %zmm8
; AVX512F-ONLY-FAST-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm8
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <5,u,u,u,6,u,u,6>
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%r8), %ymm11
-; AVX512F-ONLY-FAST-NEXT: vpermd %ymm11, %ymm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%r8), %ymm9
+; AVX512F-ONLY-FAST-NEXT: vpermd %ymm9, %ymm0, %ymm0
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <u,4,u,u,u,5,u,u>
-; AVX512F-ONLY-FAST-NEXT: vpermd %ymm11, %ymm0, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm11[14,15],zero,zero,ymm11[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm11[16,17],zero,zero,ymm11[u,u],zero,zero
+; AVX512F-ONLY-FAST-NEXT: vpermd %ymm9, %ymm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm9[14,15],zero,zero,ymm9[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm9[16,17],zero,zero,ymm9[u,u],zero,zero
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} ymm16 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm16, %ymm8, %ymm11
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %ymm16, %ymm8, %ymm9
; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%r9), %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm5, %ymm1, %ymm15
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm11, %ymm15
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm10, %ymm1, %ymm15
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm9, %ymm15
; AVX512F-ONLY-FAST-NEXT: vextracti64x4 $1, %zmm8, %ymm8
; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm0
; AVX512F-ONLY-FAST-NEXT: vprold $16, %ymm1, %ymm8
@@ -11929,13 +11926,13 @@
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535]
; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 72(%rax), %ymm0
; AVX512F-ONLY-FAST-NEXT: vpandn %ymm0, %ymm1, %ymm8
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rax), %ymm9
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm9, %ymm11
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm8, %zmm27
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rax), %ymm10
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm10, %ymm9
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm8, %zmm27
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm24, %zmm0, %zmm15
; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%r9), %xmm2
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%r8), %xmm11
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm11[4],xmm2[4],xmm11[5],xmm2[5],xmm11[6],xmm2[6],xmm11[7],xmm2[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%r8), %xmm9
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm9[4],xmm2[4],xmm9[5],xmm2[5],xmm9[6],xmm2[6],xmm9[7],xmm2[7]
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm2, %xmm28
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
@@ -11948,16 +11945,15 @@
; AVX512F-ONLY-FAST-NEXT: vpandn %ymm0, %ymm1, %ymm0
; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rax), %ymm1
; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm15
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm1, %ymm30
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm1, %ymm19
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm15, %zmm0, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm19, %zmm0, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa (%r9), %xmm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm31, %zmm0, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa (%r9), %xmm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqa (%r8), %xmm8
; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm15 = xmm8[4],xmm5[4],xmm8[5],xmm5[5],xmm8[6],xmm5[6],xmm8[7],xmm5[7]
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm15 = xmm8[4],xmm4[4],xmm8[5],xmm4[5],xmm8[6],xmm4[6],xmm8[7],xmm4[7]
; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm15, %xmm15
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm2, %xmm19
; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm0, %zmm24, %zmm15
; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, %zmm26, %zmm15, %zmm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
@@ -11965,13 +11961,13 @@
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm15 = ymm12[0,1,1,3,4,5,5,7]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm15[0,1],ymm0[2],ymm15[3,4],ymm0[5],ymm15[6,7,8,9],ymm0[10],ymm15[11,12],ymm0[13],ymm15[14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm18, %ymm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm5, %ymm10, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm18, %ymm4
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm4, %ymm11, %ymm0
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm15 = ymm7[1,1,1,1,5,5,5,5]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm15[2],ymm0[3,4],ymm15[5],ymm0[6,7,8,9],ymm15[10],ymm0[11,12],ymm15[13],ymm0[14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-ONLY-FAST-NEXT: vpbroadcastd {{.*#+}} ymm8 = [18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm8, %ymm4, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm8, %ymm5, %ymm0
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm6[0,0,2,1,4,4,6,5]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7,8,9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = <0,1,u,3,10,10,11,11>
@@ -11983,660 +11979,658 @@
; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm0, %ymm1
; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm29, %ymm4
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm4, %ymm10, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm29, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm5, %ymm11, %ymm1
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm7[2,2,2,2,6,6,6,6]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm25, %ymm15
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm15, %ymm10, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm20, %ymm6
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm6, %ymm11, %ymm1
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5],ymm1[6],ymm2[7,8,9,10],ymm1[11],ymm2[12,13],ymm1[14],ymm2[15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm20, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm21, %ymm3
; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm13, %ymm1
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm12[2,2,2,2,6,6,6,6]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm21, %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm6, %ymm13, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm25, %ymm15
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm15, %ymm13, %ymm1
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm12[3,3,3,3,7,7,7,7]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 (%rax), %zmm10
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 (%rax), %zmm21
; AVX512F-ONLY-FAST-NEXT: vbroadcasti64x4 {{.*#+}} zmm1 = [14,21,0,0,15,22,0,15,14,21,0,0,15,22,0,15]
; AVX512F-ONLY-FAST-NEXT: # zmm1 = mem[0,1,2,3,0,1,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpermi2d %zmm0, %zmm10, %zmm1
+; AVX512F-ONLY-FAST-NEXT: vpermi2d %zmm0, %zmm21, %zmm1
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm5, %ymm12, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm30, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm0
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm14[1,1,1,1,5,5,5,5]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7,8,9],ymm1[10],ymm0[11,12],ymm1[13],ymm0[14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm4, %ymm12, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm5, %ymm2, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm30, %ymm4
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm14[2,2,2,2,6,6,6,6]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6,7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13,14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm4[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm5[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm2[0,1,1,3,4,5,5,7]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7,8,9],ymm0[10],ymm1[11,12],ymm0[13],ymm1[14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm4, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm4, %ymm7
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm3, %ymm5, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm5, %ymm11
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm2[2,2,2,2,6,6,6,6]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm2, %ymm5
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm2, %ymm7
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7,8,9],ymm1[10],ymm0[11,12],ymm1[13],ymm0[14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX512F-ONLY-FAST-NEXT: vprold $16, %ymm2, %ymm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm4[1,2,2,3,5,6,6,7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm5[1,2,2,3,5,6,6,7]
; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7,8,9],ymm0[10],ymm1[11,12],ymm0[13],ymm1[14,15]
; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm8, %ymm2, %ymm1
; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm2, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[0,0,2,1,4,4,6,5]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7,8,9,10],ymm2[11],ymm1[12,13],ymm2[14],ymm1[15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm21 = [2,2,3,3,10,9,11,10]
-; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm0, %zmm21, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm23, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vbroadcasti64x4 {{.*#+}} zmm13 = [0,13,4,0,0,14,5,0,0,13,4,0,0,14,5,0]
-; AVX512F-ONLY-FAST-NEXT: # zmm13 = mem[0,1,2,3,0,1,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm0, %zmm13, %zmm2
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm1, %zmm0, %zmm2
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm6, %ymm7, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm5[3,3,3,3,7,7,7,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm15, %ymm12, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm5[0,0,2,1,4,4,6,5]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7,8,9,10],ymm2[11],ymm1[12,13],ymm2[14],ymm1[15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm12 = [2,2,3,3,10,9,11,10]
+; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm0, %zmm12, %zmm2
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm23, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vbroadcasti64x4 {{.*#+}} zmm1 = [0,13,4,0,0,14,5,0,0,13,4,0,0,14,5,0]
+; AVX512F-ONLY-FAST-NEXT: # zmm1 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm0, %zmm1, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm2, %zmm1, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm15, %ymm11, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6,7,8],ymm0[9],ymm2[10,11],ymm0[12],ymm2[13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm6, %ymm4, %ymm0
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm14[3,3,3,3,7,7,7,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5],ymm1[6],ymm2[7,8,9,10],ymm1[11],ymm2[12,13],ymm1[14],ymm2[15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm31, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[3,3,3,3,7,7,7,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [2,2,2,3,8,8,8,9]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3],ymm2[4,5],ymm0[6],ymm2[7,8,9,10],ymm0[11],ymm2[12,13],ymm0[14],ymm2[15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vbroadcasti32x4 {{.*#+}} ymm22 = [26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
+; AVX512F-ONLY-FAST-NEXT: # ymm22 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm22, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm5[3,3,3,3,7,7,7,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7,8,9],ymm2[10],ymm0[11,12],ymm2[13],ymm0[14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = [2,2,2,3,8,8,8,9]
; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%r9), %xmm5
; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%r8), %xmm6
; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,6,7,4,5,6,7,4,5,12,13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm7, %xmm1
-; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm1, %zmm3, %zmm4
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [6,7,3,3,7,7,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpermd %ymm9, %ymm1, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 96(%rax), %ymm3
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm1, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,2,3,6,7,4,5,6,7,4,5,12,13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm3, %xmm7, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm3, %xmm14
+; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm2, %zmm4, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [6,7,3,3,7,7,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpermd %ymm10, %ymm2, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 96(%rax), %ymm4
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm3, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm3
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rsi), %xmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rsi), %xmm0
; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rdi), %xmm4
-; AVX512F-ONLY-FAST-NEXT: vprold $16, %xmm3, %xmm9
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm12 = xmm4[1,1,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm12[0,1],xmm9[2],xmm12[3,4],xmm9[5],xmm12[6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm9, %xmm23
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rcx), %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rdx), %xmm4
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastq {{.*#+}} xmm12 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm12, %xmm3, %xmm9
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm12, %xmm8
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm12 = xmm4[1,1,2,2]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm12[0],xmm9[1],xmm12[2,3],xmm9[4],xmm12[5,6],xmm9[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm19, %xmm4
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm4, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm5, %xmm7, %xmm4
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm5, %xmm7
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = [0,1,1,3,8,8,9,9]
-; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm3, %zmm5, %zmm4
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 100(%rax), %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 104(%rax), %ymm5
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm3, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rcx), %xmm3
+; AVX512F-ONLY-FAST-NEXT: vprold $16, %xmm0, %xmm10
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm11 = xmm4[1,1,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm11[0,1],xmm10[2],xmm11[3,4],xmm10[5],xmm11[6,7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm3, %xmm30
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm0, %xmm23
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rcx), %xmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 96(%rdx), %xmm10
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastq {{.*#+}} xmm15 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm15, %xmm0, %xmm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm15, %xmm8
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm11 = xmm10[1,1,2,2]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm11[0],xmm4[1],xmm11[2,3],xmm4[4],xmm11[5,6],xmm4[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm10[4],xmm0[4],xmm10[5],xmm0[5],xmm10[6],xmm0[6],xmm10[7],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm15, %xmm7, %xmm5
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm15, %xmm7
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,1,1,3,8,8,9,9]
+; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm0, %zmm3, %zmm5
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 100(%rax), %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 104(%rax), %ymm6
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm5, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rcx), %xmm0
; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rdx), %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm4, %xmm31
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm8, %xmm9
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm8, %xmm3, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm3, %xmm31
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm8, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm8, %xmm0, %xmm0
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[1,1,2,2]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm5[0],xmm3[1],xmm5[2,3],xmm3[4],xmm5[5,6],xmm3[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rdi), %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0],xmm0[1],xmm5[2,3],xmm0[4],xmm5[5,6],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm18
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rdi), %xmm0
; AVX512F-ONLY-FAST-NEXT: vmovdqa 64(%rsi), %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm6, %xmm24
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm6, %xmm22
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm15 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm15, %xmm17
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm15 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
; AVX512F-ONLY-FAST-NEXT: vprold $16, %xmm5, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[1,1,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm5[2],xmm3[3,4],xmm5[5],xmm3[6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm28, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm11[0],xmm3[0],xmm11[1],xmm3[1],xmm11[2],xmm3[2],xmm11[3],xmm3[3]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm7, %xmm11
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm7, %xmm3, %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm5[2],xmm0[3,4],xmm5[5],xmm0[6,7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm28, %xmm0
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm7, %xmm13
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm7, %xmm0, %xmm5
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm14, %xmm0, %xmm0
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm6 = [0,0,0,1,8,9,9,11]
-; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm5, %zmm6, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm5, %zmm6, %zmm0
; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 64(%rax), %ymm5
; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 68(%rax), %ymm7
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm5, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm7 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm3, %zmm7, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rcx), %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rdx), %xmm8
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm8[4],xmm3[4],xmm8[5],xmm3[5],xmm8[6],xmm3[6],xmm8[7],xmm3[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm5, %xmm19
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm8[0],xmm3[0],xmm8[1],xmm3[1],xmm8[2],xmm3[2],xmm8[3],xmm3[3]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm9, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[1,1,2,2]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm8[0],xmm3[1],xmm8[2,3],xmm3[4],xmm8[5,6],xmm3[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm3, %ymm20
-; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rdi), %xmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rsi), %xmm8
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm9 = xmm8[4],xmm3[4],xmm8[5],xmm3[5],xmm8[6],xmm3[6],xmm8[7],xmm3[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm9, %xmm28
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm9, %xmm18
-; AVX512F-ONLY-FAST-NEXT: vprold $16, %xmm8, %xmm8
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[1,1,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm8[2],xmm3[3,4],xmm8[5],xmm3[6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1],xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm11, %xmm3, %xmm8
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm8, %zmm6, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm5, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm8 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm0, %zmm8, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rcx), %xmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rdx), %xmm5
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm3, %xmm16
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[1,1,2,2]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0],xmm0[1],xmm5[2,3],xmm0[4],xmm5[5,6],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rdi), %xmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa (%rsi), %xmm9
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm9[4],xmm0[4],xmm9[5],xmm0[5],xmm9[6],xmm0[6],xmm9[7],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm2, %xmm28
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3]
+; AVX512F-ONLY-FAST-NEXT: vprold $16, %xmm9, %xmm9
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm9[2],xmm0[3,4],xmm9[5],xmm0[6,7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm27
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm13, %xmm0, %xmm9
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm13, %xmm29
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm14, %xmm0, %xmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm14, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm9, %zmm6, %zmm0
; AVX512F-ONLY-FAST-NEXT: vpbroadcastd (%rax), %ymm6
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 4(%rax), %ymm8
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm3, %zmm7, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm12[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm7[1,1,1,1,5,5,5,5]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm3, %ymm29
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm12[10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm7[2,2,2,2,6,6,6,6]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm7, %ymm14
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm6[0],ymm3[1],ymm6[2,3],ymm3[4],ymm6[5,6,7,8],ymm3[9],ymm6[10,11],ymm3[12],ymm6[13,14,15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm3, %ymm26
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vprold $16, %ymm8, %ymm3
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 4(%rax), %ymm9
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm6, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm0, %zmm8, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm3[1,1,1,1,5,5,5,5]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm6[2],ymm0[3,4],ymm6[5],ymm0[6,7,8,9],ymm6[10],ymm0[11,12],ymm6[13],ymm0[14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm26
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm13[10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm3[2,2,2,2,6,6,6,6]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm3, %ymm14
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm0[1],ymm6[2,3],ymm0[4],ymm6[5,6,7,8],ymm0[9],ymm6[10,11],ymm0[12],ymm6[13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm25
+; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vprold $16, %ymm3, %ymm0
; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm9[1,2,2,3,5,6,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm6[0,1],ymm3[2],ymm6[3,4],ymm3[5],ymm6[6,7,8,9],ymm3[10],ymm6[11,12],ymm3[13],ymm6[14,15]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm7 = ymm9[0,0,2,1,4,4,6,5]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3],ymm6[4,5],ymm7[6],ymm6[7,8,9,10],ymm7[11],ymm6[12,13],ymm7[14],ymm6[15]
-; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm3, %zmm21, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm11[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm7 = ymm15[0,1,1,3,4,5,5,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm7[0,1],ymm3[2],ymm7[3,4],ymm3[5],ymm7[6,7,8,9],ymm3[10],ymm7[11,12],ymm3[13],ymm7[14,15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm3, %ymm27
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,26,27,u,u,24,25,26,27,u,u,26,27,26,27]
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm7 = ymm15[2,2,2,2,6,6,6,6]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm7[2],ymm3[3,4],ymm7[5],ymm3[6,7,8,9],ymm7[10],ymm3[11,12],ymm7[13],ymm3[14,15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm3, %ymm25
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm30, %zmm10, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vpermd %zmm3, %zmm13, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm6, %zmm0, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm11[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm3 = ymm15[3,3,3,3,7,7,7,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3],ymm0[4],ymm3[5,6,7,8],ymm0[9],ymm3[10,11],ymm0[12],ymm3[13,14,15]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0,1],ymm0[2],ymm6[3,4],ymm0[5],ymm6[6,7,8,9],ymm0[10],ymm6[11,12],ymm0[13],ymm6[14,15]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm9[0,0,2,1,4,4,6,5]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %ymm9, %ymm10
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3],ymm6[4,5],ymm8[6],ymm6[7,8,9,10],ymm8[11],ymm6[12,13],ymm8[14],ymm6[15]
+; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm0, %zmm12, %zmm6
+; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm9[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm12[0,1,1,3,4,5,5,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0,1],ymm0[2],ymm8[3,4],ymm0[5],ymm8[6,7,8,9],ymm0[10],ymm8[11,12],ymm0[13],ymm8[14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm24
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,26,27,u,u,24,25,26,27,u,u,26,27,26,27]
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm12[2,2,2,2,6,6,6,6]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm8[2],ymm0[3,4],ymm8[5],ymm0[6,7,8,9],ymm8[10],ymm0[11,12],ymm8[13],ymm0[14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm20
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm19, %zmm21, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vbroadcasti64x4 {{.*#+}} zmm8 = [0,13,4,0,0,14,5,0,0,13,4,0,0,14,5,0]
+; AVX512F-ONLY-FAST-NEXT: # zmm8 = mem[0,1,2,3,0,1,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermd %zmm0, %zmm8, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm6, %zmm1, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm9[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6,7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13,14,15]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,2,3,0,1,2,3,6,7,4,5,6,7,4,5]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm3, %xmm1, %xmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,0,1,2,3,6,7,4,5,6,7,4,5]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm1, %xmm4, %xmm0
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm16
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm3, %xmm5, %xmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm17
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rcx), %xmm6
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rdx), %xmm10
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm10[0],xmm6[0],xmm10[1],xmm6[1],xmm10[2],xmm6[2],xmm10[3],xmm6[3]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm1, %xmm11, %xmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm0, %ymm21
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm1, %xmm7, %xmm12
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rcx), %xmm9
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rdx), %xmm6
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm1, %xmm4, %xmm0
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm9[3,3,3,3,7,7,7,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7,8,9],ymm4[10],ymm3[11,12],ymm4[13],ymm3[14,15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%r9), %xmm4
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%r8), %xmm5
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm11
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29]
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0,1,2],ymm1[3],ymm4[4,5],ymm1[6],ymm4[7,8,9,10],ymm1[11],ymm4[12,13],ymm1[14],ymm4[15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %ymm22, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb %ymm0, %ymm3, %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm10[3,3,3,3,7,7,7,7]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%r9), %xmm7
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%r8), %xmm8
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm4, %xmm11
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [2,2,2,3,8,8,8,9]
-; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm11, %zmm0, %zmm7
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm12[u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29]
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} ymm13 = ymm14[3,3,3,3,7,7,7,7]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm13[0,1,2],ymm11[3],ymm13[4,5],ymm11[6],ymm13[7,8,9,10],ymm11[11],ymm13[12,13],ymm11[14],ymm13[15]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm11, %zmm0, %zmm1
; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [6,7,3,3,7,7,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpermd %ymm30, %ymm0, %ymm2
+; AVX512F-ONLY-FAST-NEXT: vpermd %ymm19, %ymm0, %ymm2
; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 32(%rax), %ymm11
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm2, %zmm21
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm7, %zmm21
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rdi), %xmm1
-; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rsi), %xmm2
-; AVX512F-ONLY-FAST-NEXT: vprold $16, %xmm2, %xmm7
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm12 = xmm1[1,1,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm12[0,1],xmm7[2],xmm12[3,4],xmm7[5],xmm12[6,7]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm2, %zmm22
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm1, %zmm22
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm8[4],xmm7[4],xmm8[5],xmm7[5],xmm8[6],xmm7[6],xmm8[7],xmm7[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm30, %xmm0
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm23, %xmm4
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm0, %xmm4, %xmm4
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm22, %xmm4
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm0, %xmm4, %xmm11
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm18, %xmm4
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm0, %xmm4, %xmm8
-; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm0, %xmm7, %xmm4
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm0, %xmm12, %xmm12
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,1,1,3,8,8,9,9]
-; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm12, %zmm0, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 36(%rax), %ymm12
-; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 40(%rax), %ymm23
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm23, %zmm12, %zmm23
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm3, %zmm23
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [4,5,2,3,4,5,6,7,8,9,10,11,10,11,8,9]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm24, %xmm0
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm3, %xmm0, %xmm12
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm28, %xmm0
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm3, %xmm0, %xmm0
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm3, %xmm1, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm15, %xmm8
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm5, %xmm11
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rdi), %xmm7
+; AVX512F-ONLY-FAST-NEXT: vmovdqa 32(%rsi), %xmm0
+; AVX512F-ONLY-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm5, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm1, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm29, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm2, %xmm4, %xmm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,1,1,3,8,8,9,9]
+; AVX512F-ONLY-FAST-NEXT: vpermt2q %zmm1, %zmm2, %zmm4
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 36(%rax), %ymm1
+; AVX512F-ONLY-FAST-NEXT: vpbroadcastd 40(%rax), %ymm5
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm10
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm10
+; AVX512F-ONLY-FAST-NEXT: vprold $16, %xmm0, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm4 = xmm7[1,1,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm4[0,1],xmm1[2],xmm4[3,4],xmm1[5],xmm4[6,7]
; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm3, %xmm2, %xmm1
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm6[u,u,4,5,u,u,u,u,6,7,u,u,u,u,8,9]
-; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm15 = xmm10[1,1,2,2]
-; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm15[0],xmm2[1],xmm15[2,3],xmm2[4],xmm15[5,6],xmm2[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm15 = mem[0,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm3 = mem[2,1,3,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm2 = mem[2,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm1 = mem[0,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm10[4],xmm6[4],xmm10[5],xmm6[5],xmm10[6],xmm6[6],xmm10[7],xmm6[7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpshufhw {{.*#+}} ymm10 = ymm4[0,1,2,3,5,5,7,6,8,9,10,11,13,13,15,14]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm10 = ymm10[3,3,3,3]
-; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm4[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,24,25,22,23,22,23,u,u,u,u,u,u,u,u]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm14 = ymm14[2,2,2,2]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm31, %xmm4
-; AVX512F-ONLY-FAST-NEXT: vpshuflw {{.*#+}} xmm13 = xmm4[0,2,3,3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = [4,5,2,3,4,5,6,7,8,9,10,11,10,11,8,9]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm17, %xmm15
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm4, %xmm15, %xmm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm28, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm4, %xmm2, %xmm2
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm23, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm4, %xmm3, %xmm3
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpshufb %xmm4, %xmm0, %xmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm9[u,u,4,5,u,u,u,u,6,7,u,u,u,u,8,9]
+; AVX512F-ONLY-FAST-NEXT: vpshufd {{.*#+}} xmm4 = xmm6[1,1,2,2]
+; AVX512F-ONLY-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm4[0],xmm0[1],xmm4[2,3],xmm0[4],xmm4[5,6],xmm0[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm23 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm23 = mem[0,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm4 = mem[2,1,3,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm3 = mem[2,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm0 = mem[0,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpshufhw {{.*#+}} ymm9 = ymm5[0,1,2,3,5,5,7,6,8,9,10,11,13,13,15,14]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm9[3,3,3,3]
+; AVX512F-ONLY-FAST-NEXT: vpshufb {{.*#+}} ymm15 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,24,25,22,23,22,23,u,u,u,u,u,u,u,u]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm15 = ymm15[2,2,2,2]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm31, %xmm5
+; AVX512F-ONLY-FAST-NEXT: vpshuflw {{.*#+}} xmm14 = xmm5[0,2,3,3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm14 = ymm14[0,0,2,1]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,1,3]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm16, %xmm5
+; AVX512F-ONLY-FAST-NEXT: vpshuflw {{.*#+}} xmm13 = xmm5[0,2,3,3,4,5,6,7]
; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm13 = ymm13[0,0,2,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm12 = ymm12[0,0,1,3]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %xmm19, %xmm4
-; AVX512F-ONLY-FAST-NEXT: vpshuflw {{.*#+}} xmm9 = xmm4[0,2,3,3,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,0,2,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,1,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm4 = mem[2,1,3,2]
-; AVX512F-ONLY-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm5 = mem[2,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm6 = mem[0,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vmovups %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpermpd $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm6 = mem[2,1,3,3]
-; AVX512F-ONLY-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpermpd $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm6 = mem[2,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vmovups %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm6 = mem[0,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpermpd $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm6 = mem[2,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vmovups %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm6 = mem[0,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512F-ONLY-FAST-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm31 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm31 = mem[2,1,3,2]
-; AVX512F-ONLY-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm30 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm30 = mem[2,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm24 = ymm16[0,1,1,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm19 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm19 = mem[0,0,1,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm18 = ymm11[0,0,1,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm6 = mem[0,0,2,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm22 = ymm17[0,1,1,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm20 = ymm20[0,0,1,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm17 = ymm8[0,0,1,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm16 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm16 = mem[0,0,2,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm28 = ymm29[2,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm26 = ymm26[0,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm27 = ymm27[2,1,3,2]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm25 = ymm25[2,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm15, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm3, %zmm2, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm3
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm14
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm13, %zmm3 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm12, %zmm12 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm29 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm3, %zmm29, %zmm12
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm12 # 64-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm9 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm9, %zmm29, %zmm0
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm0 # 64-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm3, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm4, %zmm7
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm3, %zmm7
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm11 # 64-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm4 # 64-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vextracti64x4 $1, %zmm1, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogd $226, 124(%r8){1to8}, %ymm3, %ymm1
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm10
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm0, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vshufi64x2 {{.*#+}} zmm1 = zmm14[0,1,2,3],zmm1[4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm1, %zmm3, %zmm8
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm14 # 64-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm19, %zmm24, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm18, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = [65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm1, %zmm5, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm20, %zmm22, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm16, %zmm17, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm1, %zmm5, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm3, %zmm1, %zmm20
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm6, %zmm1, %zmm22
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm30, %zmm31, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = [0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm1, %zmm5, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm26, %zmm28, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm25, %zmm27, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm1, %zmm5, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm3, %zmm1, %zmm25
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm6, %zmm1, %zmm28
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm3 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm1, %zmm2, %zmm3
-; AVX512F-ONLY-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm1 = mem[0,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm5 = mem[0,1,1,3]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,0,1,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm6 = mem[2,1,3,2]
+; AVX512F-ONLY-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm7 = mem[2,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm5 = mem[0,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vmovups %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpermpd $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX512F-ONLY-FAST-NEXT: # ymm5 = mem[2,1,3,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm6 = mem[0,0,1,1]
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm5, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm6 = mem[0,0,2,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq $208, {{[-0-9]+}}(%r{{[sb]}}p), %ymm26 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm26 = mem[0,0,1,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm27 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm27 = mem[0,0,1,1]
-; AVX512F-ONLY-FAST-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # xmm9 = mem[0,2,3,3,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,0,2,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm10 = mem[0,2,2,3]
-; AVX512F-ONLY-FAST-NEXT: vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm24 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm24 = mem[0,1,1,3]
+; AVX512F-ONLY-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpermpd $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm5 = mem[2,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vmovups %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm5 = mem[0,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512F-ONLY-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm31 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm31 = mem[2,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm30 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm30 = mem[0,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm29 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm29 = mem[2,1,3,2]
+; AVX512F-ONLY-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm28 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm28 = mem[2,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm19 = ymm21[0,1,1,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm18 = ymm18[0,0,1,1]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm17 = ymm8[0,0,1,1]
+; AVX512F-ONLY-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm8 = mem[0,0,2,1]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm21 = ymm12[0,1,1,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm12 = mem[0,0,1,1]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm16 = ymm11[0,0,1,1]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm11 = ymm27[0,0,2,1]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm27 = ymm26[2,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm26 = ymm25[0,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm25 = ymm24[2,1,3,2]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm24 = ymm20[2,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm23, %zmm4
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm3, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm4, %zmm3, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm20 # 32-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm20
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm20, %ymm15
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm14, %zmm4 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm23 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm4, %zmm23, %zmm1
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm1 # 64-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm13, %zmm13 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm2 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm13, %zmm23, %zmm2
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm2 # 64-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm4, %zmm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm6, %zmm6
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm6
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm20 # 64-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm5 # 64-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vpternlogd $226, 124(%r8){1to8}, %ymm4, %ymm0
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm9
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm0, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm15[0,1,2,3],zmm0[4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm0, %zmm4, %zmm9
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm13 # 64-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm18, %zmm19, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm17, %zmm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm7 = [65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm0, %zmm7, %zmm4
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm12, %zmm21, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm16, %zmm8
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm0, %zmm7, %zmm8
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm4, %zmm0, %zmm21
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm8, %zmm0, %zmm17
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm30, %zmm31, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm28, %zmm29, %zmm4
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm7 = [0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm0, %zmm7, %zmm4
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm26, %zmm27, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm24, %zmm25, %zmm8
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm0, %zmm7, %zmm8
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm4, %zmm0, %zmm26
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm8, %zmm0, %zmm29
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm4 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm0, %zmm3, %zmm4
+; AVX512F-ONLY-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm0 = mem[0,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm7 = mem[0,1,1,3]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm0, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm7 = mem[2,1,3,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm8 = mem[0,0,1,1]
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm7, %zmm7
+; AVX512F-ONLY-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm8 = mem[0,0,2,1]
+; AVX512F-ONLY-FAST-NEXT: vpermq $208, {{[-0-9]+}}(%r{{[sb]}}p), %ymm27 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm27 = mem[0,0,1,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm24 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm24 = mem[0,0,1,1]
+; AVX512F-ONLY-FAST-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # xmm11 = mem[0,2,3,3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,0,2,1]
+; AVX512F-ONLY-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm12 = mem[0,2,2,3]
+; AVX512F-ONLY-FAST-NEXT: vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm25 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm25 = mem[0,1,1,3]
; AVX512F-ONLY-FAST-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX512F-ONLY-FAST-NEXT: # ymm15 = mem[2,1,3,3]
; AVX512F-ONLY-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm16 # 32-byte Folded Reload
; AVX512F-ONLY-FAST-NEXT: # ymm16 = mem[0,0,1,1]
-; AVX512F-ONLY-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm17 # 32-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # ymm17 = mem[0,0,2,1]
+; AVX512F-ONLY-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm28 # 32-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # ymm28 = mem[0,0,2,1]
; AVX512F-ONLY-FAST-NEXT: vpermq $208, {{[-0-9]+}}(%r{{[sb]}}p), %ymm18 # 32-byte Folded Reload
; AVX512F-ONLY-FAST-NEXT: # ymm18 = mem[0,0,1,3]
; AVX512F-ONLY-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm19 # 32-byte Folded Reload
; AVX512F-ONLY-FAST-NEXT: # ymm19 = mem[0,0,1,1]
-; AVX512F-ONLY-FAST-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX512F-ONLY-FAST-NEXT: # xmm13 = mem[0,2,3,3,4,5,6,7]
-; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm13 = ymm13[0,0,2,1]
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm1, %zmm2, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm24, %zmm10, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm16, %zmm15, %zmm10
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm1, %zmm2, %zmm10
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535]
+; AVX512F-ONLY-FAST-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; AVX512F-ONLY-FAST-NEXT: # xmm14 = mem[0,2,3,3,4,5,6,7]
+; AVX512F-ONLY-FAST-NEXT: vpermq {{.*#+}} ymm14 = ymm14[0,0,2,1]
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm0, %zmm3, %zmm7
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm25, %zmm12, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm16, %zmm15, %zmm12
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm0, %zmm3, %zmm12
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535]
; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm5, %zmm1, %zmm15
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm10, %zmm1, %zmm21
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm26, %zmm6, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm27, %zmm2
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm1, %zmm29, %zmm2
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm18, %zmm17, %zmm1
-; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm13, %zmm19, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm1, %zmm29, %zmm5
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0]
-; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm2, %zmm1, %zmm6
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm5, %zmm1, %zmm23
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm7, %zmm4
-; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm3, %zmm14
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm7, %zmm0, %zmm15
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm12, %zmm0, %zmm22
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm27, %zmm8, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm24, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm0, %zmm23, %zmm3
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm18, %zmm28, %zmm0
+; AVX512F-ONLY-FAST-NEXT: vinserti64x4 $1, %ymm14, %zmm19, %zmm7
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $226, %zmm0, %zmm23, %zmm7
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0]
+; AVX512F-ONLY-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm3, %zmm0, %zmm8
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $184, %zmm7, %zmm0, %zmm10
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm6, %zmm5
+; AVX512F-ONLY-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm13
; AVX512F-ONLY-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm4, 320(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm23, 256(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm21, 192(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm28, 128(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm0, 64(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm22, (%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm20, 448(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm6, 704(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm5, 320(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm10, 256(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm22, 192(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm29, 128(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm2, 64(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm17, (%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm21, 448(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm8, 704(%rax)
; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm15, 640(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm25, 576(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm12, 512(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm14, 384(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm11, 768(%rax)
-; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm8, 832(%rax)
-; AVX512F-ONLY-FAST-NEXT: addq $2264, %rsp # imm = 0x8D8
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm26, 576(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm1, 512(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm13, 384(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm20, 768(%rax)
+; AVX512F-ONLY-FAST-NEXT: vmovdqa64 %zmm9, 832(%rax)
+; AVX512F-ONLY-FAST-NEXT: addq $2200, %rsp # imm = 0x898
; AVX512F-ONLY-FAST-NEXT: vzeroupper
; AVX512F-ONLY-FAST-NEXT: retq
;
; AVX512DQ-FAST-LABEL: store_i16_stride7_vf64:
; AVX512DQ-FAST: # %bb.0:
-; AVX512DQ-FAST-NEXT: subq $2264, %rsp # imm = 0x8D8
-; AVX512DQ-FAST-NEXT: vmovdqa 96(%rsi), %ymm9
-; AVX512DQ-FAST-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX512DQ-FAST-NEXT: vmovdqa 96(%rcx), %ymm2
-; AVX512DQ-FAST-NEXT: vmovdqa 96(%rdx), %ymm8
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128]
-; AVX512DQ-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm0
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm23
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u>
-; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm8, %ymm2
-; AVX512DQ-FAST-NEXT: vporq %ymm0, %ymm2, %ymm16
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128,128,128,128,128]
-; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm9, %ymm5
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u,16,17,18,19>
-; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm6
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm1, %ymm18
-; AVX512DQ-FAST-NEXT: vporq %ymm5, %ymm6, %ymm17
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm6 = [128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128]
+; AVX512DQ-FAST-NEXT: subq $2200, %rsp # imm = 0x898
+; AVX512DQ-FAST-NEXT: vmovdqa 96(%rsi), %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa 96(%rdi), %ymm6
+; AVX512DQ-FAST-NEXT: vmovdqa 96(%rcx), %ymm3
+; AVX512DQ-FAST-NEXT: vmovdqa 96(%rdx), %ymm15
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128]
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm3, %ymm25
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm7 = <u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u>
+; AVX512DQ-FAST-NEXT: vpshufb %ymm7, %ymm15, %ymm3
+; AVX512DQ-FAST-NEXT: vporq %ymm0, %ymm3, %ymm17
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128,128,128,128,128,128,128,128,128]
+; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm4
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm1, %ymm20
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u,u,u,u,u,16,17,18,19>
+; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm6, %ymm5
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm6, %ymm23
+; AVX512DQ-FAST-NEXT: vporq %ymm4, %ymm5, %ymm16
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm5 = [128,128,128,128,128,128,128,128,128,128,128,128,14,15,128,128,128,128,128,128,128,128,128,128,128,128,16,17,128,128,128,128]
; AVX512DQ-FAST-NEXT: vmovdqa 64(%r9), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm1, %ymm5
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm7
+; AVX512DQ-FAST-NEXT: vpshufb %ymm5, %ymm1, %ymm4
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm6
; AVX512DQ-FAST-NEXT: vmovdqa 64(%r8), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm15 = <u,u,u,u,u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u>
-; AVX512DQ-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm6
-; AVX512DQ-FAST-NEXT: vporq %ymm5, %ymm6, %ymm24
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm8 = <u,u,u,u,u,u,u,u,12,13,14,15,128,128,u,u,u,u,u,u,u,u,u,u,16,17,128,128,u,u,u,u>
+; AVX512DQ-FAST-NEXT: vpshufb %ymm8, %ymm1, %ymm5
+; AVX512DQ-FAST-NEXT: vporq %ymm4, %ymm5, %ymm24
; AVX512DQ-FAST-NEXT: vmovdqa 64(%rcx), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm5
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm4
; AVX512DQ-FAST-NEXT: vmovdqa 64(%rdx), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm6
-; AVX512DQ-FAST-NEXT: vpor %ymm5, %ymm6, %ymm1
+; AVX512DQ-FAST-NEXT: vpshufb %ymm7, %ymm1, %ymm5
+; AVX512DQ-FAST-NEXT: vpor %ymm4, %ymm5, %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqa 64(%rsi), %ymm1
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm1, %ymm5
+; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm4
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm1, %ymm30
; AVX512DQ-FAST-NEXT: vmovdqa 64(%rdi), %ymm14
-; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm14, %ymm10
-; AVX512DQ-FAST-NEXT: vpor %ymm5, %ymm10, %ymm1
+; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm14, %ymm5
+; AVX512DQ-FAST-NEXT: vpor %ymm4, %ymm5, %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqa (%r9), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm7, %ymm1, %ymm5
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm7, %ymm11
+; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm1, %ymm4
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm6, %ymm10
; AVX512DQ-FAST-NEXT: vmovdqa (%r8), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm15, %ymm1, %ymm10
-; AVX512DQ-FAST-NEXT: vporq %ymm5, %ymm10, %ymm19
+; AVX512DQ-FAST-NEXT: vpshufb %ymm8, %ymm1, %ymm5
+; AVX512DQ-FAST-NEXT: vporq %ymm4, %ymm5, %ymm31
; AVX512DQ-FAST-NEXT: vmovdqa (%rcx), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm4, %ymm1, %ymm5
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm4
; AVX512DQ-FAST-NEXT: vmovdqa (%rdx), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm10
-; AVX512DQ-FAST-NEXT: vpor %ymm5, %ymm10, %ymm1
+; AVX512DQ-FAST-NEXT: vpshufb %ymm7, %ymm1, %ymm5
+; AVX512DQ-FAST-NEXT: vpor %ymm4, %ymm5, %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqa (%rsi), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm1, %ymm5
+; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm4
; AVX512DQ-FAST-NEXT: vmovdqa (%rdi), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm10
-; AVX512DQ-FAST-NEXT: vpor %ymm5, %ymm10, %ymm1
+; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm1, %ymm5
+; AVX512DQ-FAST-NEXT: vpor %ymm4, %ymm5, %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqa 32(%rcx), %ymm13
-; AVX512DQ-FAST-NEXT: vpshufb %ymm4, %ymm13, %ymm4
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm13, %ymm2
; AVX512DQ-FAST-NEXT: vmovdqa 32(%rdx), %ymm12
-; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm12, %ymm3
-; AVX512DQ-FAST-NEXT: vpor %ymm4, %ymm3, %ymm1
+; AVX512DQ-FAST-NEXT: vpshufb %ymm7, %ymm12, %ymm4
+; AVX512DQ-FAST-NEXT: vpor %ymm2, %ymm4, %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa 32(%rsi), %ymm10
-; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm10, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa 32(%rsi), %ymm11
+; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm11, %ymm2
; AVX512DQ-FAST-NEXT: vmovdqa 32(%rdi), %ymm7
-; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm7, %ymm2
-; AVX512DQ-FAST-NEXT: vpor %ymm0, %ymm2, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm7, %ymm0
+; AVX512DQ-FAST-NEXT: vpor %ymm2, %ymm0, %ymm0
; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqa 32(%r8), %ymm6
-; AVX512DQ-FAST-NEXT: vpshufb %ymm15, %ymm6, %ymm0
-; AVX512DQ-FAST-NEXT: vmovdqa 32(%r9), %ymm4
-; AVX512DQ-FAST-NEXT: vpshufb %ymm11, %ymm4, %ymm15
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm11, %ymm5
-; AVX512DQ-FAST-NEXT: vporq %ymm15, %ymm0, %ymm22
-; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
-; AVX512DQ-FAST-NEXT: # ymm1 = mem[0,1,0,1]
-; AVX512DQ-FAST-NEXT: vpshufb %ymm1, %ymm4, %ymm0
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm1, %ymm31
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm15 = ymm6[3,3,3,3,7,7,7,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm15[2],ymm0[3,4],ymm15[5],ymm0[6,7,8,9],ymm15[10],ymm0[11,12],ymm15[13],ymm0[14,15]
-; AVX512DQ-FAST-NEXT: vprold $16, %ymm4, %ymm15
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm6[1,2,2,3,5,6,6,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm11 = ymm11[0,1],ymm15[2],ymm11[3,4],ymm15[5],ymm11[6,7,8,9],ymm15[10],ymm11[11,12],ymm15[13],ymm11[14,15]
+; AVX512DQ-FAST-NEXT: vpshufb %ymm8, %ymm6, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa 32(%r9), %ymm5
+; AVX512DQ-FAST-NEXT: vpshufb %ymm10, %ymm5, %ymm9
+; AVX512DQ-FAST-NEXT: vporq %ymm9, %ymm0, %ymm22
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm9 = ymm6[3,3,3,3,7,7,7,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm9[2],ymm0[3,4],ymm9[5],ymm0[6,7,8,9],ymm9[10],ymm0[11,12],ymm9[13],ymm0[14,15]
+; AVX512DQ-FAST-NEXT: vprold $16, %ymm5, %ymm9
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm6[1,2,2,3,5,6,6,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4],ymm9[5],ymm8[6,7,8,9],ymm9[10],ymm8[11,12],ymm9[13],ymm8[14,15]
; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [2,1,3,2,10,10,10,11]
-; AVX512DQ-FAST-NEXT: vpermi2q %zmm0, %zmm11, %zmm1
+; AVX512DQ-FAST-NEXT: vpermi2q %zmm0, %zmm8, %zmm1
; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = <10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u>
-; AVX512DQ-FAST-NEXT: vpshufb %ymm1, %ymm9, %ymm0
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm1, %ymm29
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm18, %ymm1
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm18[2,2,2,2,6,6,6,6]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0],ymm0[1],ymm11[2,3],ymm0[4],ymm11[5,6,7,8],ymm0[9],ymm11[10,11],ymm0[12],ymm11[13,14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u>
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm20, %ymm1
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm29
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm23[2,2,2,2,6,6,6,6]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm0[1],ymm8[2,3],ymm0[4],ymm8[5,6,7,8],ymm0[9],ymm8[10,11],ymm0[12],ymm8[13,14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29>
-; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm9, %ymm0
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm3, %ymm25
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm18[3,3,3,3,7,7,7,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0,1,2],ymm0[3],ymm11[4,5],ymm0[6],ymm11[7,8,9,10],ymm0[11],ymm11[12,13],ymm0[14],ymm11[15]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29>
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm20
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm23[3,3,3,3,7,7,7,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0,1,2],ymm0[3],ymm8[4,5],ymm0[6],ymm8[7,8,9,10],ymm0[11],ymm8[12,13],ymm0[14],ymm8[15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27,22,23,26,27,0,0,24,25,26,27,0,0,26,27,26,27]
; AVX512DQ-FAST-NEXT: # ymm2 = mem[0,1,0,1]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm23, %ymm3
-; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm0
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm20
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm8[2,2,2,2,6,6,6,6]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm11[2],ymm0[3,4],ymm11[5],ymm0[6,7,8,9],ymm11[10],ymm0[11,12],ymm11[13],ymm0[14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm25, %ymm4
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm21
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm15[2,2,2,2,6,6,6,6]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm8[2],ymm0[3,4],ymm8[5],ymm0[6,7,8,9],ymm8[10],ymm0[11,12],ymm8[13],ymm0[14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = <14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u>
-; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm3, %ymm0
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm21
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm8[3,3,3,3,7,7,7,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm11[0],ymm0[1],ymm11[2,3],ymm0[4],ymm11[5,6,7,8],ymm0[9],ymm11[10,11],ymm0[12],ymm11[13,14,15]
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm4, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm25
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm15[3,3,3,3,7,7,7,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0],ymm0[1],ymm8[2,3],ymm0[4],ymm8[5,6,7,8],ymm0[9],ymm8[10,11],ymm0[12],ymm8[13,14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm3[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm8[0,1,1,3,4,5,5,7]
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm4[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm15[0,1,1,3,4,5,5,7]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0,1],ymm0[2],ymm8[3,4],ymm0[5],ymm8[6,7,8,9],ymm0[10],ymm8[11,12],ymm0[13],ymm8[14,15]
; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,1,3,2]
; AVX512DQ-FAST-NEXT: vbroadcasti128 {{.*#+}} ymm2 = [18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
; AVX512DQ-FAST-NEXT: # ymm2 = mem[0,1,0,1]
-; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm9, %ymm8
+; AVX512DQ-FAST-NEXT: vpshufb %ymm2, %ymm1, %ymm8
; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm2, %ymm18
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm11 = ymm1[1,1,1,1,5,5,5,5]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm8[0,1],ymm11[2],ymm8[3,4],ymm11[5],ymm8[6,7,8,9],ymm11[10],ymm8[11,12],ymm11[13],ymm8[14,15]
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm9 = ymm23[1,1,1,1,5,5,5,5]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm8 = ymm8[0,1],ymm9[2],ymm8[3,4],ymm9[5],ymm8[6,7,8,9],ymm9[10],ymm8[11,12],ymm9[13],ymm8[14,15]
; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm8 = ymm8[2,2,2,3]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm16, %zmm0
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm17, %zmm8
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm17, %zmm0
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm16, %zmm8
; AVX512DQ-FAST-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm8
; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <5,u,u,u,6,u,u,6>
-; AVX512DQ-FAST-NEXT: vmovdqa 96(%r8), %ymm11
-; AVX512DQ-FAST-NEXT: vpermd %ymm11, %ymm0, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa 96(%r8), %ymm9
+; AVX512DQ-FAST-NEXT: vpermd %ymm9, %ymm0, %ymm0
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = <u,4,u,u,u,5,u,u>
-; AVX512DQ-FAST-NEXT: vpermd %ymm11, %ymm0, %ymm0
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm11[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm11[14,15],zero,zero,ymm11[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm11[16,17],zero,zero,ymm11[u,u],zero,zero
+; AVX512DQ-FAST-NEXT: vpermd %ymm9, %ymm0, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm9 = ymm9[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm9[14,15],zero,zero,ymm9[u,u],zero,zero,zero,zero,zero,zero,zero,zero,ymm9[16,17],zero,zero,ymm9[u,u],zero,zero
; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} ymm16 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm16, %ymm8, %ymm11
+; AVX512DQ-FAST-NEXT: vpternlogq $248, %ymm16, %ymm8, %ymm9
; AVX512DQ-FAST-NEXT: vmovdqa 96(%r9), %ymm1
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm5, %ymm1, %ymm15
-; AVX512DQ-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm11, %ymm15
+; AVX512DQ-FAST-NEXT: vpshufb %ymm10, %ymm1, %ymm15
+; AVX512DQ-FAST-NEXT: vpternlogq $248, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm9, %ymm15
; AVX512DQ-FAST-NEXT: vextracti64x4 $1, %zmm8, %ymm8
; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm8, %ymm0
; AVX512DQ-FAST-NEXT: vprold $16, %ymm1, %ymm8
@@ -12662,13 +12656,13 @@
; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535]
; AVX512DQ-FAST-NEXT: vpbroadcastd 72(%rax), %ymm0
; AVX512DQ-FAST-NEXT: vpandn %ymm0, %ymm1, %ymm8
-; AVX512DQ-FAST-NEXT: vmovdqa 64(%rax), %ymm9
-; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm9, %ymm11
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm8, %zmm27
+; AVX512DQ-FAST-NEXT: vmovdqa 64(%rax), %ymm10
+; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm10, %ymm9
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm8, %zmm27
; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm24, %zmm0, %zmm15
; AVX512DQ-FAST-NEXT: vmovdqa 64(%r9), %xmm2
-; AVX512DQ-FAST-NEXT: vmovdqa 64(%r8), %xmm11
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm11[4],xmm2[4],xmm11[5],xmm2[5],xmm11[6],xmm2[6],xmm11[7],xmm2[7]
+; AVX512DQ-FAST-NEXT: vmovdqa 64(%r8), %xmm9
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm9[4],xmm2[4],xmm9[5],xmm2[5],xmm9[6],xmm2[6],xmm9[7],xmm2[7]
; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm2, %xmm28
; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
@@ -12681,16 +12675,15 @@
; AVX512DQ-FAST-NEXT: vpandn %ymm0, %ymm1, %ymm0
; AVX512DQ-FAST-NEXT: vmovdqa (%rax), %ymm1
; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm1, %ymm15
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm1, %ymm30
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm1, %ymm19
; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm15, %zmm0, %zmm1
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm19, %zmm0, %zmm0
-; AVX512DQ-FAST-NEXT: vmovdqa (%r9), %xmm5
-; AVX512DQ-FAST-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm31, %zmm0, %zmm0
+; AVX512DQ-FAST-NEXT: vmovdqa (%r9), %xmm4
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqa (%r8), %xmm8
; AVX512DQ-FAST-NEXT: vmovdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm15 = xmm8[4],xmm5[4],xmm8[5],xmm5[5],xmm8[6],xmm5[6],xmm8[7],xmm5[7]
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm15 = xmm8[4],xmm4[4],xmm8[5],xmm4[5],xmm8[6],xmm4[6],xmm8[7],xmm4[7]
; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm15, %xmm15
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm2, %xmm19
; AVX512DQ-FAST-NEXT: vpermt2q %zmm0, %zmm24, %zmm15
; AVX512DQ-FAST-NEXT: vpternlogq $248, %zmm26, %zmm15, %zmm1
; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
@@ -12698,13 +12691,13 @@
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm15 = ymm12[0,1,1,3,4,5,5,7]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm15[0,1],ymm0[2],ymm15[3,4],ymm0[5],ymm15[6,7,8,9],ymm0[10],ymm15[11,12],ymm0[13],ymm15[14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm18, %ymm5
-; AVX512DQ-FAST-NEXT: vpshufb %ymm5, %ymm10, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm18, %ymm4
+; AVX512DQ-FAST-NEXT: vpshufb %ymm4, %ymm11, %ymm0
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm15 = ymm7[1,1,1,1,5,5,5,5]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm15[2],ymm0[3,4],ymm15[5],ymm0[6,7,8,9],ymm15[10],ymm0[11,12],ymm15[13],ymm0[14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vpbroadcastd {{.*#+}} ymm8 = [18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
-; AVX512DQ-FAST-NEXT: vpshufb %ymm8, %ymm4, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb %ymm8, %ymm5, %ymm0
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm6[0,0,2,1,4,4,6,5]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3],ymm0[4,5],ymm2[6],ymm0[7,8,9,10],ymm2[11],ymm0[12,13],ymm2[14],ymm0[15]
; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = <0,1,u,3,10,10,11,11>
@@ -12716,505 +12709,506 @@
; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm0, %ymm1
; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm2, %zmm1, %zmm1
; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm29, %ymm4
-; AVX512DQ-FAST-NEXT: vpshufb %ymm4, %ymm10, %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm29, %ymm5
+; AVX512DQ-FAST-NEXT: vpshufb %ymm5, %ymm11, %ymm1
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm7[2,2,2,2,6,6,6,6]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm25, %ymm15
-; AVX512DQ-FAST-NEXT: vpshufb %ymm15, %ymm10, %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm20, %ymm6
+; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm11, %ymm1
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5],ymm1[6],ymm2[7,8,9,10],ymm1[11],ymm2[12,13],ymm1[14],ymm2[15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm20, %ymm3
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm21, %ymm3
; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm13, %ymm1
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm12[2,2,2,2,6,6,6,6]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm21, %ymm6
-; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm13, %ymm1
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm25, %ymm15
+; AVX512DQ-FAST-NEXT: vpshufb %ymm15, %ymm13, %ymm1
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm12[3,3,3,3,7,7,7,7]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa64 (%rax), %zmm10
+; AVX512DQ-FAST-NEXT: vmovdqa64 (%rax), %zmm21
; AVX512DQ-FAST-NEXT: vbroadcasti32x8 {{.*#+}} zmm1 = [14,21,0,0,15,22,0,15,14,21,0,0,15,22,0,15]
; AVX512DQ-FAST-NEXT: # zmm1 = mem[0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpermi2d %zmm0, %zmm10, %zmm1
+; AVX512DQ-FAST-NEXT: vpermi2d %zmm0, %zmm21, %zmm1
; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vpshufb %ymm5, %ymm12, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm30, %ymm2
+; AVX512DQ-FAST-NEXT: vpshufb %ymm4, %ymm2, %ymm0
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm14[1,1,1,1,5,5,5,5]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7,8,9],ymm1[10],ymm0[11,12],ymm1[13],ymm0[14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm4, %ymm12, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb %ymm5, %ymm2, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm30, %ymm4
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm14[2,2,2,2,6,6,6,6]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6,7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13,14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm4[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm5[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm2[0,1,1,3,4,5,5,7]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7,8,9],ymm0[10],ymm1[11,12],ymm0[13],ymm1[14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm4, %ymm0
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm4, %ymm7
+; AVX512DQ-FAST-NEXT: vpshufb %ymm3, %ymm5, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm5, %ymm11
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm2[2,2,2,2,6,6,6,6]
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm2, %ymm5
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm2, %ymm7
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7,8,9],ymm1[10],ymm0[11,12],ymm1[13],ymm0[14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX512DQ-FAST-NEXT: vprold $16, %ymm2, %ymm0
-; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm4[1,2,2,3,5,6,6,7]
+; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm5[1,2,2,3,5,6,6,7]
; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7,8,9],ymm0[10],ymm1[11,12],ymm0[13],ymm1[14,15]
; AVX512DQ-FAST-NEXT: vpshufb %ymm8, %ymm2, %ymm1
; AVX512DQ-FAST-NEXT: vmovdqa %ymm2, %ymm3
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[0,0,2,1,4,4,6,5]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7,8,9,10],ymm2[11],ymm1[12,13],ymm2[14],ymm1[15]
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm21 = [2,2,3,3,10,9,11,10]
-; AVX512DQ-FAST-NEXT: vpermt2q %zmm0, %zmm21, %zmm1
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm23, %zmm0
-; AVX512DQ-FAST-NEXT: vbroadcasti32x8 {{.*#+}} zmm13 = [0,13,4,0,0,14,5,0,0,13,4,0,0,14,5,0]
-; AVX512DQ-FAST-NEXT: # zmm13 = mem[0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpermd %zmm0, %zmm13, %zmm2
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm1, %zmm0, %zmm2
-; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm7, %ymm1
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm5[3,3,3,3,7,7,7,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6,7,8],ymm1[9],ymm2[10,11],ymm1[12],ymm2[13,14,15]
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %ymm15, %ymm12, %ymm1
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm5[0,0,2,1,4,4,6,5]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm2 = ymm1[0,1,2],ymm2[3],ymm1[4,5],ymm2[6],ymm1[7,8,9,10],ymm2[11],ymm1[12,13],ymm2[14],ymm1[15]
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm12 = [2,2,3,3,10,9,11,10]
+; AVX512DQ-FAST-NEXT: vpermt2q %zmm0, %zmm12, %zmm2
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm23, %zmm0
+; AVX512DQ-FAST-NEXT: vbroadcasti32x8 {{.*#+}} zmm1 = [0,13,4,0,0,14,5,0,0,13,4,0,0,14,5,0]
+; AVX512DQ-FAST-NEXT: # zmm1 = mem[0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpermd %zmm0, %zmm1, %zmm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm2, %zmm1, %zmm0
+; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FAST-NEXT: vpshufb %ymm15, %ymm11, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm7[3,3,3,3,7,7,7,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6,7,8],ymm0[9],ymm2[10,11],ymm0[12],ymm2[13,14,15]
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpshufb %ymm6, %ymm4, %ymm0
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm14[3,3,3,3,7,7,7,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm2[0,1,2],ymm1[3],ymm2[4,5],ymm1[6],ymm2[7,8,9,10],ymm1[11],ymm2[12,13],ymm1[14],ymm2[15]
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm31, %ymm1
-; AVX512DQ-FAST-NEXT: vpshufb %ymm1, %ymm3, %ymm1
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[3,3,3,3,7,7,7,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm4 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7,8,9],ymm2[10],ymm1[11,12],ymm2[13],ymm1[14,15]
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [2,2,2,3,8,8,8,9]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3],ymm2[4,5],ymm0[6],ymm2[7,8,9,10],ymm0[11],ymm2[12,13],ymm0[14],ymm2[15]
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vbroadcasti64x2 {{.*#+}} ymm22 = [26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
+; AVX512DQ-FAST-NEXT: # ymm22 = mem[0,1,0,1]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm22, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm3, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm2 = ymm5[3,3,3,3,7,7,7,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3,4],ymm2[5],ymm0[6,7,8,9],ymm2[10],ymm0[11,12],ymm2[13],ymm0[14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = [2,2,2,3,8,8,8,9]
; AVX512DQ-FAST-NEXT: vmovdqa 96(%r9), %xmm5
; AVX512DQ-FAST-NEXT: vmovdqa 96(%r8), %xmm6
; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,6,7,4,5,6,7,4,5,12,13,14,15]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm7, %xmm1
-; AVX512DQ-FAST-NEXT: vpermt2q %zmm1, %zmm3, %zmm4
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm1 = [6,7,3,3,7,7,6,7]
-; AVX512DQ-FAST-NEXT: vpermd %ymm9, %ymm1, %ymm1
-; AVX512DQ-FAST-NEXT: vpbroadcastd 96(%rax), %ymm3
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm1, %zmm3
-; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm3
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,2,3,6,7,4,5,6,7,4,5,12,13,14,15]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm3, %xmm7, %xmm2
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm3, %xmm14
+; AVX512DQ-FAST-NEXT: vpermt2q %zmm2, %zmm4, %zmm0
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm2 = [6,7,3,3,7,7,6,7]
+; AVX512DQ-FAST-NEXT: vpermd %ymm10, %ymm2, %ymm3
+; AVX512DQ-FAST-NEXT: vpbroadcastd 96(%rax), %ymm4
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm3, %zmm3
+; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm3
; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa 96(%rsi), %xmm3
+; AVX512DQ-FAST-NEXT: vmovdqa 96(%rsi), %xmm0
; AVX512DQ-FAST-NEXT: vmovdqa 96(%rdi), %xmm4
-; AVX512DQ-FAST-NEXT: vprold $16, %xmm3, %xmm9
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm12 = xmm4[1,1,2,3]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm12[0,1],xmm9[2],xmm12[3,4],xmm9[5],xmm12[6,7]
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm9, %xmm23
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
-; AVX512DQ-FAST-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa 96(%rcx), %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa 96(%rdx), %xmm4
-; AVX512DQ-FAST-NEXT: vpbroadcastq {{.*#+}} xmm12 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm12, %xmm3, %xmm9
-; AVX512DQ-FAST-NEXT: vmovdqa %xmm12, %xmm8
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm12 = xmm4[1,1,2,2]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm9 = xmm12[0],xmm9[1],xmm12[2,3],xmm9[4],xmm12[5,6],xmm9[7]
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
-; AVX512DQ-FAST-NEXT: vmovdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm19, %xmm4
-; AVX512DQ-FAST-NEXT: vpshufb %xmm4, %xmm3, %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm5 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm5, %xmm7, %xmm4
-; AVX512DQ-FAST-NEXT: vmovdqa %xmm5, %xmm7
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = [0,1,1,3,8,8,9,9]
-; AVX512DQ-FAST-NEXT: vpermt2q %zmm3, %zmm5, %zmm4
-; AVX512DQ-FAST-NEXT: vpbroadcastd 100(%rax), %ymm3
-; AVX512DQ-FAST-NEXT: vpbroadcastd 104(%rax), %ymm5
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm3, %zmm3
-; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm3
-; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa 64(%rcx), %xmm3
+; AVX512DQ-FAST-NEXT: vprold $16, %xmm0, %xmm10
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm11 = xmm4[1,1,2,3]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm11[0,1],xmm10[2],xmm11[3,4],xmm10[5],xmm11[6,7]
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm3, %xmm30
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm0, %xmm23
+; AVX512DQ-FAST-NEXT: vmovdqa 96(%rcx), %xmm0
+; AVX512DQ-FAST-NEXT: vmovdqa 96(%rdx), %xmm10
+; AVX512DQ-FAST-NEXT: vpbroadcastq {{.*#+}} xmm15 = [6,7,4,5,0,0,8,9,6,7,4,5,0,0,8,9]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm15, %xmm0, %xmm4
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm15, %xmm8
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm11 = xmm10[1,1,2,2]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm11[0],xmm4[1],xmm11[2,3],xmm4[4],xmm11[5,6],xmm4[7]
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm10[4],xmm0[4],xmm10[5],xmm0[5],xmm10[6],xmm0[6],xmm10[7],xmm0[7]
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm15 = [0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm15, %xmm7, %xmm5
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm15, %xmm7
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,1,1,3,8,8,9,9]
+; AVX512DQ-FAST-NEXT: vpermt2q %zmm0, %zmm3, %zmm5
+; AVX512DQ-FAST-NEXT: vpbroadcastd 100(%rax), %ymm0
+; AVX512DQ-FAST-NEXT: vpbroadcastd 104(%rax), %ymm6
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm0, %zmm0
+; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm5, %zmm0
+; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FAST-NEXT: vmovdqa 64(%rcx), %xmm0
; AVX512DQ-FAST-NEXT: vmovdqa 64(%rdx), %xmm5
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm4, %xmm31
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
-; AVX512DQ-FAST-NEXT: vmovdqa %xmm8, %xmm9
-; AVX512DQ-FAST-NEXT: vpshufb %xmm8, %xmm3, %xmm3
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm3, %xmm31
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm11 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm8, %xmm2
+; AVX512DQ-FAST-NEXT: vpshufb %xmm8, %xmm0, %xmm0
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[1,1,2,2]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm5[0],xmm3[1],xmm5[2,3],xmm3[4],xmm5[5,6],xmm3[7]
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa 64(%rdi), %xmm3
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0],xmm0[1],xmm5[2,3],xmm0[4],xmm5[5,6],xmm0[7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm18
+; AVX512DQ-FAST-NEXT: vmovdqa 64(%rdi), %xmm0
; AVX512DQ-FAST-NEXT: vmovdqa 64(%rsi), %xmm5
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm6 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm6, %xmm24
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm6, %xmm22
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm15 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm15, %xmm17
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm15 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
; AVX512DQ-FAST-NEXT: vprold $16, %xmm5, %xmm5
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[1,1,2,3]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm5[2],xmm3[3,4],xmm5[5],xmm3[6,7]
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm28, %xmm3
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm11[0],xmm3[0],xmm11[1],xmm3[1],xmm11[2],xmm3[2],xmm11[3],xmm3[3]
-; AVX512DQ-FAST-NEXT: vmovdqa %xmm7, %xmm11
-; AVX512DQ-FAST-NEXT: vpshufb %xmm7, %xmm3, %xmm5
-; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm5[2],xmm0[3,4],xmm5[5],xmm0[6,7]
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm28, %xmm0
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm7, %xmm13
+; AVX512DQ-FAST-NEXT: vpshufb %xmm7, %xmm0, %xmm5
+; AVX512DQ-FAST-NEXT: vpshufb %xmm14, %xmm0, %xmm0
; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm6 = [0,0,0,1,8,9,9,11]
-; AVX512DQ-FAST-NEXT: vpermt2q %zmm5, %zmm6, %zmm3
+; AVX512DQ-FAST-NEXT: vpermt2q %zmm5, %zmm6, %zmm0
; AVX512DQ-FAST-NEXT: vpbroadcastd 64(%rax), %ymm5
; AVX512DQ-FAST-NEXT: vpbroadcastd 68(%rax), %ymm7
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm5, %zmm5
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm7 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm3, %zmm7, %zmm5
-; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa (%rcx), %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa (%rdx), %xmm8
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm8[4],xmm3[4],xmm8[5],xmm3[5],xmm8[6],xmm3[6],xmm8[7],xmm3[7]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm5, %xmm19
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm8[0],xmm3[0],xmm8[1],xmm3[1],xmm8[2],xmm3[2],xmm8[3],xmm3[3]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm9, %xmm3, %xmm3
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm8 = xmm8[1,1,2,2]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm8[0],xmm3[1],xmm8[2,3],xmm3[4],xmm8[5,6],xmm3[7]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm3, %ymm20
-; AVX512DQ-FAST-NEXT: vmovdqa (%rdi), %xmm3
-; AVX512DQ-FAST-NEXT: vmovdqa (%rsi), %xmm8
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm9 = xmm8[4],xmm3[4],xmm8[5],xmm3[5],xmm8[6],xmm3[6],xmm8[7],xmm3[7]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm9, %xmm28
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm9 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm9, %xmm18
-; AVX512DQ-FAST-NEXT: vprold $16, %xmm8, %xmm8
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[1,1,2,3]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm8[2],xmm3[3,4],xmm8[5],xmm3[6,7]
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1],xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm11, %xmm3, %xmm8
-; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm3
-; AVX512DQ-FAST-NEXT: vpermt2q %zmm8, %zmm6, %zmm3
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm5, %zmm3
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm8 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm0, %zmm8, %zmm3
+; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FAST-NEXT: vmovdqa (%rcx), %xmm0
+; AVX512DQ-FAST-NEXT: vmovdqa (%rdx), %xmm5
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm3 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm3, %xmm16
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[1,1,2,2]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm5[0],xmm0[1],xmm5[2,3],xmm0[4],xmm5[5,6],xmm0[7]
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vmovdqa (%rdi), %xmm0
+; AVX512DQ-FAST-NEXT: vmovdqa (%rsi), %xmm9
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm9[4],xmm0[4],xmm9[5],xmm0[5],xmm9[6],xmm0[6],xmm9[7],xmm0[7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm2, %xmm28
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3]
+; AVX512DQ-FAST-NEXT: vprold $16, %xmm9, %xmm9
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm9[2],xmm0[3,4],xmm9[5],xmm0[6,7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm27
+; AVX512DQ-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm13, %xmm0, %xmm9
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm13, %xmm29
+; AVX512DQ-FAST-NEXT: vpshufb %xmm14, %xmm0, %xmm0
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm14, %xmm2
+; AVX512DQ-FAST-NEXT: vpermt2q %zmm9, %zmm6, %zmm0
; AVX512DQ-FAST-NEXT: vpbroadcastd (%rax), %ymm6
-; AVX512DQ-FAST-NEXT: vpbroadcastd 4(%rax), %ymm8
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm6, %zmm6
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm3, %zmm7, %zmm6
-; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm12[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
-; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm7[1,1,1,1,5,5,5,5]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm6[2],ymm3[3,4],ymm6[5],ymm3[6,7,8,9],ymm6[10],ymm3[11,12],ymm6[13],ymm3[14,15]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm3, %ymm29
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm12[10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm7[2,2,2,2,6,6,6,6]
-; AVX512DQ-FAST-NEXT: vmovdqa %ymm7, %ymm14
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm6[0],ymm3[1],ymm6[2,3],ymm3[4],ymm6[5,6,7,8],ymm3[9],ymm6[10,11],ymm3[12],ymm6[13,14,15]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm3, %ymm26
-; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vprold $16, %ymm8, %ymm3
+; AVX512DQ-FAST-NEXT: vpbroadcastd 4(%rax), %ymm9
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm6, %zmm3
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm0, %zmm8, %zmm3
+; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,24,25,26,27,22,23,22,23]
+; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm3[1,1,1,1,5,5,5,5]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm6[2],ymm0[3,4],ymm6[5],ymm0[6,7,8,9],ymm6[10],ymm0[11,12],ymm6[13],ymm0[14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm26
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm13[10,11,8,9,u,u,u,u,u,u,u,u,u,u,u,u,26,27,24,25,u,u,u,u,26,27,24,25,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm3[2,2,2,2,6,6,6,6]
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm3, %ymm14
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0],ymm0[1],ymm6[2,3],ymm0[4],ymm6[5,6,7,8],ymm0[9],ymm6[10,11],ymm0[12],ymm6[13,14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm25
+; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX512DQ-FAST-NEXT: vprold $16, %ymm3, %ymm0
; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm6 = ymm9[1,2,2,3,5,6,6,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm6[0,1],ymm3[2],ymm6[3,4],ymm3[5],ymm6[6,7,8,9],ymm3[10],ymm6[11,12],ymm3[13],ymm6[14,15]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm7 = ymm9[0,0,2,1,4,4,6,5]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0,1,2],ymm7[3],ymm6[4,5],ymm7[6],ymm6[7,8,9,10],ymm7[11],ymm6[12,13],ymm7[14],ymm6[15]
-; AVX512DQ-FAST-NEXT: vpermt2q %zmm3, %zmm21, %zmm6
-; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} ymm3 = ymm11[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
-; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm7 = ymm15[0,1,1,3,4,5,5,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm7[0,1],ymm3[2],ymm7[3,4],ymm3[5],ymm7[6,7,8,9],ymm3[10],ymm7[11,12],ymm3[13],ymm7[14,15]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm3, %ymm27
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm11[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,26,27,u,u,24,25,26,27,u,u,26,27,26,27]
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm7 = ymm15[2,2,2,2,6,6,6,6]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm3 = ymm3[0,1],ymm7[2],ymm3[3,4],ymm7[5],ymm3[6,7,8,9],ymm7[10],ymm3[11,12],ymm7[13],ymm3[14,15]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm3, %ymm25
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm30, %zmm10, %zmm3
-; AVX512DQ-FAST-NEXT: vpermd %zmm3, %zmm13, %zmm3
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm6, %zmm0, %zmm3
-; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm11[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm3 = ymm15[3,3,3,3,7,7,7,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3],ymm0[4],ymm3[5,6,7,8],ymm0[9],ymm3[10,11],ymm0[12],ymm3[13,14,15]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm6[0,1],ymm0[2],ymm6[3,4],ymm0[5],ymm6[6,7,8,9],ymm0[10],ymm6[11,12],ymm0[13],ymm6[14,15]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm6 = ymm3[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,18,19,20,21,18,19,20,21,18,19,20,21,18,19,20,21]
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm9[0,0,2,1,4,4,6,5]
+; AVX512DQ-FAST-NEXT: vmovdqa %ymm9, %ymm10
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3],ymm6[4,5],ymm8[6],ymm6[7,8,9,10],ymm8[11],ymm6[12,13],ymm8[14],ymm6[15]
+; AVX512DQ-FAST-NEXT: vpermt2q %zmm0, %zmm12, %zmm6
+; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm9[0,0,1,1,2,2,3,3,8,8,9,9,10,10,11,11]
+; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm12[0,1,1,3,4,5,5,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm8[0,1],ymm0[2],ymm8[3,4],ymm0[5],ymm8[6,7,8,9],ymm0[10],ymm8[11,12],ymm0[13],ymm8[14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm24
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm9[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,22,23,26,27,u,u,24,25,26,27,u,u,26,27,26,27]
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm8 = ymm12[2,2,2,2,6,6,6,6]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm8[2],ymm0[3,4],ymm8[5],ymm0[6,7,8,9],ymm8[10],ymm0[11,12],ymm8[13],ymm0[14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm20
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm19, %zmm21, %zmm0
+; AVX512DQ-FAST-NEXT: vbroadcasti32x8 {{.*#+}} zmm8 = [0,13,4,0,0,14,5,0,0,13,4,0,0,14,5,0]
+; AVX512DQ-FAST-NEXT: # zmm8 = mem[0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpermd %zmm0, %zmm8, %zmm0
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm6, %zmm1, %zmm0
+; AVX512DQ-FAST-NEXT: vmovdqu64 %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm0 = ymm9[14,15,12,13,u,u,u,u,u,u,u,u,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm1 = ymm12[3,3,3,3,7,7,7,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2,3],ymm0[4],ymm1[5,6,7,8],ymm0[9],ymm1[10,11],ymm0[12],ymm1[13,14,15]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [0,1,2,3,0,1,2,3,6,7,4,5,6,7,4,5]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm3, %xmm1, %xmm0
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm1 = [0,1,2,3,0,1,2,3,6,7,4,5,6,7,4,5]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm1, %xmm4, %xmm0
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm0
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm16
-; AVX512DQ-FAST-NEXT: vpshufb %xmm3, %xmm5, %xmm0
-; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm17
-; AVX512DQ-FAST-NEXT: vmovdqa 32(%rcx), %xmm6
-; AVX512DQ-FAST-NEXT: vmovdqa 32(%rdx), %xmm10
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm10[0],xmm6[0],xmm10[1],xmm6[1],xmm10[2],xmm6[2],xmm10[3],xmm6[3]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm3, %xmm4, %xmm0
+; AVX512DQ-FAST-NEXT: vpshufb %xmm1, %xmm11, %xmm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm0, %ymm21
+; AVX512DQ-FAST-NEXT: vpshufb %xmm1, %xmm7, %xmm12
+; AVX512DQ-FAST-NEXT: vmovdqa 32(%rcx), %xmm9
+; AVX512DQ-FAST-NEXT: vmovdqa 32(%rdx), %xmm6
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm1, %xmm4, %xmm0
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm3 = ymm8[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,26,27,28,29,26,27,28,29,26,27,28,29,30,31,30,31]
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm9[3,3,3,3,7,7,7,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm7 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7,8,9],ymm4[10],ymm3[11,12],ymm4[13],ymm3[14,15]
-; AVX512DQ-FAST-NEXT: vmovdqa 32(%r9), %xmm4
-; AVX512DQ-FAST-NEXT: vmovdqa 32(%r8), %xmm5
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm3, %xmm11
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm1 = ymm13[u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29]
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm14[3,3,3,3,7,7,7,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm4[0,1,2],ymm1[3],ymm4[4,5],ymm1[6],ymm4[7,8,9,10],ymm1[11],ymm4[12,13],ymm1[14],ymm4[15]
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vmovdqa64 %ymm22, %ymm0
+; AVX512DQ-FAST-NEXT: vpshufb %ymm0, %ymm3, %ymm1
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm4 = ymm10[3,3,3,3,7,7,7,7]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm1 = ymm1[0,1],ymm4[2],ymm1[3,4],ymm4[5],ymm1[6,7,8,9],ymm4[10],ymm1[11,12],ymm4[13],ymm1[14,15]
+; AVX512DQ-FAST-NEXT: vmovdqa 32(%r9), %xmm7
+; AVX512DQ-FAST-NEXT: vmovdqa 32(%r8), %xmm8
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm4, %xmm11
; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [2,2,2,3,8,8,8,9]
-; AVX512DQ-FAST-NEXT: vpermt2q %zmm11, %zmm0, %zmm7
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm11 = ymm12[u,u,u,u,u,u,u,u,u,u,u,u,14,15,12,13,u,u,u,u,30,31,28,29,u,u,u,u,30,31,28,29]
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} ymm13 = ymm14[3,3,3,3,7,7,7,7]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} ymm0 = ymm13[0,1,2],ymm11[3],ymm13[4,5],ymm11[6],ymm13[7,8,9,10],ymm11[11],ymm13[12,13],ymm11[14],ymm13[15]
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpermt2q %zmm11, %zmm0, %zmm1
; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm0 = [6,7,3,3,7,7,6,7]
-; AVX512DQ-FAST-NEXT: vpermd %ymm30, %ymm0, %ymm2
+; AVX512DQ-FAST-NEXT: vpermd %ymm19, %ymm0, %ymm2
; AVX512DQ-FAST-NEXT: vpbroadcastd 32(%rax), %ymm11
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm2, %zmm21
-; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm7, %zmm21
-; AVX512DQ-FAST-NEXT: vmovdqa 32(%rdi), %xmm1
-; AVX512DQ-FAST-NEXT: vmovdqa 32(%rsi), %xmm2
-; AVX512DQ-FAST-NEXT: vprold $16, %xmm2, %xmm7
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm12 = xmm1[1,1,2,3]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm12[0,1],xmm7[2],xmm12[3,4],xmm7[5],xmm12[6,7]
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm2, %zmm22
+; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm1, %zmm22
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm1 = xmm8[4],xmm7[4],xmm8[5],xmm7[5],xmm8[6],xmm7[6],xmm8[7],xmm7[7]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm2 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm30, %xmm0
+; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm0, %xmm0
; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm12 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm0 = [0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm23, %xmm4
-; AVX512DQ-FAST-NEXT: vpshufb %xmm0, %xmm4, %xmm4
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm22, %xmm4
-; AVX512DQ-FAST-NEXT: vpshufb %xmm0, %xmm4, %xmm11
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm18, %xmm4
-; AVX512DQ-FAST-NEXT: vpshufb %xmm0, %xmm4, %xmm8
-; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; AVX512DQ-FAST-NEXT: vpshufb %xmm0, %xmm7, %xmm4
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %xmm0, %xmm12, %xmm12
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[0,1,2,3,8,9,10,11,14,15,12,13,14,15,12,13]
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,1,1,3,8,8,9,9]
-; AVX512DQ-FAST-NEXT: vpermt2q %zmm12, %zmm0, %zmm3
-; AVX512DQ-FAST-NEXT: vpbroadcastd 36(%rax), %ymm12
-; AVX512DQ-FAST-NEXT: vpbroadcastd 40(%rax), %ymm23
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm23, %zmm12, %zmm23
-; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm3, %zmm23
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm3 = [4,5,2,3,4,5,6,7,8,9,10,11,10,11,8,9]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm24, %xmm0
-; AVX512DQ-FAST-NEXT: vpshufb %xmm3, %xmm0, %xmm12
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm28, %xmm0
-; AVX512DQ-FAST-NEXT: vpshufb %xmm3, %xmm0, %xmm0
-; AVX512DQ-FAST-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX512DQ-FAST-NEXT: vpshufb %xmm3, %xmm1, %xmm1
+; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm15, %xmm8
+; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm5, %xmm11
+; AVX512DQ-FAST-NEXT: vmovdqa 32(%rdi), %xmm7
+; AVX512DQ-FAST-NEXT: vmovdqa 32(%rsi), %xmm0
+; AVX512DQ-FAST-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
+; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm5, %xmm3
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm1, %xmm1
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm29, %xmm2
+; AVX512DQ-FAST-NEXT: vpshufb %xmm2, %xmm4, %xmm4
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,1,1,3,8,8,9,9]
+; AVX512DQ-FAST-NEXT: vpermt2q %zmm1, %zmm2, %zmm4
+; AVX512DQ-FAST-NEXT: vpbroadcastd 36(%rax), %ymm1
+; AVX512DQ-FAST-NEXT: vpbroadcastd 40(%rax), %ymm5
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm10
+; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm10
+; AVX512DQ-FAST-NEXT: vprold $16, %xmm0, %xmm1
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm4 = xmm7[1,1,2,3]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm4[0,1],xmm1[2],xmm4[3,4],xmm1[5],xmm4[6,7]
; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb %xmm3, %xmm2, %xmm1
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm2 = xmm6[u,u,4,5,u,u,u,u,6,7,u,u,u,u,8,9]
-; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm15 = xmm10[1,1,2,2]
-; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm1 = xmm15[0],xmm2[1],xmm15[2,3],xmm2[4],xmm15[5,6],xmm2[7]
-; AVX512DQ-FAST-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm15 = mem[0,2,2,3]
-; AVX512DQ-FAST-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm3 = mem[2,1,3,3]
-; AVX512DQ-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm2 = mem[2,2,2,3]
-; AVX512DQ-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm1 = mem[0,2,2,3]
-; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm10[4],xmm6[4],xmm10[5],xmm6[5],xmm10[6],xmm6[6],xmm10[7],xmm6[7]
-; AVX512DQ-FAST-NEXT: vmovdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vpshufhw {{.*#+}} ymm10 = ymm4[0,1,2,3,5,5,7,6,8,9,10,11,13,13,15,14]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm10 = ymm10[3,3,3,3]
-; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm14 = ymm4[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,24,25,22,23,22,23,u,u,u,u,u,u,u,u]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm14 = ymm14[2,2,2,2]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm31, %xmm4
-; AVX512DQ-FAST-NEXT: vpshuflw {{.*#+}} xmm13 = xmm4[0,2,3,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} xmm4 = [4,5,2,3,4,5,6,7,8,9,10,11,10,11,8,9]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm17, %xmm15
+; AVX512DQ-FAST-NEXT: vpshufb %xmm4, %xmm15, %xmm1
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm28, %xmm2
+; AVX512DQ-FAST-NEXT: vpshufb %xmm4, %xmm2, %xmm2
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm23, %xmm3
+; AVX512DQ-FAST-NEXT: vpshufb %xmm4, %xmm3, %xmm3
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpshufb %xmm4, %xmm0, %xmm0
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm9[u,u,4,5,u,u,u,u,6,7,u,u,u,u,8,9]
+; AVX512DQ-FAST-NEXT: vpshufd {{.*#+}} xmm4 = xmm6[1,1,2,2]
+; AVX512DQ-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm4[0],xmm0[1],xmm4[2,3],xmm0[4],xmm4[5,6],xmm0[7]
+; AVX512DQ-FAST-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm23 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm23 = mem[0,2,2,3]
+; AVX512DQ-FAST-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm4 = mem[2,1,3,3]
+; AVX512DQ-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm3 = mem[2,2,2,3]
+; AVX512DQ-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm0 = mem[0,2,2,3]
+; AVX512DQ-FAST-NEXT: vpunpckhwd {{.*#+}} xmm5 = xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
+; AVX512DQ-FAST-NEXT: vmovdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX512DQ-FAST-NEXT: vpshufhw {{.*#+}} ymm9 = ymm5[0,1,2,3,5,5,7,6,8,9,10,11,13,13,15,14]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm9[3,3,3,3]
+; AVX512DQ-FAST-NEXT: vpshufb {{.*#+}} ymm15 = ymm5[u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,24,25,24,25,22,23,22,23,u,u,u,u,u,u,u,u]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm15 = ymm15[2,2,2,2]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm31, %xmm5
+; AVX512DQ-FAST-NEXT: vpshuflw {{.*#+}} xmm14 = xmm5[0,2,3,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm14 = ymm14[0,0,2,1]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,0,1,3]
+; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm16, %xmm5
+; AVX512DQ-FAST-NEXT: vpshuflw {{.*#+}} xmm13 = xmm5[0,2,3,3,4,5,6,7]
; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm13 = ymm13[0,0,2,1]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm12 = ymm12[0,0,1,3]
-; AVX512DQ-FAST-NEXT: vmovdqa64 %xmm19, %xmm4
-; AVX512DQ-FAST-NEXT: vpshuflw {{.*#+}} xmm9 = xmm4[0,2,3,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,0,2,1]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,0,1,3]
-; AVX512DQ-FAST-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm4 = mem[2,1,3,2]
-; AVX512DQ-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm5 = mem[2,2,2,3]
-; AVX512DQ-FAST-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm6 = mem[0,2,2,3]
-; AVX512DQ-FAST-NEXT: vmovups %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vpermpd $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm6 = mem[2,1,3,3]
-; AVX512DQ-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpermpd $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm6 = mem[2,2,2,3]
-; AVX512DQ-FAST-NEXT: vmovups %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm6 = mem[0,2,2,3]
-; AVX512DQ-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpermpd $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm6 = mem[2,2,2,3]
-; AVX512DQ-FAST-NEXT: vmovups %zmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
-; AVX512DQ-FAST-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm6 = mem[0,2,2,3]
-; AVX512DQ-FAST-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX512DQ-FAST-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm31 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm31 = mem[2,1,3,2]
-; AVX512DQ-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm30 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm30 = mem[2,2,2,3]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm24 = ymm16[0,1,1,3]
-; AVX512DQ-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm19 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm19 = mem[0,0,1,1]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm18 = ymm11[0,0,1,1]
-; AVX512DQ-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm6 = mem[0,0,2,1]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm22 = ymm17[0,1,1,3]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm20 = ymm20[0,0,1,1]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm17 = ymm8[0,0,1,1]
-; AVX512DQ-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm16 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm16 = mem[0,0,2,1]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm28 = ymm29[2,2,2,3]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm26 = ymm26[0,2,2,3]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm27 = ymm27[2,1,3,2]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm25 = ymm25[2,2,2,3]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm3, %zmm15, %zmm3
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm1, %zmm2, %zmm1
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm2 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm3, %zmm2, %zmm1
-; AVX512DQ-FAST-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm3
-; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm3, %ymm14
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm13, %zmm3 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm12, %zmm12 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm29 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm3, %zmm29, %zmm12
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $226, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm12 # 64-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm9, %zmm9 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm9, %zmm29, %zmm0
-; AVX512DQ-FAST-NEXT: vpternlogq $226, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm0 # 64-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm3, %zmm3
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm4, %zmm7
-; AVX512DQ-FAST-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm3, %zmm7
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm11 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm11 # 64-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm4 # 64-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vextracti64x4 $1, %zmm1, %ymm1
-; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm3 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogd $226, 124(%r8){1to8}, %ymm3, %ymm1
-; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm10
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm10, %zmm0, %zmm1
-; AVX512DQ-FAST-NEXT: vshufi64x2 {{.*#+}} zmm1 = zmm14[0,1,2,3],zmm1[4,5,6,7]
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0]
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm1, %zmm3, %zmm8
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm14 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm14 # 64-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm19, %zmm24, %zmm1
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm18, %zmm3
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = [65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0]
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm1, %zmm5, %zmm3
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm20, %zmm22, %zmm1
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm16, %zmm17, %zmm6
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm1, %zmm5, %zmm6
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm3, %zmm1, %zmm20
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm22 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm6, %zmm1, %zmm22
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm30, %zmm31, %zmm3
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm5 = [0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535]
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm1, %zmm5, %zmm3
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm26, %zmm28, %zmm1
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm25, %zmm27, %zmm6
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm1, %zmm5, %zmm6
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0]
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm25 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm3, %zmm1, %zmm25
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm28 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm6, %zmm1, %zmm28
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm3 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm3, %zmm3 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm1, %zmm2, %zmm3
-; AVX512DQ-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm1 = mem[0,2,2,3]
-; AVX512DQ-FAST-NEXT: vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm5 = mem[0,1,1,3]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm1
-; AVX512DQ-FAST-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,0,1,3]
+; AVX512DQ-FAST-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm6 = mem[2,1,3,2]
+; AVX512DQ-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm7 = mem[2,2,2,3]
+; AVX512DQ-FAST-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm5 = mem[0,2,2,3]
+; AVX512DQ-FAST-NEXT: vmovups %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FAST-NEXT: vpermpd $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
; AVX512DQ-FAST-NEXT: # ymm5 = mem[2,1,3,3]
-; AVX512DQ-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm6 = mem[0,0,1,1]
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm5, %zmm5
-; AVX512DQ-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm6 = mem[0,0,2,1]
-; AVX512DQ-FAST-NEXT: vpermq $208, {{[-0-9]+}}(%r{{[sb]}}p), %ymm26 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm26 = mem[0,0,1,3]
-; AVX512DQ-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm27 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm27 = mem[0,0,1,1]
-; AVX512DQ-FAST-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # xmm9 = mem[0,2,3,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm9 = ymm9[0,0,2,1]
-; AVX512DQ-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm10 = mem[0,2,2,3]
-; AVX512DQ-FAST-NEXT: vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm24 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm24 = mem[0,1,1,3]
+; AVX512DQ-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpermpd $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm5 = mem[2,2,2,3]
+; AVX512DQ-FAST-NEXT: vmovups %zmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
+; AVX512DQ-FAST-NEXT: vpermpd $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm5 = mem[0,2,2,3]
+; AVX512DQ-FAST-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX512DQ-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm31 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm31 = mem[2,2,2,3]
+; AVX512DQ-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm30 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm30 = mem[0,2,2,3]
+; AVX512DQ-FAST-NEXT: vpermq $182, {{[-0-9]+}}(%r{{[sb]}}p), %ymm29 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm29 = mem[2,1,3,2]
+; AVX512DQ-FAST-NEXT: vpermq $234, {{[-0-9]+}}(%r{{[sb]}}p), %ymm28 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm28 = mem[2,2,2,3]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm19 = ymm21[0,1,1,3]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm18 = ymm18[0,0,1,1]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm17 = ymm8[0,0,1,1]
+; AVX512DQ-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm8 = mem[0,0,2,1]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm21 = ymm12[0,1,1,3]
+; AVX512DQ-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm12 = mem[0,0,1,1]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm16 = ymm11[0,0,1,1]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm11 = ymm27[0,0,2,1]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm27 = ymm26[2,2,2,3]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm26 = ymm25[0,2,2,3]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm25 = ymm24[2,1,3,2]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm24 = ymm20[2,2,2,3]
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm4, %zmm23, %zmm4
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm0, %zmm3, %zmm0
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm3 = [65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm4, %zmm3, %zmm0
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %ymm20 # 32-byte Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm20
+; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm20, %ymm15
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm14, %zmm4 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm1 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm23 = [65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm4, %zmm23, %zmm1
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535]
+; AVX512DQ-FAST-NEXT: vpternlogq $226, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm1 # 64-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm13, %zmm13 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm2, %zmm2 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm13, %zmm23, %zmm2
+; AVX512DQ-FAST-NEXT: vpternlogq $226, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm2 # 64-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm6, %zmm4, %zmm4
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm6, %zmm6
+; AVX512DQ-FAST-NEXT: vpternlogq $228, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm6
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = [0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm20 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm20 # 64-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm5 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $248, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm5 # 64-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vextracti64x4 $1, %zmm0, %ymm0
+; AVX512DQ-FAST-NEXT: vmovdqa {{.*#+}} ymm4 = [65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535]
+; AVX512DQ-FAST-NEXT: vpternlogd $226, 124(%r8){1to8}, %ymm4, %ymm0
+; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm9
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm0, %zmm0
+; AVX512DQ-FAST-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm15[0,1,2,3],zmm0[4,5,6,7]
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm4 = [65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0,65535,65535,65535,65535,65535,65535,0]
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm9 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm0, %zmm4, %zmm9
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm13 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $184, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm13 # 64-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm18, %zmm19, %zmm0
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm17, %zmm4
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm7 = [65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0]
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm0, %zmm7, %zmm4
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm12, %zmm21, %zmm0
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm16, %zmm8
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm0, %zmm7, %zmm8
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm21 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm4, %zmm0, %zmm21
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm17 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm8, %zmm0, %zmm17
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm30, %zmm31, %zmm0
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm28, %zmm29, %zmm4
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm7 = [0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535,65535,65535,0,0,65535,65535,65535]
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm0, %zmm7, %zmm4
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm26, %zmm27, %zmm0
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm24, %zmm25, %zmm8
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm0, %zmm7, %zmm8
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0]
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm26 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm4, %zmm0, %zmm26
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm29 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm8, %zmm0, %zmm29
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm4 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm4, %zmm4 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm0, %zmm3, %zmm4
+; AVX512DQ-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm0 = mem[0,2,2,3]
+; AVX512DQ-FAST-NEXT: vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm7 = mem[0,1,1,3]
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm7, %zmm0, %zmm0
+; AVX512DQ-FAST-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm7 = mem[2,1,3,3]
+; AVX512DQ-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm8 = mem[0,0,1,1]
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm8, %zmm7, %zmm7
+; AVX512DQ-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm8 = mem[0,0,2,1]
+; AVX512DQ-FAST-NEXT: vpermq $208, {{[-0-9]+}}(%r{{[sb]}}p), %ymm27 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm27 = mem[0,0,1,3]
+; AVX512DQ-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm24 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm24 = mem[0,0,1,1]
+; AVX512DQ-FAST-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # xmm11 = mem[0,2,3,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,0,2,1]
+; AVX512DQ-FAST-NEXT: vpermq $232, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm12 = mem[0,2,2,3]
+; AVX512DQ-FAST-NEXT: vpermq $212, {{[-0-9]+}}(%r{{[sb]}}p), %ymm25 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm25 = mem[0,1,1,3]
; AVX512DQ-FAST-NEXT: vpermq $246, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Folded Reload
; AVX512DQ-FAST-NEXT: # ymm15 = mem[2,1,3,3]
; AVX512DQ-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm16 # 32-byte Folded Reload
; AVX512DQ-FAST-NEXT: # ymm16 = mem[0,0,1,1]
-; AVX512DQ-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm17 # 32-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # ymm17 = mem[0,0,2,1]
+; AVX512DQ-FAST-NEXT: vpermq $96, {{[-0-9]+}}(%r{{[sb]}}p), %ymm28 # 32-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # ymm28 = mem[0,0,2,1]
; AVX512DQ-FAST-NEXT: vpermq $208, {{[-0-9]+}}(%r{{[sb]}}p), %ymm18 # 32-byte Folded Reload
; AVX512DQ-FAST-NEXT: # ymm18 = mem[0,0,1,3]
; AVX512DQ-FAST-NEXT: vpermq $80, {{[-0-9]+}}(%r{{[sb]}}p), %ymm19 # 32-byte Folded Reload
; AVX512DQ-FAST-NEXT: # ymm19 = mem[0,0,1,1]
-; AVX512DQ-FAST-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX512DQ-FAST-NEXT: # xmm13 = mem[0,2,3,3,4,5,6,7]
-; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm13 = ymm13[0,0,2,1]
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm1, %zmm2, %zmm5
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm24, %zmm10, %zmm1
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm16, %zmm15, %zmm10
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm1, %zmm2, %zmm10
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535]
+; AVX512DQ-FAST-NEXT: vpshuflw $248, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; AVX512DQ-FAST-NEXT: # xmm14 = mem[0,2,3,3,4,5,6,7]
+; AVX512DQ-FAST-NEXT: vpermq {{.*#+}} ymm14 = ymm14[0,0,2,1]
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm0, %zmm3, %zmm7
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm25, %zmm12, %zmm0
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm16, %zmm15, %zmm12
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm0, %zmm3, %zmm12
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535]
; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm15 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm5, %zmm1, %zmm15
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm10, %zmm1, %zmm21
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm26, %zmm6, %zmm1
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm9, %zmm27, %zmm2
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm1, %zmm29, %zmm2
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm18, %zmm17, %zmm1
-; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm13, %zmm19, %zmm5
-; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm1, %zmm29, %zmm5
-; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm1 = [65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0]
-; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm6 # 64-byte Reload
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm2, %zmm1, %zmm6
-; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm5, %zmm1, %zmm23
-; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm7, %zmm4
-; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm3, %zmm14
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm7, %zmm0, %zmm15
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm12, %zmm0, %zmm22
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm27, %zmm8, %zmm0
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm11, %zmm24, %zmm3
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm0, %zmm23, %zmm3
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm18, %zmm28, %zmm0
+; AVX512DQ-FAST-NEXT: vinserti64x4 $1, %ymm14, %zmm19, %zmm7
+; AVX512DQ-FAST-NEXT: vpternlogq $226, %zmm0, %zmm23, %zmm7
+; AVX512DQ-FAST-NEXT: vmovdqa64 {{.*#+}} zmm0 = [65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0,0,65535,65535,65535,65535,0,0]
+; AVX512DQ-FAST-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm8 # 64-byte Reload
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm3, %zmm0, %zmm8
+; AVX512DQ-FAST-NEXT: vpternlogq $184, %zmm7, %zmm0, %zmm10
+; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm6, %zmm5
+; AVX512DQ-FAST-NEXT: vpternlogq $216, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm4, %zmm13
; AVX512DQ-FAST-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm4, 320(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm23, 256(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm21, 192(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm28, 128(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm0, 64(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm22, (%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm20, 448(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm6, 704(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm5, 320(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm10, 256(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm22, 192(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm29, 128(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm2, 64(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm17, (%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm21, 448(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm8, 704(%rax)
; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm15, 640(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm25, 576(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm12, 512(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm14, 384(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm11, 768(%rax)
-; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm8, 832(%rax)
-; AVX512DQ-FAST-NEXT: addq $2264, %rsp # imm = 0x8D8
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm26, 576(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm1, 512(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm13, 384(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm20, 768(%rax)
+; AVX512DQ-FAST-NEXT: vmovdqa64 %zmm9, 832(%rax)
+; AVX512DQ-FAST-NEXT: addq $2200, %rsp # imm = 0x898
; AVX512DQ-FAST-NEXT: vzeroupper
; AVX512DQ-FAST-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
index 1180f88..8d92086 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
@@ -58,24 +58,24 @@
; AVX1-ONLY-NEXT: vmovq {{.*#+}} xmm4 = mem[0],zero
; AVX1-ONLY-NEXT: vmovq {{.*#+}} xmm5 = mem[0],zero
; AVX1-ONLY-NEXT: vpunpcklqdq {{.*#+}} xmm4 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT: vmovsd {{.*#+}} xmm6 = mem[0],zero
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovsd {{.*#+}} xmm5 = mem[0],zero
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3,0,1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm1[3,0],ymm0[1,0],ymm1[7,4],ymm0[5,4]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[0,2],ymm0[2,1],ymm6[4,6],ymm0[6,5]
-; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm5 = xmm2[12,13,14,15],xmm5[0,1,2,3,4,5,6,7,8,9,10,11]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[0,2],xmm4[3,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1,2],xmm6[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm1[3,0],ymm0[1,0],ymm1[7,4],ymm0[5,4]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm5[0,2],ymm0[2,1],ymm5[4,6],ymm0[6,5]
+; AVX1-ONLY-NEXT: vpalignr {{.*#+}} xmm6 = xmm2[12,13,14,15],xmm4[0,1,2,3,4,5,6,7,8,9,10,11]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,2],xmm4[3,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1,2],xmm5[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[0,2],ymm0[0,3],ymm1[4,6],ymm0[4,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm1
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[u,u,0,2,u,u,u,5]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2],ymm1[2,3],ymm2[4,6],ymm1[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rax)
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm0
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm0
; AVX1-ONLY-NEXT: vmovlps %xmm0, 48(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm5, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm6, 32(%rax)
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -3435,20 +3435,20 @@
;
; AVX1-ONLY-LABEL: store_i32_stride7_vf32:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $1656, %rsp # imm = 0x678
+; AVX1-ONLY-NEXT: subq $1624, %rsp # imm = 0x658
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %ymm0
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %ymm4
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rax), %ymm3
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm4[1],ymm2[1],ymm4[3],ymm2[3]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm4[2],ymm0[2],ymm4[3],ymm0[3],ymm4[6],ymm0[6],ymm4[7],ymm0[7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,0],ymm2[4,5],ymm1[6,4]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm1[2,3],ymm0[2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
@@ -3501,14 +3501,14 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm1[1,1],ymm0[5,5],ymm1[5,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rcx), %ymm1
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm1[1,1],ymm2[5,5],ymm1[5,5]
+; AVX1-ONLY-NEXT: vmovaps (%rcx), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm2[1,1],ymm1[5,5],ymm2[5,5]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%r9), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
@@ -3521,18 +3521,18 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm5
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm5[0],xmm1[0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm8
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm8[0],xmm1[0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1]
-; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm7
+; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm6
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %xmm2
; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm3
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm3, %xmm6
+; AVX1-ONLY-NEXT: vmovaps %xmm3, %xmm9
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm8
+; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm7
; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
@@ -3551,31 +3551,30 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm4[1,1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm5[1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm5[1,1],xmm1[0,2]
+; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm6[1],xmm8[1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm8[1,1],xmm1[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm6[1],xmm8[1],zero
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm9[1],xmm7[1],zero
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm0
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm8
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %ymm0
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,1],ymm0[1,1],ymm1[5,5],ymm0[5,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm8[1,1],ymm0[5,5],ymm8[5,5]
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %ymm11
-; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm11[1,1],ymm1[1,1],ymm11[5,5],ymm1[5,5]
-; AVX1-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%rcx), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm2[1,1],ymm1[5,5],ymm2[5,5]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
; AVX1-ONLY-NEXT: vmovaps 32(%r8), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r9), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX1-ONLY-NEXT: vmovaps 32(%r9), %ymm11
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm11[0],ymm2[0],ymm11[2],ymm2[2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,1],ymm1[6,4],ymm2[6,5]
; AVX1-ONLY-NEXT: vmovaps 32(%rax), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -3585,18 +3584,19 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm6
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm6[0],xmm1[0]
+; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm7[0],xmm1[0]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1]
-; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm7
+; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm6
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm5
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm3, %xmm8
+; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm3
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX1-ONLY-NEXT: vmovaps %xmm3, %xmm9
; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm2, %xmm10
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,1,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 64(%r9), %xmm3
@@ -3614,31 +3614,31 @@
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm4[1,1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm6[1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm6[1,1],xmm1[0,2]
+; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm6[1],xmm7[1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm7[1,1],xmm1[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm5[1],xmm8[1],zero
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm9[1],xmm10[1],zero
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm13
-; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %ymm14
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[1,1],ymm13[1,1],ymm14[5,5],ymm13[5,5]
-; AVX1-ONLY-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm15
+; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %ymm0
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm15[1,1],ymm0[5,5],ymm15[5,5]
+; AVX1-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %ymm9
-; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %ymm8
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm9[1,1],ymm8[1,1],ymm9[5,5],ymm8[5,5]
-; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
-; AVX1-ONLY-NEXT: vmovaps 64(%r8), %ymm10
-; AVX1-ONLY-NEXT: vmovaps 64(%r9), %ymm12
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm12[0],ymm10[0],ymm12[2],ymm10[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm10[2,1],ymm1[6,4],ymm10[6,5]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %ymm10
+; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %ymm9
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm10[1,1],ymm9[1,1],ymm10[5,5],ymm9[5,5]
; AVX1-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
+; AVX1-ONLY-NEXT: vmovaps 64(%r8), %ymm12
+; AVX1-ONLY-NEXT: vmovaps 64(%r9), %ymm13
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm13[0],ymm12[0],ymm13[2],ymm12[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm12[2,1],ymm1[6,4],ymm12[6,5]
+; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rax), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm1[2,3]
@@ -3646,39 +3646,38 @@
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm2[0],xmm7[0]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[2,1]
-; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm0 = xmm7[0],xmm6[0]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm6[2,1]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 96(%rcx), %xmm5
; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm3
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm15 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; AVX1-ONLY-NEXT: vmovaps %xmm4, %xmm6
-; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm15[0,1,0,1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1],ymm15[2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 96(%r9), %xmm5
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm14 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps %xmm3, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm14[0,1,0,1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1],ymm14[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 96(%r9), %xmm0
+; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%r8), %xmm4
; AVX1-ONLY-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm15 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rax), %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm15[4,5,6],ymm1[7]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} xmm14 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rax), %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm14[0],ymm2[0],ymm14[2],ymm2[2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm5[1,1],xmm4[1,1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm0[1,1],xmm4[1,1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm2[1]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm2[1,1],xmm1[0,2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm2[5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm6[1],xmm7[1]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm7[1,1],xmm1[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm15 = zero,xmm3[1],xmm6[1],zero
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm3[1],xmm5[1],zero
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -3690,49 +3689,49 @@
; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm3[2],ymm4[3],ymm3[3],ymm4[6],ymm3[6],ymm4[7],ymm3[7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm1[0,2],ymm2[5,5],ymm1[4,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 16(%rax), %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 16(%rax), %xmm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm11[2],ymm5[2],ymm11[3],ymm5[3],ymm11[6],ymm5[6],ymm11[7],ymm5[7]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm5[2],ymm8[3],ymm5[3],ymm8[6],ymm5[6],ymm8[7],ymm5[7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm11[1],ymm1[3],ymm11[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm11[1,1],ymm1[0,2],ymm11[5,5],ymm1[4,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 48(%rax), %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 48(%rax), %xmm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm14[1],ymm13[1],ymm14[3],ymm13[3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm15[1],ymm0[3],ymm15[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm9[2],ymm8[2],ymm9[3],ymm8[3],ymm9[6],ymm8[6],ymm9[7],ymm8[7]
+; AVX1-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm10[2],ymm9[2],ymm10[3],ymm9[3],ymm10[6],ymm9[6],ymm10[7],ymm9[7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm10[1],ymm12[1],ymm10[3],ymm12[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[1,1],ymm1[0,2],ymm12[5,5],ymm1[4,6]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm12[1],ymm13[1],ymm12[3],ymm13[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm13[1,1],ymm1[0,2],ymm13[5,5],ymm1[4,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 80(%rax), %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 80(%rax), %xmm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[3,3],ymm7[3,3],ymm14[7,7],ymm7[7,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm7[3,3],ymm2[3,3],ymm7[7,7],ymm2[7,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,3],ymm2[3,3],ymm6[7,7],ymm2[7,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm6[3,3],ymm14[3,3],ymm6[7,7],ymm14[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vbroadcastss 124(%r8), %ymm1
@@ -3742,9 +3741,9 @@
; AVX1-ONLY-NEXT: vbroadcastsd 120(%rax), %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm14[0],ymm7[2],ymm14[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm14[3,1],ymm0[0,2],ymm14[7,5],ymm0[4,6]
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm6[0],ymm2[0],ymm6[1],ymm2[1],ymm6[4],ymm2[4],ymm6[5],ymm2[5]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm2[0],ymm7[0],ymm2[2],ymm7[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm7[3,1],ymm0[0,2],ymm7[7,5],ymm0[4,6]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm6[0],ymm14[0],ymm6[1],ymm14[1],ymm6[4],ymm14[4],ymm6[5],ymm14[5]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vbroadcastss 108(%r8), %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7]
@@ -3753,8 +3752,8 @@
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[0,1,2],mem[3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm7[1,1],ymm14[1,1],ymm7[5,5],ymm14[5,5]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm6[1,1],ymm2[5,5],ymm6[5,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm2[1,1],ymm7[1,1],ymm2[5,5],ymm7[5,5]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm14[1,1],ymm6[1,1],ymm14[5,5],ymm6[5,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
; AVX1-ONLY-NEXT: vbroadcastsd 112(%r8), %ymm1
@@ -3777,8 +3776,8 @@
; AVX1-ONLY-NEXT: # xmm1 = mem[2,2,2,2]
; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm1 = mem[0,1,2],xmm1[3]
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%rax), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm15[4,5,6,7]
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%rax), %ymm14
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm3[3,3],ymm4[3,3],ymm3[7,7],ymm4[7,7]
@@ -3788,7 +3787,7 @@
; AVX1-ONLY-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm1 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -3796,63 +3795,63 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,2,3,1,4,6,7,5]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm0[3,3],mem[3,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm15 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm15 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm15, %ymm13
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2,3,4],ymm1[5,6],ymm13[7]
-; AVX1-ONLY-NEXT: vpermilps $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[2,2,2,2]
-; AVX1-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm13 = mem[0,1,2],xmm13[3]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%rax), %ymm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm15[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm13[2,3,4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm5[3,3],ymm14[3,3],ymm5[7,7],ymm14[7,7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm13[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm0 = xmm0[3,3],mem[3,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm1 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm15[2,2,2,2]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm14[0,1,2],xmm1[3]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%rax), %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm13[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm5[3,3],ymm8[3,3],ymm5[7,7],ymm8[7,7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm12[3,3],ymm7[3,3],ymm12[7,7],ymm7[7,7]
-; AVX1-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm15[0,1,2],ymm13[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm0[3,3],ymm11[3,3],ymm0[7,7],ymm11[7,7]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm0[2,3],ymm11[1,2],ymm0[6,7],ymm11[5,6]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm7[3,3],ymm0[3,3],ymm7[7,7],ymm0[7,7]
+; AVX1-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1,2],ymm1[3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm2[3,3],ymm11[3,3],ymm2[7,7],ymm11[7,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm2[2,3],ymm11[1,2],ymm2[6,7],ymm11[5,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm11[2,3,2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm11 = ymm11[0,2,3,1,4,6,7,5]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0],ymm13[1,2,3,4],ymm11[5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm9 = xmm0[3,3],mem[3,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm10 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm10 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm10, %ymm8
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0],ymm1[1,2,3,4],ymm11[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm9 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm9 = xmm1[3,3],mem[3,3]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm10 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm10 = xmm1[2],mem[2],xmm1[3],mem[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm10, %ymm8
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm9[5,6],ymm8[7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm12[2,2,2,2]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm13[2,2,2,2]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm15[0,1,2],xmm9[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm13[0,1,2],xmm9[3]
; AVX1-ONLY-NEXT: vbroadcastsd 72(%rax), %ymm10
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm8[0,1],ymm9[2,3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm9[2,3,4],ymm8[5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm6[3,3],ymm8[3,3],ymm6[7,7],ymm8[7,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm8[3,3],ymm11[3,3],ymm8[7,7],ymm11[7,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm9[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm1[3,3],ymm0[3,3],ymm1[7,7],ymm0[7,7]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm6[3,3],ymm1[3,3],ymm6[7,7],ymm1[7,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2],ymm9[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -3866,7 +3865,7 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = xmm2[3,3],mem[3,3]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm2 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = xmm2[2],mem[2],xmm2[3],mem[3]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm5, %ymm2
@@ -3893,22 +3892,21 @@
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm5 = xmm5[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm5[1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm12[0],ymm7[0],ymm12[1],ymm7[1],ymm12[4],ymm7[4],ymm12[5],ymm7[5]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm14[0],ymm9[0],ymm14[2],ymm9[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm9[3,1],ymm7[0,2],ymm9[7,5],ymm7[4,6]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm7[0],ymm0[0],ymm7[1],ymm0[1],ymm7[4],ymm0[4],ymm7[5],ymm0[5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm7 = ymm7[0],ymm0[0],ymm7[2],ymm0[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm0[3,1],ymm7[0,2],ymm0[7,5],ymm7[4,6]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm5[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm7 = xmm7[3,3],mem[3,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm14[3,3],xmm15[3,3]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = xmm7[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm7[1,2,3],ymm5[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm7 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm8[0],ymm6[0],ymm8[2],ymm6[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[3,1],ymm9[0,2],ymm6[7,5],ymm9[4,6]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm7 = ymm6[0],ymm1[0],ymm6[1],ymm1[1],ymm6[4],ymm1[4],ymm6[5],ymm1[5]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm11[0],ymm8[0],ymm11[2],ymm8[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm8[3,1],ymm9[0,2],ymm8[7,5],ymm9[4,6]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm15[3,3],xmm13[3,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm13[3,3],xmm12[3,3]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm7 = xmm7[0,1,2],mem[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4,5,6,7]
@@ -3920,14 +3918,15 @@
; AVX1-ONLY-NEXT: vmovaps %ymm4, 640(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 608(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm11, 512(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm0, 512(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 416(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 384(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 288(%rax)
-; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 192(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 160(%rax)
@@ -3963,7 +3962,7 @@
; AVX1-ONLY-NEXT: vmovaps %ymm0, 864(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 832(%rax)
-; AVX1-ONLY-NEXT: addq $1656, %rsp # imm = 0x678
+; AVX1-ONLY-NEXT: addq $1624, %rsp # imm = 0x658
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -7221,7 +7220,7 @@
;
; AVX1-ONLY-LABEL: store_i32_stride7_vf64:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: subq $3416, %rsp # imm = 0xD58
+; AVX1-ONLY-NEXT: subq $3432, %rsp # imm = 0xD68
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %ymm5
; AVX1-ONLY-NEXT: vmovaps 224(%rsi), %ymm2
@@ -7289,17 +7288,16 @@
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm1[1,1],ymm0[5,5],ymm1[5,5]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %ymm12
-; AVX1-ONLY-NEXT: vmovaps (%rcx), %ymm1
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[1,1],ymm1[1,1],ymm12[5,5],ymm1[5,5]
-; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%rcx), %ymm2
+; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm2[1,1],ymm1[5,5],ymm2[5,5]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%r9), %ymm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX1-ONLY-NEXT: vmovaps (%r9), %ymm12
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm12[0],ymm2[0],ymm12[2],ymm2[2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,0],ymm2[2,1],ymm1[6,4],ymm2[6,5]
; AVX1-ONLY-NEXT: vmovaps (%rax), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -7683,13 +7681,14 @@
; AVX1-ONLY-NEXT: vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4]
-; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm2[1,1],ymm1[0,2],ymm2[5,5],ymm1[4,6]
+; AVX1-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm12[1],ymm1[3],ymm12[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm12[1,1],ymm1[0,2],ymm12[5,5],ymm1[4,6]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
; AVX1-ONLY-NEXT: vmovaps 16(%rax), %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
@@ -8114,67 +8113,67 @@
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm7[0],ymm0[1],ymm7[1],ymm0[4],ymm7[4],ymm0[5],ymm7[5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm7[0],ymm2[2],ymm7[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm7[3,1],ymm2[0,2],ymm7[7,5],ymm2[4,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm7[0],ymm2[2],ymm7[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[3,1],ymm3[0,2],ymm7[7,5],ymm3[4,6]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[3,3],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm2[3,3],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm7[0],ymm2[2],ymm7[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm7[3,1],ymm2[0,2],ymm7[7,5],ymm2[4,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $255, (%rsp), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[3,3],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0],ymm2[1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm2[0],ymm7[0],ymm2[2],ymm7[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm7[3,1],ymm3[0,2],ymm7[7,5],ymm3[4,6]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vshufps $255, (%rsp), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[3,3],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0],ymm3[1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[1],ymm6[1],ymm0[4],ymm6[4],ymm0[5],ymm6[5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm6[0],ymm2[2],ymm6[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm6[3,1],ymm2[0,2],ymm6[7,5],ymm2[4,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[3,3],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm0[0],ymm2[1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm6[0],ymm3[2],ymm6[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm6[3,1],ymm3[0,2],ymm6[7,5],ymm3[4,6]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[3,3],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm0[0],ymm3[1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[1],ymm4[1],ymm0[4],ymm4[4],ymm0[5],ymm4[5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[2],ymm4[2]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm4[3,1],ymm2[0,2],ymm4[7,5],ymm2[4,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[3,3],mem[3,3]
-; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm2 = xmm2[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm12[0],ymm1[0],ymm12[1],ymm1[1],ymm12[4],ymm1[4],ymm12[5],ymm1[5]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[3,1],ymm3[0,2],ymm4[7,5],ymm3[4,6]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[3,3],mem[3,3]
+; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm3 = xmm3[0,1,2],mem[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm3 = ymm12[0],ymm1[0],ymm12[1],ymm1[1],ymm12[4],ymm1[4],ymm12[5],ymm1[5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm12 = ymm4[0],ymm14[0],ymm4[2],ymm14[2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm14[3,1],ymm12[0,2],ymm14[7,5],ymm12[4,6]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm12[0,1,2,3],ymm2[4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm12[0,1,2,3],ymm3[4,5],ymm12[6,7]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX1-ONLY-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm12 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm12 = xmm1[3,3],mem[3,3]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm12 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm12 = xmm12[0,1,2],mem[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm12[1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm12[1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm12 = ymm9[0],ymm10[0],ymm9[1],ymm10[1],ymm9[4],ymm10[4],ymm9[5],ymm10[5]
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm11[0],ymm1[0],ymm11[2],ymm1[2]
@@ -8199,10 +8198,10 @@
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: vmovaps %ymm9, 1440(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm10, 1216(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 992(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 992(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm0, 768(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm7, 544(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 320(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 320(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 96(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -8303,37 +8302,37 @@
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1632(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1600(%rax)
-; AVX1-ONLY-NEXT: addq $3416, %rsp # imm = 0xD58
+; AVX1-ONLY-NEXT: addq $3432, %rsp # imm = 0xD68
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
; AVX2-SLOW-LABEL: store_i32_stride7_vf64:
; AVX2-SLOW: # %bb.0:
-; AVX2-SLOW-NEXT: subq $3000, %rsp # imm = 0xBB8
+; AVX2-SLOW-NEXT: subq $2968, %rsp # imm = 0xB98
; AVX2-SLOW-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-SLOW-NEXT: vmovaps (%rax), %xmm0
; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vmovaps (%r8), %xmm2
-; AVX2-SLOW-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 32(%r8), %xmm4
-; AVX2-SLOW-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vmovaps (%r8), %xmm5
+; AVX2-SLOW-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 32(%r8), %xmm3
+; AVX2-SLOW-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vmovaps (%r9), %xmm1
; AVX2-SLOW-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 32(%r9), %xmm7
-; AVX2-SLOW-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 32(%r9), %xmm4
+; AVX2-SLOW-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
; AVX2-SLOW-NEXT: vbroadcastsd %xmm1, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovaps (%rcx), %xmm9
-; AVX2-SLOW-NEXT: vmovaps 32(%rcx), %xmm3
-; AVX2-SLOW-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 32(%rcx), %xmm10
+; AVX2-SLOW-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vmovaps (%rdx), %xmm8
; AVX2-SLOW-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm8[1],xmm9[1],zero
; AVX2-SLOW-NEXT: vmovaps (%rdi), %xmm6
-; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %xmm10
-; AVX2-SLOW-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-SLOW-NEXT: vmovaps 32(%rdi), %xmm7
+; AVX2-SLOW-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vmovaps (%rsi), %xmm5
; AVX2-SLOW-NEXT: vmovaps 32(%rsi), %xmm11
; AVX2-SLOW-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8346,15 +8345,15 @@
; AVX2-SLOW-NEXT: vmovaps 32(%rax), %xmm0
; AVX2-SLOW-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm7[1,1,1,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
; AVX2-SLOW-NEXT: vbroadcastsd %xmm1, %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm11[1,1,2,2]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm10[2],xmm1[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2],xmm1[3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
; AVX2-SLOW-NEXT: vmovaps 32(%rdx), %xmm7
-; AVX2-SLOW-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm7[1],xmm3[1],zero
+; AVX2-SLOW-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm7[1],xmm10[1],zero
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8533,11 +8532,11 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm1
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 64(%rsi), %ymm0
+; AVX2-SLOW-NEXT: vmovaps 64(%rdi), %ymm0
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
+; AVX2-SLOW-NEXT: vmovaps 64(%rsi), %ymm1
+; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
; AVX2-SLOW-NEXT: vmovaps 64(%rdx), %ymm1
; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -8664,21 +8663,22 @@
; AVX2-SLOW-NEXT: vmovaps 224(%r9), %xmm3
; AVX2-SLOW-NEXT: vbroadcastss %xmm3, %ymm14
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5],ymm15[6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 224(%rax), %ymm15
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm14[0,1,2,3,4,5],ymm15[6],ymm14[7]
+; AVX2-SLOW-NEXT: vmovaps 224(%rax), %xmm15
+; AVX2-SLOW-NEXT: vbroadcastss %xmm15, %ymm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm14[0,1,2,3,4,5],ymm13[6],ymm14[7]
; AVX2-SLOW-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm14 = xmm4[1,1,2,2]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0,1],xmm2[2],xmm14[3]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[0,1,2,1]
-; AVX2-SLOW-NEXT: vinsertps {{.*#+}} xmm15 = zero,xmm1[1],xmm0[1],zero
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0],ymm15[1,2],ymm14[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 228(%r8), %ymm13
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm14 = xmm3[1,1,1,1]
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5],ymm13[6,7]
-; AVX2-SLOW-NEXT: vinsertf128 $1, 224(%rax), %ymm15, %ymm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm13 = xmm4[1,1,2,2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm2[2],xmm13[3]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
+; AVX2-SLOW-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm1[1],xmm0[1],zero
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2],ymm13[3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 228(%r8), %ymm12
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm13 = xmm3[1,1,1,1]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm15, %ymm14, %ymm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3],xmm4[3,3]
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
@@ -8694,75 +8694,74 @@
; AVX2-SLOW-NEXT: vbroadcastss 232(%rax), %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 224(%rdi), %ymm12
-; AVX2-SLOW-NEXT: vmovaps 224(%rsi), %ymm10
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm13 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT: vmovaps 224(%rdx), %ymm0
+; AVX2-SLOW-NEXT: vmovaps 224(%rdi), %ymm4
+; AVX2-SLOW-NEXT: vmovaps 224(%rsi), %ymm2
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm12 = ymm0[2,2,2,2]
+; AVX2-SLOW-NEXT: vmovaps 224(%rdx), %ymm10
+; AVX2-SLOW-NEXT: vmovaps 224(%rcx), %ymm1
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm13 = ymm10[1,1],ymm1[1,1],ymm10[5,5],ymm1[5,5]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm13[5,6],ymm12[7]
+; AVX2-SLOW-NEXT: vbroadcastsd 240(%r8), %ymm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0],ymm12[1,2,3,4,5,6],ymm13[7]
+; AVX2-SLOW-NEXT: vbroadcastss 240(%r9), %xmm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1],ymm12[2,3,4,5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 240(%rax), %ymm13
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1],ymm13[2],ymm12[3,4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps 224(%rcx), %ymm2
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm14 = ymm0[1,1],ymm2[1,1],ymm0[5,5],ymm2[5,5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6],ymm13[7]
-; AVX2-SLOW-NEXT: vbroadcastsd 240(%r8), %ymm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1,2,3,4,5,6],ymm14[7]
-; AVX2-SLOW-NEXT: vbroadcastss 240(%r9), %xmm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4,5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 240(%rax), %ymm14
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm14[2],ymm13[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vbroadcastss %xmm9, %xmm13
-; AVX2-SLOW-NEXT: vbroadcastss %xmm8, %xmm14
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm14 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm14 = xmm14[0,1,2,2]
-; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[0,1,2,1]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1],ymm13[2,3],ymm14[4,5,6,7]
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-SLOW-NEXT: vbroadcastss %xmm9, %xmm12
+; AVX2-SLOW-NEXT: vbroadcastss %xmm8, %xmm13
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm13 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3],ymm13[4,5,6,7]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm14 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-SLOW-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 16-byte Folded Reload
-; AVX2-SLOW-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm14[0],ymm15[0],ymm14[2],ymm15[2]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6],ymm1[7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm13 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX2-SLOW-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 16-byte Folded Reload
+; AVX2-SLOW-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm13[0],ymm14[0],ymm13[2],ymm14[2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6],ymm12[7]
+; AVX2-SLOW-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm6[3,3],xmm5[3,3]
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm4[2,2,2,2]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 8(%rax), %ymm8
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-SLOW-NEXT: vbroadcastss %xmm1, %xmm5
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT: vbroadcastss %xmm12, %xmm5
; AVX2-SLOW-NEXT: vbroadcastss %xmm7, %xmm6
; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm4[0],xmm14[0],xmm4[1],xmm14[1]
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm13[0],xmm9[1],xmm13[1]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm15[0],xmm9[1],xmm15[1]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-SLOW-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm8[0],ymm6[2],ymm8[2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm4[3,3],xmm14[3,3]
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm13[3,3],xmm14[3,3]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm12[2],xmm7[3],xmm12[3]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm13[2,2,2,2]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm15[2,2,2,2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 40(%rax), %ymm7
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -8772,40 +8771,40 @@
; AVX2-SLOW-NEXT: vbroadcastss %xmm13, %xmm5
; AVX2-SLOW-NEXT: vbroadcastss %xmm11, %xmm6
; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-SLOW-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 72(%rax), %ymm7
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-SLOW-NEXT: vbroadcastss %xmm9, %xmm5
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-SLOW-NEXT: vbroadcastss %xmm13, %xmm6
+; AVX2-SLOW-NEXT: vbroadcastss %xmm13, %xmm5
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX2-SLOW-NEXT: vbroadcastss %xmm14, %xmm6
; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
@@ -8817,8 +8816,8 @@
; AVX2-SLOW-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm4[3,3],xmm1[3,3]
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm9[2],xmm13[3],xmm9[3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
@@ -8829,14 +8828,14 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-SLOW-NEXT: vbroadcastss %xmm9, %xmm5
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT: vbroadcastss %xmm12, %xmm5
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-SLOW-NEXT: vbroadcastss %xmm13, %xmm6
; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
@@ -8846,10 +8845,10 @@
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-SLOW-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm8[3,3],xmm4[3,3]
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm9[2],xmm13[3],xmm9[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
@@ -8860,14 +8859,14 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-SLOW-NEXT: vbroadcastss %xmm9, %xmm5
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT: vbroadcastss %xmm12, %xmm5
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-SLOW-NEXT: vbroadcastss %xmm13, %xmm6
; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
@@ -8877,10 +8876,10 @@
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-SLOW-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
-; AVX2-SLOW-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm8[3,3],xmm4[3,3]
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm9[2],xmm13[3],xmm9[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
+; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
@@ -8891,40 +8890,40 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-SLOW-NEXT: vbroadcastss %xmm9, %xmm5
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX2-SLOW-NEXT: vbroadcastss %xmm11, %xmm6
+; AVX2-SLOW-NEXT: vbroadcastss %xmm11, %xmm5
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX2-SLOW-NEXT: vbroadcastss %xmm13, %xmm6
; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-SLOW-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
; AVX2-SLOW-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm8[3,3],xmm4[3,3]
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm11[2],xmm9[2],xmm11[3],xmm9[3]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm11[2],xmm13[3],xmm11[3]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm6 = xmm12[0,1,2],xmm6[3]
; AVX2-SLOW-NEXT: vbroadcastsd 200(%rax), %ymm7
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm5 = ymm4[2],mem[2],ymm4[3],mem[3],ymm4[6],mem[6],ymm4[7],mem[7]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm5 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
@@ -8939,18 +8938,17 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0],ymm5[1,2,3,4,5,6],ymm6[7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vbroadcastss 240(%rdx), %ymm5
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm2[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm6 = ymm1[3,1,2,0,7,5,6,4]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6],ymm6[7]
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} ymm6 = ymm12[0],ymm10[0],ymm12[1],ymm10[1],ymm12[4],ymm10[4],ymm12[5],ymm10[5]
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} ymm6 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
; AVX2-SLOW-NEXT: vbroadcastss 236(%r8), %ymm6
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2,3,4,5,6,7]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm3[2,2,3,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3],ymm5[4,5,6,7]
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm5 = ymm12[2],ymm10[2],ymm12[3],ymm10[3],ymm12[6],ymm10[6],ymm12[7],ymm10[7]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm5 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm6 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm6 = ymm10[2],ymm1[2],ymm10[3],ymm1[3],ymm10[6],ymm1[6],ymm10[7],ymm1[7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
@@ -8963,8 +8961,8 @@
; AVX2-SLOW-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm7[2,3],ymm6[2,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0],ymm3[1],ymm5[2,3,4],ymm3[5],ymm5[6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm0 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm10[2],ymm12[2],ymm10[3],ymm12[3],ymm10[6],ymm12[6],ymm10[7],ymm12[7]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm10[2],ymm1[3],ymm10[3],ymm1[6],ymm10[6],ymm1[7],ymm10[7]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm4[2],ymm2[3],ymm4[3],ymm2[6],ymm4[6],ymm2[7],ymm4[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -8975,17 +8973,17 @@
; AVX2-SLOW-NEXT: vbroadcastsd 248(%rax), %ymm1
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4],ymm5[5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm8[1,1],ymm3[1,1],ymm8[5,5],ymm3[5,5]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,0,0,0,4,4,4,4]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,0,0,0,4,4,4,4]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,1,0,1,4,5,4,5]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7]
@@ -8994,8 +8992,8 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm15[1,1,1,1,5,5,5,5]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
@@ -9013,14 +9011,14 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-SLOW-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = mem[1,1,1,1,5,5,5,5]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
; AVX2-SLOW-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm15[1,1],ymm1[5,5],ymm15[5,5]
+; AVX2-SLOW-NEXT: vshufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm1 = ymm1[1,1],mem[1,1],ymm1[5,5],mem[5,5]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
; AVX2-SLOW-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
@@ -9090,7 +9088,7 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm4[1,1],ymm0[5,5],ymm4[5,5]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm12[1,1],ymm0[5,5],ymm12[5,5]
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm13[1,1,1,1,5,5,5,5]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
@@ -9108,7 +9106,7 @@
; AVX2-SLOW-NEXT: vbroadcastss 16(%rdx), %ymm0
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm3[3,1,2,0,7,5,6,4]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} ymm1 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} ymm1 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
@@ -9118,11 +9116,11 @@
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm0 = ymm3[2],ymm8[2],ymm3[3],ymm8[3],ymm3[6],ymm8[6],ymm3[7],ymm8[7]
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm5[2],ymm6[3],ymm5[3],ymm6[6],ymm5[6],ymm6[7],ymm5[7]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,3],ymm12[3,3],ymm9[7,7],ymm12[7,7]
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,3],ymm14[3,3],ymm9[7,7],ymm14[7,7]
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
@@ -9130,17 +9128,17 @@
; AVX2-SLOW-NEXT: vbroadcastss 48(%rdx), %ymm0
; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm7[3,1,2,0,7,5,6,4]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} ymm1 = ymm10[0],ymm14[0],ymm10[1],ymm14[1],ymm10[4],ymm14[4],ymm10[5],ymm14[5]
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} ymm1 = ymm10[0],ymm15[0],ymm10[1],ymm15[1],ymm10[4],ymm15[4],ymm10[5],ymm15[5]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm1 = xmm1[3,3],mem[3,3]
; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
; AVX2-SLOW-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
; AVX2-SLOW-NEXT: # ymm0 = ymm7[2],mem[2],ymm7[3],mem[3],ymm7[6],mem[6],ymm7[7],mem[7]
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm14[2],ymm10[2],ymm14[3],ymm10[3],ymm14[6],ymm10[6],ymm14[7],ymm10[7]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm15[2],ymm10[2],ymm15[3],ymm10[3],ymm15[6],ymm10[6],ymm15[7],ymm10[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -9150,24 +9148,23 @@
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
; AVX2-SLOW-NEXT: vbroadcastss 80(%rdx), %ymm0
-; AVX2-SLOW-NEXT: vmovaps %ymm15, %ymm13
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm12[3,1,2,0,7,5,6,4]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm4[0],ymm2[1],ymm4[1],ymm2[4],ymm4[4],ymm2[5],ymm4[5]
+; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm6[0],ymm2[1],ymm6[1],ymm2[4],ymm6[4],ymm2[5],ymm6[5]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm1 = xmm1[3,3],mem[3,3]
; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm13[2],mem[2],ymm13[3],mem[3],ymm13[6],mem[6],ymm13[7],mem[7]
-; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
+; AVX2-SLOW-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
+; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm2[2],ymm6[3],ymm2[3],ymm6[6],ymm2[6],ymm6[7],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -9177,23 +9174,23 @@
; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
-; AVX2-SLOW-NEXT: vbroadcastss 112(%rdx), %ymm0
-; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm4[3,1,2,0,7,5,6,4]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm12 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
+; AVX2-SLOW-NEXT: vbroadcastss 112(%rdx), %ymm2
+; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-SLOW-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-SLOW-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm1[4,5],ymm6[6,7]
; AVX2-SLOW-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-SLOW-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm1 = xmm1[3,3],mem[3,3]
; AVX2-SLOW-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-SLOW-NEXT: # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
-; AVX2-SLOW-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
-; AVX2-SLOW-NEXT: # ymm0 = ymm4[2],mem[2],ymm4[3],mem[3],ymm4[6],mem[6],ymm4[7],mem[7]
+; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm1[1,2,3],ymm6[4,5,6,7]
+; AVX2-SLOW-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-SLOW-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
; AVX2-SLOW-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -9280,10 +9277,10 @@
; AVX2-SLOW-NEXT: vmovaps %ymm3, 992(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm4, 864(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm6, 768(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm13, 640(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm15, 544(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm14, 416(%rax)
-; AVX2-SLOW-NEXT: vmovaps %ymm12, 320(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm12, 640(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm13, 544(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm15, 416(%rax)
+; AVX2-SLOW-NEXT: vmovaps %ymm14, 320(%rax)
; AVX2-SLOW-NEXT: vmovaps %ymm9, 192(%rax)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 96(%rax)
@@ -9373,7 +9370,7 @@
; AVX2-SLOW-NEXT: vmovaps %ymm0, 1600(%rax)
; AVX2-SLOW-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-SLOW-NEXT: vmovaps %ymm0, 1568(%rax)
-; AVX2-SLOW-NEXT: addq $3000, %rsp # imm = 0xBB8
+; AVX2-SLOW-NEXT: addq $2968, %rsp # imm = 0xB98
; AVX2-SLOW-NEXT: vzeroupper
; AVX2-SLOW-NEXT: retq
;
@@ -9569,11 +9566,11 @@
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT: vmovaps (%rdx), %ymm11
+; AVX2-FAST-NEXT: vmovaps (%rdx), %ymm14
; AVX2-FAST-NEXT: vmovaps (%rcx), %ymm1
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm11[2],ymm1[2],ymm11[3],ymm1[3],ymm11[6],ymm1[6],ymm11[7],ymm1[7]
-; AVX2-FAST-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm14[2],ymm1[2],ymm14[3],ymm1[3],ymm14[6],ymm1[6],ymm14[7],ymm1[7]
+; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps (%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -9702,17 +9699,17 @@
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
; AVX2-FAST-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps 192(%rdi), %ymm1
-; AVX2-FAST-NEXT: vmovaps 192(%rsi), %ymm9
-; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm9[2],ymm1[3],ymm9[3],ymm1[6],ymm9[6],ymm1[7],ymm9[7]
-; AVX2-FAST-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 192(%rsi), %ymm11
+; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm11[2],ymm1[3],ymm11[3],ymm1[6],ymm11[6],ymm1[7],ymm11[7]
+; AVX2-FAST-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vmovaps %ymm1, %ymm13
; AVX2-FAST-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT: vmovaps 192(%rdx), %ymm7
-; AVX2-FAST-NEXT: vmovaps 192(%rcx), %ymm8
-; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm7[2],ymm8[2],ymm7[3],ymm8[3],ymm7[6],ymm8[6],ymm7[7],ymm8[7]
-; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovaps 192(%rdx), %ymm8
+; AVX2-FAST-NEXT: vmovaps 192(%rcx), %ymm9
+; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm1 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
; AVX2-FAST-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-FAST-NEXT: vmovaps 192(%r8), %ymm2
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -9736,27 +9733,28 @@
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1,2],ymm2[3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastss 228(%r8), %ymm4
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps 224(%r9), %xmm4
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm15 = xmm4[1,1,1,1]
+; AVX2-FAST-NEXT: vmovaps 224(%r9), %xmm7
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm15 = xmm7[1,1,1,1]
; AVX2-FAST-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm15[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vinsertf128 $1, 224(%rax), %ymm5, %ymm5
+; AVX2-FAST-NEXT: vmovaps 224(%rax), %xmm15
+; AVX2-FAST-NEXT: vinsertf128 $1, %xmm15, %ymm5, %ymm5
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm5[1],ymm2[2,3,4],ymm5[5],ymm2[6,7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vbroadcastss %xmm3, %xmm2
; AVX2-FAST-NEXT: vbroadcastss %xmm6, %xmm5
-; AVX2-FAST-NEXT: vunpcklps {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; AVX2-FAST-NEXT: vunpcklps {{.*#+}} xmm15 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; AVX2-FAST-NEXT: vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; AVX2-FAST-NEXT: vunpcklps {{.*#+}} xmm2 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; AVX2-FAST-NEXT: vbroadcastf128 {{.*#+}} ymm10 = [0,1,2,2,0,1,2,2]
; AVX2-FAST-NEXT: # ymm10 = mem[0,1,0,1]
-; AVX2-FAST-NEXT: vpermps %ymm15, %ymm10, %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm15[0,1],ymm2[2,3],ymm15[4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastsd 224(%r8), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm15[4,5],ymm2[6,7]
-; AVX2-FAST-NEXT: vbroadcastss %xmm4, %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm15[5],ymm2[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 224(%rax), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm15[6],ymm2[7]
+; AVX2-FAST-NEXT: vpermps %ymm2, %ymm10, %ymm2
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastsd 224(%r8), %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5],ymm2[6,7]
+; AVX2-FAST-NEXT: vbroadcastss %xmm7, %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5],ymm2[6,7]
+; AVX2-FAST-NEXT: vbroadcastss %xmm15, %ymm4
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6],ymm2[7]
; AVX2-FAST-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} xmm1 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
@@ -9765,7 +9763,7 @@
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6],ymm1[7]
; AVX2-FAST-NEXT: vmovaps 224(%r8), %ymm6
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm4[2,2,2,2]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm7[2,2,2,2]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3],ymm0[4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastss 232(%rax), %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
@@ -9774,51 +9772,51 @@
; AVX2-FAST-NEXT: vmovaps 224(%rsi), %ymm1
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm0 = ymm1[1,1,1,1,5,5,5,5]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3,4],ymm2[5],ymm0[6,7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm15 = ymm0[2,2,2,2]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm0[2,2,2,2]
; AVX2-FAST-NEXT: vmovaps 224(%rdx), %ymm3
; AVX2-FAST-NEXT: vmovaps 224(%rcx), %ymm0
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm14 = ymm3[1,1],ymm0[1,1],ymm3[5,5],ymm0[5,5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5,6],ymm15[7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm15 = ymm3[1,1],ymm0[1,1],ymm3[5,5],ymm0[5,5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm15[5,6],ymm4[7]
; AVX2-FAST-NEXT: vbroadcastsd 240(%r8), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0],ymm14[1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm15[0],ymm4[1,2,3,4,5,6],ymm15[7]
; AVX2-FAST-NEXT: vbroadcastss 240(%r9), %xmm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0],ymm15[1],ymm14[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm15[1],ymm4[2,3,4,5,6,7]
; AVX2-FAST-NEXT: vbroadcastss 240(%rax), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm14 = ymm8[2],ymm7[2],ymm8[3],ymm7[3],ymm8[6],ymm7[6],ymm8[7],ymm7[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[3,3,3,3]
-; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm15 = ymm9[2],ymm13[2],ymm9[3],ymm13[3],ymm9[6],ymm13[6],ymm9[7],ymm13[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2],ymm14[3,4,5,6,7]
-; AVX2-FAST-NEXT: vbroadcastss 220(%r8), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4],ymm15[5],ymm14[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 220(%r9), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-FAST-NEXT: vbroadcastsd 216(%rax), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm15[0],ymm14[1,2,3,4,5,6],ymm15[7]
-; AVX2-FAST-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vbroadcastss 240(%rdx), %ymm14
-; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm15 = ymm0[3,1,2,0,7,5,6,4]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4,5],ymm14[6],ymm15[7]
-; AVX2-FAST-NEXT: vunpcklps {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
-; AVX2-FAST-NEXT: vbroadcastss 236(%r8), %ymm15
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0],ymm15[1],ymm14[2,3,4,5,6,7]
-; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm4 = xmm4[2,2,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm14[0,1],ymm4[2,3],ymm14[4,5,6,7]
-; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm14 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[2,2,2,2]
-; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm15 = ymm3[2],ymm0[2],ymm3[3],ymm0[3],ymm3[6],ymm0[6],ymm3[7],ymm0[7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm14[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm14 = [5,6,5,6,5,6,5,6]
-; AVX2-FAST-NEXT: vpermps 224(%r9), %ymm14, %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm14[0],ymm6[1,2,3,4,5,6],ymm14[7]
-; AVX2-FAST-NEXT: vmovaps 224(%rax), %ymm14
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm14[3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm15[2],ymm4[3,4,5,6,7]
; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm14[2,3],ymm15[2,3]
+; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm4 = ymm9[2],ymm8[2],ymm9[3],ymm8[3],ymm9[6],ymm8[6],ymm9[7],ymm8[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm4 = ymm4[3,3,3,3]
+; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm15 = ymm11[2],ymm13[2],ymm11[3],ymm13[3],ymm11[6],ymm13[6],ymm11[7],ymm13[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm15 = ymm15[3,3,3,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm15[0,1,2],ymm4[3,4,5,6,7]
+; AVX2-FAST-NEXT: vbroadcastss 220(%r8), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm15[5],ymm4[6,7]
+; AVX2-FAST-NEXT: vbroadcastss 220(%r9), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-FAST-NEXT: vbroadcastsd 216(%rax), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm15[0],ymm4[1,2,3,4,5,6],ymm15[7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vbroadcastss 240(%rdx), %ymm4
+; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm15 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm15[0,1,2,3,4,5],ymm4[6],ymm15[7]
+; AVX2-FAST-NEXT: vunpcklps {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm15[4,5],ymm4[6,7]
+; AVX2-FAST-NEXT: vbroadcastss 236(%r8), %ymm15
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm15[1],ymm4[2,3,4,5,6,7]
+; AVX2-FAST-NEXT: vshufps {{.*#+}} xmm7 = xmm7[2,2,3,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm7[2,3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm7 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm7 = ymm7[2,2,2,2]
+; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm15 = ymm3[2],ymm0[2],ymm3[3],ymm0[3],ymm3[6],ymm0[6],ymm3[7],ymm0[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-FAST-NEXT: vbroadcastsd {{.*#+}} ymm7 = [5,6,5,6,5,6,5,6]
+; AVX2-FAST-NEXT: vpermps 224(%r9), %ymm7, %ymm7
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0],ymm6[1,2,3,4,5,6],ymm7[7]
+; AVX2-FAST-NEXT: vmovaps 224(%rax), %ymm7
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2],ymm7[3],ymm4[4,5,6,7]
+; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm7[2,3],ymm15[2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0],ymm4[1],ymm6[2,3,4],ymm4[5],ymm6[6,7]
; AVX2-FAST-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
@@ -9867,8 +9865,8 @@
; AVX2-FAST-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT: vshufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm1 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm1 = ymm11[1,1],mem[1,1],ymm11[5,5],mem[5,5]
+; AVX2-FAST-NEXT: vshufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm1 = ymm14[1,1],mem[1,1],ymm14[5,5],mem[5,5]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,0,0,0,4,4,4,4]
; AVX2-FAST-NEXT: vpermilps $68, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
@@ -9937,8 +9935,8 @@
; AVX2-FAST-NEXT: vunpcklps {{.*#+}} xmm1 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-FAST-NEXT: vpermps %ymm1, %ymm10, %ymm1
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
-; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-FAST-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX2-FAST-NEXT: vunpcklps {{.*#+}} xmm1 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
; AVX2-FAST-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
@@ -10009,8 +10007,8 @@
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm11[1,1],ymm1[5,5],ymm11[5,5]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
; AVX2-FAST-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
@@ -10253,6 +10251,7 @@
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm10 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-NEXT: vbroadcastss 112(%rdx), %ymm0
+; AVX2-FAST-NEXT: vmovaps %ymm11, %ymm3
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm5 = ymm11[3,1,2,0,7,5,6,4]
; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6],ymm5[7]
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
@@ -10263,20 +10262,20 @@
; AVX2-FAST-NEXT: # xmm5 = xmm4[3,3],mem[3,3]
; AVX2-FAST-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
; AVX2-FAST-NEXT: # xmm5 = xmm5[0,1,2],mem[3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0],ymm5[1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm5 = ymm11[2],mem[2],ymm11[3],mem[3],ymm11[6],mem[6],ymm11[7],mem[7]
-; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm11 = ymm2[2],ymm12[2],ymm2[3],ymm12[3],ymm2[6],ymm12[6],ymm2[7],ymm12[7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0],ymm5[1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm0 = ymm3[2],mem[2],ymm3[3],mem[3],ymm3[6],mem[6],ymm3[7],mem[7]
+; AVX2-FAST-NEXT: vunpckhps {{.*#+}} ymm5 = ymm2[2],ymm12[2],ymm2[3],ymm12[3],ymm2[6],ymm12[6],ymm2[7],ymm12[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[3,3,3,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm11[0,1,2],ymm5[3,4,5,6,7]
-; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm11 # 32-byte Folded Reload
-; AVX2-FAST-NEXT: # ymm11 = ymm0[3,3],mem[3,3],ymm0[7,7],mem[7,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3,4,5,6,7]
+; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-FAST-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-NEXT: # ymm5 = ymm1[3,3],mem[3,3],ymm1[7,7],mem[7,7]
; AVX2-FAST-NEXT: vpermilps {{.*#+}} ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0],ymm11[1,2],ymm12[3,4],ymm11[5,6],ymm12[7]
-; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm11 = ymm11[2,1,2,3]
-; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0],ymm5[1,2,3,4],ymm11[5,6,7]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm5 = ymm12[0],ymm5[1,2],ymm12[3,4],ymm5[5,6],ymm12[7]
+; AVX2-FAST-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,1,2,3]
+; AVX2-FAST-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0],ymm0[1,2,3,4],ymm5[5,6,7]
; AVX2-FAST-NEXT: vbroadcastss 144(%rdx), %ymm5
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-NEXT: vshufps {{.*#+}} ymm12 = ymm2[3,1,2,0,7,5,6,4]
@@ -10346,8 +10345,8 @@
; AVX2-FAST-NEXT: vmovaps %ymm5, 1216(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm3, 1088(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm12, 992(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm11, 864(%rax)
-; AVX2-FAST-NEXT: vmovaps %ymm4, 768(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm4, 864(%rax)
+; AVX2-FAST-NEXT: vmovaps %ymm11, 768(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm10, 640(%rax)
; AVX2-FAST-NEXT: vmovaps %ymm15, 544(%rax)
; AVX2-FAST-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -10450,31 +10449,31 @@
;
; AVX2-FAST-PERLANE-LABEL: store_i32_stride7_vf64:
; AVX2-FAST-PERLANE: # %bb.0:
-; AVX2-FAST-PERLANE-NEXT: subq $3000, %rsp # imm = 0xBB8
+; AVX2-FAST-PERLANE-NEXT: subq $2968, %rsp # imm = 0xB98
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rax), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovaps (%r8), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%r8), %xmm4
-; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps (%r8), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%r8), %xmm3
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps (%r9), %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%r9), %xmm7
-; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%r9), %xmm4
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm5[1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd %xmm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rcx), %xmm9
-; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rcx), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rcx), %xmm10
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdx), %xmm8
; AVX2-FAST-PERLANE-NEXT: vinsertps {{.*#+}} xmm1 = zero,xmm8[1],xmm9[1],zero
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rdi), %xmm6
-; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %xmm10
-; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdi), %xmm7
+; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovaps (%rsi), %xmm5
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rsi), %xmm11
; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -10487,15 +10486,15 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rax), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm7[1,1,1,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm4[1],xmm1[2,3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0],xmm3[1],xmm1[2,3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd %xmm1, %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm1 = xmm11[1,1,2,2]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm10[2],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2],xmm1[3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vmovaps 32(%rdx), %xmm7
-; AVX2-FAST-PERLANE-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm7[1],xmm3[1],zero
+; AVX2-FAST-PERLANE-NEXT: vinsertps {{.*#+}} xmm2 = zero,xmm7[1],xmm10[1],zero
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2],ymm1[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10674,11 +10673,11 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm1
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rsi), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdi), %ymm0
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rsi), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vmovaps 64(%rdx), %ymm1
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -10805,21 +10804,22 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%r9), %xmm3
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm3, %ymm14
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm15[0,1,2,3,4],ymm14[5],ymm15[6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 224(%rax), %ymm15
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm14[0,1,2,3,4,5],ymm15[6],ymm14[7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rax), %xmm15
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm15, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm14[0,1,2,3,4,5],ymm13[6],ymm14[7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm14 = xmm4[1,1,2,2]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0,1],xmm2[2],xmm14[3]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[0,1,2,1]
-; AVX2-FAST-PERLANE-NEXT: vinsertps {{.*#+}} xmm15 = zero,xmm1[1],xmm0[1],zero
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0],ymm15[1,2],ymm14[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 228(%r8), %ymm13
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2],ymm13[3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm14 = xmm3[1,1,1,1]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm14[4,5],ymm13[6,7]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, 224(%rax), %ymm15, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm13[0],ymm14[1],ymm13[2,3,4],ymm14[5],ymm13[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm13 = xmm4[1,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm2[2],xmm13[3]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT: vinsertps {{.*#+}} xmm14 = zero,xmm1[1],xmm0[1],zero
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1,2],ymm13[3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 228(%r8), %ymm12
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1,2],ymm12[3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm13 = xmm3[1,1,1,1]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm15, %ymm14, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0],ymm13[1],ymm12[2,3,4],ymm13[5],ymm12[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3],xmm4[3,3]
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
@@ -10835,75 +10835,74 @@
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 232(%rax), %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdi), %ymm12
-; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rsi), %ymm10
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm10[1,1,1,1,5,5,5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1],ymm0[2,3,4],ymm12[5],ymm0[6,7]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm13 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdx), %ymm0
+; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdi), %ymm4
+; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rsi), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm12 = ymm0[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rdx), %ymm10
+; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rcx), %ymm1
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm13 = ymm10[1,1],ymm1[1,1],ymm10[5,5],ymm1[5,5]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4],ymm13[5,6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 240(%r8), %ymm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0],ymm12[1,2,3,4,5,6],ymm13[7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 240(%r9), %xmm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0],ymm13[1],ymm12[2,3,4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 240(%rax), %ymm13
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1],ymm13[2],ymm12[3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps 224(%rcx), %ymm2
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm14 = ymm0[1,1],ymm2[1,1],ymm0[5,5],ymm2[5,5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm14[5,6],ymm13[7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 240(%r8), %ymm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0],ymm13[1,2,3,4,5,6],ymm14[7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 240(%r9), %xmm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0],ymm14[1],ymm13[2,3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 240(%rax), %ymm14
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm14[2],ymm13[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm9, %xmm13
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm8, %xmm14
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm13 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm14 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm14 = xmm14[0,1,2,2]
-; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm14 = ymm14[0,1,2,1]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm14[0,1],ymm13[2,3],ymm14[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm9, %xmm12
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm8, %xmm13
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm13 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm13 = xmm13[0,1,2,2]
+; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm13 = ymm13[0,1,2,1]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2,3],ymm13[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm14 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 16-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm14[0],ymm15[0],ymm14[2],ymm15[2]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm14[4,5,6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm13 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm13, %ymm0, %ymm13
+; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 16-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm13[0],ymm14[0],ymm13[2],ymm14[2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm12, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm6[3,3],xmm5[3,3]
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm4[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm0[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm15[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 8(%rax), %ymm8
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm1, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm12, %xmm5
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm7, %xmm6
; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm4[0],xmm14[0],xmm4[1],xmm14[1]
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm13[0],xmm9[1],xmm13[1]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm15[0],xmm9[1],xmm15[1]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm8[0],ymm6[2],ymm8[2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm4[3,3],xmm14[3,3]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm13[3,3],xmm14[3,3]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm7[2],xmm12[2],xmm7[3],xmm12[3]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm13[2,2,2,2]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm15[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm9[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 40(%rax), %ymm7
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
@@ -10913,40 +10912,40 @@
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm13, %xmm5
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm11, %xmm6
; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm8[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 72(%rax), %ymm7
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm9, %xmm5
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm13, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm13, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm14, %xmm6
; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm9[0],xmm12[1],xmm9[1]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
@@ -10958,8 +10957,8 @@
; AVX2-FAST-PERLANE-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm4[3,3],xmm1[3,3]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm9[2],xmm13[3],xmm9[3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm12[3,3],xmm9[3,3]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
@@ -10970,14 +10969,14 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm9, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm12, %xmm5
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm13, %xmm6
; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
@@ -10987,10 +10986,10 @@
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm8[3,3],xmm4[3,3]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm9[2],xmm13[3],xmm9[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
@@ -11001,14 +11000,14 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm9, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm12, %xmm5
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm13, %xmm6
; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
@@ -11018,10 +11017,10 @@
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
-; AVX2-FAST-PERLANE-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm8[3,3],xmm4[3,3]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm9[2],xmm13[3],xmm9[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
@@ -11032,40 +11031,40 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm9, %xmm5
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm11, %xmm6
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm11, %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss %xmm13, %xmm6
; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6],ymm5[7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm8[3,3],xmm4[3,3]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm11[2],xmm9[2],xmm11[3],xmm9[3]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm5 = xmm9[3,3],xmm8[3,3]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} xmm6 = xmm13[2],xmm11[2],xmm13[3],xmm11[3]
; AVX2-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm6[0,1,2,2]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm6 = ymm6[0,1,2,1]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm6 = xmm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm6 = xmm12[0,1,2],xmm6[3]
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 200(%rax), %ymm7
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm6[2,3,4],ymm5[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm4[2],mem[2],ymm4[3],mem[3],ymm4[6],mem[6],ymm4[7],mem[7]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
@@ -11080,18 +11079,17 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0],ymm5[1,2,3,4,5,6],ymm6[7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 240(%rdx), %ymm5
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm2[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm6 = ymm1[3,1,2,0,7,5,6,4]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6],ymm6[7]
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} ymm6 = ymm12[0],ymm10[0],ymm12[1],ymm10[1],ymm12[4],ymm10[4],ymm12[5],ymm10[5]
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} ymm6 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 236(%r8), %ymm6
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2,3,4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm3 = xmm3[2,2,3,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3],ymm5[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm5 = ymm12[2],ymm10[2],ymm12[3],ymm10[3],ymm12[6],ymm10[6],ymm12[7],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm5 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm6 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm6 = ymm10[2],ymm1[2],ymm10[3],ymm1[3],ymm10[6],ymm1[6],ymm10[7],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
@@ -11104,8 +11102,8 @@
; AVX2-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm7[2,3],ymm6[2,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0],ymm3[1],ymm5[2,3,4],ymm3[5],ymm5[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm0 = ymm2[2],ymm1[2],ymm2[3],ymm1[3],ymm2[6],ymm1[6],ymm2[7],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm10[2],ymm12[2],ymm10[3],ymm12[3],ymm10[6],ymm12[6],ymm10[7],ymm12[7]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm10[2],ymm1[3],ymm10[3],ymm1[6],ymm10[6],ymm1[7],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm2[2],ymm4[2],ymm2[3],ymm4[3],ymm2[6],ymm4[6],ymm2[7],ymm4[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -11116,17 +11114,17 @@
; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 248(%rax), %ymm1
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3,4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm5[1],ymm0[2,3,4],ymm5[5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm5[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm4[1],ymm0[2,3,4],ymm4[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm8[1,1],ymm3[1,1],ymm8[5,5],ymm3[5,5]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm12[0,0,0,0,4,4,4,4]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm14[0,0,0,0,4,4,4,4]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm2 = ymm9[0,1,0,1,4,5,4,5]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2,3,4],ymm1[5],ymm2[6,7]
@@ -11135,8 +11133,8 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm14[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm15[1,1,1,1,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm10[1],ymm0[2,3,4],ymm10[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
@@ -11154,14 +11152,14 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-FAST-PERLANE-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = mem[1,1,1,1,5,5,5,5]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm6[1,1,1,1,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[0],mem[1],ymm0[2,3,4],mem[5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[1,1],ymm15[1,1],ymm1[5,5],ymm15[5,5]
+; AVX2-FAST-PERLANE-NEXT: vshufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm1 = ymm1[1,1],mem[1,1],ymm1[5,5],mem[5,5]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6],ymm0[7]
; AVX2-FAST-PERLANE-NEXT: vpermilps $0, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm1 = mem[0,0,0,0,4,4,4,4]
@@ -11231,7 +11229,7 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm4[1,1],ymm0[5,5],ymm4[5,5]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[1,1],ymm12[1,1],ymm0[5,5],ymm12[5,5]
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm13[1,1,1,1,5,5,5,5]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm11[1],ymm1[2,3,4],ymm11[5],ymm1[6,7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
@@ -11249,7 +11247,7 @@
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 16(%rdx), %ymm0
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm3[3,1,2,0,7,5,6,4]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} ymm1 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} ymm1 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
@@ -11259,11 +11257,11 @@
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm0 = ymm3[2],ymm8[2],ymm3[3],ymm8[3],ymm3[6],ymm8[6],ymm3[7],ymm8[7]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm5[2],ymm6[3],ymm5[3],ymm6[6],ymm5[6],ymm6[7],ymm5[7]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm5[2],ymm4[2],ymm5[3],ymm4[3],ymm5[6],ymm4[6],ymm5[7],ymm4[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,3],ymm12[3,3],ymm9[7,7],ymm12[7,7]
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm9[3,3],ymm14[3,3],ymm9[7,7],ymm14[7,7]
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
@@ -11271,17 +11269,17 @@
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 48(%rdx), %ymm0
; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm7[3,1,2,0,7,5,6,4]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} ymm1 = ymm10[0],ymm14[0],ymm10[1],ymm14[1],ymm10[4],ymm14[4],ymm10[5],ymm14[5]
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} ymm1 = ymm10[0],ymm15[0],ymm10[1],ymm15[1],ymm10[4],ymm15[4],ymm10[5],ymm15[5]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm1 = xmm1[3,3],mem[3,3]
; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm7[2],mem[2],ymm7[3],mem[3],ymm7[6],mem[6],ymm7[7],mem[7]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm14[2],ymm10[2],ymm14[3],ymm10[3],ymm14[6],ymm10[6],ymm14[7],ymm10[7]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm15[2],ymm10[2],ymm15[3],ymm10[3],ymm15[6],ymm10[6],ymm15[7],ymm10[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -11291,24 +11289,23 @@
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm14 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
; AVX2-FAST-PERLANE-NEXT: vbroadcastss 80(%rdx), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm15, %ymm13
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm15[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm12[3,1,2,0,7,5,6,4]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm4[0],ymm2[1],ymm4[1],ymm2[4],ymm4[4],ymm2[5],ymm4[5]
+; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm6[0],ymm2[1],ymm6[1],ymm2[4],ymm6[4],ymm2[5],ymm6[5]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm1 = xmm1[3,3],mem[3,3]
; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm13[2],mem[2],ymm13[3],mem[3],ymm13[6],mem[6],ymm13[7],mem[7]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm12[2],mem[2],ymm12[3],mem[3],ymm12[6],mem[6],ymm12[7],mem[7]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm6[2],ymm2[2],ymm6[3],ymm2[3],ymm6[6],ymm2[6],ymm6[7],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
@@ -11318,23 +11315,23 @@
; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm2 = mem[2,3,2,3,6,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0],ymm1[1,2],ymm2[3,4],ymm1[5,6],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,1,2,3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vbroadcastss 112(%rdx), %ymm0
-; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm4[3,1,2,0,7,5,6,4]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6],ymm1[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm12 = ymm1[0],ymm0[1,2,3,4],ymm1[5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vbroadcastss 112(%rdx), %ymm2
+; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} ymm1 = ymm0[3,1,2,0,7,5,6,4]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3,4,5],ymm2[6],ymm1[7]
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm1[4,5],ymm6[6,7]
; AVX2-FAST-PERLANE-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-FAST-PERLANE-NEXT: vshufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm1 = xmm1[3,3],mem[3,3]
; AVX2-FAST-PERLANE-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
; AVX2-FAST-PERLANE-NEXT: # xmm1 = xmm1[0,1,2],mem[3]
-; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0],ymm1[1,2,3],ymm0[4,5,6,7]
-; AVX2-FAST-PERLANE-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm0 # 32-byte Folded Reload
-; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm4[2],mem[2],ymm4[3],mem[3],ymm4[6],mem[6],ymm4[7],mem[7]
+; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm1[1,2,3],ymm6[4,5,6,7]
+; AVX2-FAST-PERLANE-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-FAST-PERLANE-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
; AVX2-FAST-PERLANE-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[3,3,3,3]
@@ -11421,10 +11418,10 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm3, 992(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm4, 864(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm6, 768(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, 640(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm15, 544(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, 416(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm12, 320(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm12, 640(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm13, 544(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm15, 416(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm14, 320(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm9, 192(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 96(%rax)
@@ -11514,7 +11511,7 @@
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 1600(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm0, 1568(%rax)
-; AVX2-FAST-PERLANE-NEXT: addq $3000, %rsp # imm = 0xBB8
+; AVX2-FAST-PERLANE-NEXT: addq $2968, %rsp # imm = 0xB98
; AVX2-FAST-PERLANE-NEXT: vzeroupper
; AVX2-FAST-PERLANE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll
index 6ae1465..3619d7f 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-3.ll
@@ -111,11 +111,11 @@
; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm0
; AVX1-ONLY-NEXT: vmovapd (%rdx), %ymm1
; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm2
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovapd 16(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm4 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm4, %ymm2
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3],ymm4[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm0[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm4[2,3]
@@ -228,36 +228,36 @@
; AVX1-ONLY-NEXT: vmovapd (%rdx), %ymm2
; AVX1-ONLY-NEXT: vmovapd 32(%rdx), %ymm3
; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovapd 16(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm6
-; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm8 = xmm4[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm8, %ymm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm8, %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm8[0,1],ymm4[2,3],ymm8[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm8 = xmm6[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm8, %ymm6
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm8, %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1],ymm6[2,3],ymm8[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm1[0,0,3,2]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm3[2,3],ymm8[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2],ymm7[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm5 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm5, %ymm4
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm6 = xmm5[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm6, %ymm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm3[2,3],ymm7[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm7[0],ymm6[1],ymm7[2],ymm6[3]
+; AVX1-ONLY-NEXT: vmovapd 16(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm0[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm2[2,3],ymm8[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm8[0],ymm5[1],ymm8[2],ymm5[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2],ymm7[3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vmovapd %ymm5, 64(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm7, 160(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, 64(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 160(%rcx)
; AVX1-ONLY-NEXT: vmovapd %ymm1, 128(%rcx)
; AVX1-ONLY-NEXT: vmovapd %ymm0, 32(%rcx)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 96(%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm4, (%rcx)
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
@@ -444,27 +444,26 @@
;
; AVX1-ONLY-LABEL: store_i64_stride3_vf16:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm6
+; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm7
; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm2
-; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm5
+; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm6
; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovapd (%rdx), %ymm9
; AVX1-ONLY-NEXT: vmovapd 32(%rdx), %ymm4
; AVX1-ONLY-NEXT: vmovapd 64(%rdx), %ymm8
; AVX1-ONLY-NEXT: vmovapd 96(%rdx), %ymm3
; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm7 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm7, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm7, %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm0[2,3],ymm7[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm5 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm5, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm5, %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3],ymm5[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm7
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm11 = xmm7[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm11, %ymm7
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdx), %ymm11, %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3],ymm11[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm10 = xmm5[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm10, %ymm5
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdx), %ymm10, %ymm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm10[0,1],ymm5[2,3],ymm10[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm10
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm11 = xmm10[0],mem[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm11, %ymm10
; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm11, %ymm11
@@ -476,9 +475,10 @@
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3],ymm12[4,5],ymm11[6,7]
; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm12
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm5[0,0,3,2]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm8[2,3],ymm14[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm14[0],ymm12[1],ymm14[2],ymm12[3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm6[0,0,3,2]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm8[2,3],ymm13[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm13[0],ymm12[1],ymm13[2],ymm12[3]
+; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm13
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm13[0,1],ymm4[2,3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm2[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm4[2,3],ymm14[2,3]
@@ -490,15 +490,15 @@
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm15[0],ymm14[1],ymm15[2],ymm14[3]
; AVX1-ONLY-NEXT: vmovapd 16(%rdi), %xmm15
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm15[0,1],ymm9[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm6[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm0[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm15[1],ymm0[2],ymm15[3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[1,0,2,2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2],ymm7[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2,3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],mem[2],ymm6[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm9[1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[1,0,2,2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2],ymm5[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0],ymm8[1],ymm5[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm8[1],ymm6[2,3]
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],mem[2],ymm2[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2,3]
@@ -510,12 +510,12 @@
; AVX1-ONLY-NEXT: vmovapd %ymm1, 320(%rcx)
; AVX1-ONLY-NEXT: vmovapd %ymm13, 160(%rcx)
; AVX1-ONLY-NEXT: vmovapd %ymm2, 128(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm5, 224(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 224(%rcx)
; AVX1-ONLY-NEXT: vmovapd %ymm12, 256(%rcx)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 32(%rcx)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, 32(%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm11, 288(%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm10, 96(%rcx)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 192(%rcx)
+; AVX1-ONLY-NEXT: vmovaps %ymm5, 192(%rcx)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, (%rcx)
; AVX1-ONLY-NEXT: vzeroupper
@@ -882,79 +882,79 @@
; AVX1-ONLY-LABEL: store_i64_stride3_vf32:
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: subq $424, %rsp # imm = 0x1A8
-; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm6
-; AVX1-ONLY-NEXT: vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm5
-; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm3
+; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm2
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd (%rdx), %ymm12
; AVX1-ONLY-NEXT: vmovapd 32(%rdx), %ymm13
-; AVX1-ONLY-NEXT: vmovapd 64(%rdx), %ymm7
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovapd 16(%rdi), %xmm4
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm3, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm2
+; AVX1-ONLY-NEXT: vmovapd 64(%rdx), %ymm4
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm1, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdx), %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdx), %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, 128(%rdx), %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdx), %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, 192(%rdx), %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%rdx), %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm4[0,1],ymm12[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm6[0,0,3,2]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm12[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0],ymm1[1],ymm2[2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, 64(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, 96(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, 128(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, 160(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, 192(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 16(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm12[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm3[0,0,3,2]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm12[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm13[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm5[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm2[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm13[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm10
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm10[0,0,3,2]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm7[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 96(%rdx), %ymm4
@@ -1802,16 +1802,16 @@
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: subq $1480, %rsp # imm = 0x5C8
; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, (%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rdx), %ymm1, %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm0
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll
index c22478b..a1b3dd9 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll
@@ -148,31 +148,32 @@
; AVX1-ONLY-NEXT: vmovapd (%rdi), %ymm0
; AVX1-ONLY-NEXT: vmovapd (%r8), %ymm1
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = mem[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovapd 16(%rdx), %xmm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[2],ymm2[3]
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm3
+; AVX1-ONLY-NEXT: vmovapd 16(%rdx), %xmm4
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[2],ymm2[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm1[2],ymm2[3]
; AVX1-ONLY-NEXT: vmovddup {{.*#+}} xmm4 = mem[0,0]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm0[0],mem[0],ymm0[2],mem[2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm1[1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vmovlpd {{.*#+}} xmm3 = mem[0],xmm3[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 24(%rcx), %ymm5
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovapd 16(%rsi), %xmm5
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm5[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 24(%rcx), %ymm6
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm6
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm6[2],ymm0[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm5, 16(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm1, (%r9)
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm3, (%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm1, 16(%r9)
; AVX1-ONLY-NEXT: vmovapd %ymm4, 64(%r9)
; AVX1-ONLY-NEXT: vmovapd %ymm0, 32(%r9)
; AVX1-ONLY-NEXT: vmovapd %ymm2, 96(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm3, 128(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm5, 128(%r9)
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
;
@@ -331,63 +332,63 @@
;
; AVX1-ONLY-LABEL: store_i64_stride5_vf8:
; AVX1-ONLY: # %bb.0:
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm10
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovaps (%rcx), %ymm0
; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm3
; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm5
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm2 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm7
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm7[1],ymm0[3],ymm7[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm8
+; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm6
+; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm11
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm6
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm6[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm6 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm9[0],mem[0],ymm9[2],mem[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm5[1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = mem[2,3],ymm9[2,3]
-; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm10
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm10[0],ymm8[0],ymm10[2],ymm8[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm5[2],ymm8[3]
-; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm11
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm12
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm4
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3,4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm3[0,1],ymm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm7 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm10[0],mem[0],ymm10[2],mem[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0],ymm5[1],ymm7[2,3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = mem[2,3],ymm10[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm11[0],ymm9[0],ymm11[2],ymm9[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm5[2],ymm9[3]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm12
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm13[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1,2],ymm12[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm5[0],ymm9[1,2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm5[0],ymm10[1,2,3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm12 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm13 = ymm1[0],mem[0],ymm1[2],mem[2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm3[2,3],ymm12[4,5,6,7]
; AVX1-ONLY-NEXT: vmovapd 48(%rsi), %xmm13
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm10 = xmm13[1],xmm10[1]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm13[1],xmm11[1]
; AVX1-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm13[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm10[0,1,2],ymm5[3]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm11[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm11[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm7
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm10
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm4, 16(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm7, (%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm10, 160(%r9)
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm3 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm8
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm6, (%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm2, 16(%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm8, 160(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm3, 176(%r9)
; AVX1-ONLY-NEXT: vmovaps %ymm12, 64(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm9, 192(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm8, 256(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 224(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 32(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm10, 192(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, 256(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm7, 224(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 32(%r9)
; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%r9)
; AVX1-ONLY-NEXT: vmovaps %ymm0, 128(%r9)
; AVX1-ONLY-NEXT: vmovapd %ymm5, 288(%r9)
@@ -780,150 +781,152 @@
; AVX1-ONLY-LABEL: store_i64_stride5_vf16:
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: subq $216, %rsp
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm3
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm4
-; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm7
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm6
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm5
+; AVX1-ONLY-NEXT: vmovapd 96(%rdi), %ymm8
; AVX1-ONLY-NEXT: vmovaps (%rcx), %ymm0
-; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm5 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 64(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm2
+; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm9
+; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm3
+; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 80(%rdx), %xmm1
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm7[0],mem[0],ymm7[2],mem[2]
+; AVX1-ONLY-NEXT: vmovaps 80(%rdx), %xmm13
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm13[1],ymm1[3],ymm13[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm8[0],mem[0],ymm8[2],mem[2]
; AVX1-ONLY-NEXT: vmovapd 96(%rcx), %xmm1
; AVX1-ONLY-NEXT: vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%rsi), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm4[0],mem[0],ymm4[2],mem[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm2[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm5[0],mem[0],ymm5[2],mem[2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm1
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm3[0],mem[0],ymm3[2],mem[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm6[0],mem[0],ymm6[2],mem[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm0[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm15
; AVX1-ONLY-NEXT: vbroadcastsd 72(%rsi), %ymm0
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 64(%rdx), %xmm11
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1,2,3,4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3,4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm15[0],mem[0],ymm15[2],mem[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm10[4,5,6,7]
; AVX1-ONLY-NEXT: vbroadcastsd 104(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm7[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm1
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm12
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm10[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm8[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vmovaps 96(%rdx), %xmm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm12
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm10[0,1,2],ymm12[3]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm9[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm9
-; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm0
+; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = ymm0[0,1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm1 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5],ymm9[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm0 = ymm9[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm9[2,3],ymm13[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm9[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = mem[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm4
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[2],ymm3[3]
-; AVX1-ONLY-NEXT: vmovapd 48(%rsi), %xmm13
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm13[1],xmm4[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm6[2,3]
+; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm5
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm5[0],ymm0[0],ymm5[2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd 48(%rsi), %xmm6
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm6[1],xmm5[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm6
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm5[0,1],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm0[0],ymm14[1,2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm8[0],ymm0[1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm3[0,1],ymm0[2],ymm3[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = ymm0[0],mem[1,2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm14[0],ymm0[1],ymm14[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm1[0,1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 64(%r8), %ymm15
-; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm3 = mem[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm15[0,1],ymm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm15[2,3],ymm5[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm15[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 64(%r8), %ymm13
+; AVX1-ONLY-NEXT: vblendps $63, (%rsp), %ymm13, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5],ymm13[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm13[0,1],ymm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm13[2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm13[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm7[2,3]
-; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm7
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm7[0],ymm0[0],ymm7[2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 112(%rsi), %xmm9
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm7 = xmm9[1],xmm7[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3]
-; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm9
-; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm4 = mem[0],ymm9[1],mem[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm9[0],ymm2[1,2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm0[0,1],ymm9[2],ymm0[3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1,2],ymm9[3]
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm8
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm8[0],ymm1[0],ymm8[2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 112(%rsi), %xmm13
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm13[1],xmm8[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm13
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm13
+; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm5 = mem[0],ymm13[1],mem[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm13[0],ymm3[1,2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm13[2],ymm1[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm12
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm7
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm15
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm13, 16(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm15, (%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm1, 496(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm10, 480(%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm10, 16(%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm7, (%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm2, 496(%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm6, 480(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm0, 176(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm12, 160(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm11, 336(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm9, 320(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm2, 576(%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm13, 320(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm1, 576(%r9)
; AVX1-ONLY-NEXT: vmovapd %ymm3, 512(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 384(%r9)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 352(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm8, 256(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm14, 224(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 384(%r9)
+; AVX1-ONLY-NEXT: vmovaps %ymm15, 352(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm14, 256(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, 224(%r9)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 192(%r9)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 64(%r9)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm7, 608(%r9)
-; AVX1-ONLY-NEXT: vmovapd %ymm4, 544(%r9)
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovapd %ymm8, 608(%r9)
+; AVX1-ONLY-NEXT: vmovapd %ymm5, 544(%r9)
+; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 448(%r9)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 416(%r9)
@@ -1774,16 +1777,16 @@
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm9[0],mem[0],ymm9[2],mem[2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm13
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm14
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rsi), %ymm0
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm13[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm1
; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm13[0],mem[0],ymm13[2],mem[2]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm14[0],mem[0],ymm14[2],mem[2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %ymm2
@@ -1835,40 +1838,40 @@
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm8
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm4[0,1,2],ymm8[3]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm14[2,3]
-; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm14[0,1,2,3],ymm9[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm14
-; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = mem[0,1,2,3,4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm14[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = mem[0,1],ymm14[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm14[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm15[0,1,2,3],ymm9[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm15
+; AVX1-ONLY-NEXT: vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = mem[0,1,2,3,4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm15[0,1],mem[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm15[4,5],ymm9[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = mem[2,3],ymm13[2,3]
-; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm13
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm13[0],ymm9[0],ymm13[2],ymm9[3]
-; AVX1-ONLY-NEXT: vmovapd 48(%rsi), %xmm14
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm13 = xmm14[1],xmm13[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3]
-; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm14
-; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = ymm14[0],mem[1,2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm15 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm15 = mem[0],ymm14[1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm14[2],ymm9[3]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = mem[2,3],ymm14[2,3]
+; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm14
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm14[0],ymm9[0],ymm14[2],ymm9[3]
+; AVX1-ONLY-NEXT: vmovapd 48(%rsi), %xmm15
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm14 = xmm15[1],xmm14[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%rcx), %ymm15
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm15
+; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm15[0],mem[1,2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = mem[0],ymm15[1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm15[2],ymm9[3]
; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm13[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm14[0,1,2],ymm15[3]
; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
@@ -1889,20 +1892,20 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = mem[2,3],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovapd 112(%rdx), %xmm9
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm9[0],ymm2[0],ymm9[2],ymm2[3]
-; AVX1-ONLY-NEXT: vmovapd 112(%rsi), %xmm13
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm13[1],xmm9[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm13
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm13[2,3]
-; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm13
-; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = mem[0],ymm13[1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm13[0],mem[1,2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm13[2],ymm2[3]
+; AVX1-ONLY-NEXT: vmovapd 112(%rsi), %xmm14
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm14[1],xmm9[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 120(%rcx), %ymm14
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT: vmovapd 96(%r8), %ymm14
+; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = mem[0],ymm14[1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm13 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm13 = ymm14[0],mem[1,2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2],ymm2[3]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm9[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm9[0,1,2],ymm14[3]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
@@ -1921,18 +1924,18 @@
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm11[2,3]
; AVX1-ONLY-NEXT: vmovapd 176(%rdx), %xmm2
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm2[0],ymm1[0],ymm2[2],ymm1[3]
-; AVX1-ONLY-NEXT: vmovapd 176(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 184(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm1[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[3]
+; AVX1-ONLY-NEXT: vmovapd 176(%rsi), %xmm14
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm14[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 184(%rcx), %ymm14
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2,3]
; AVX1-ONLY-NEXT: vmovapd 160(%r8), %ymm14
-; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm1 = mem[0],ymm14[1],mem[2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm14[0],ymm12[1,2,3]
-; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm11[0,1],ymm14[2],ymm11[3]
+; AVX1-ONLY-NEXT: vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm9 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT: # ymm9 = mem[0],ymm14[1],mem[2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm14[0],ymm12[1,2,3]
+; AVX1-ONLY-NEXT: vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm14[2],ymm1[3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm2[0,1,2],ymm14[3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
@@ -1953,36 +1956,36 @@
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovapd 240(%rdx), %xmm2
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 240(%rsi), %xmm11
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm11[1],xmm2[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 248(%rcx), %ymm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm11[2,3]
-; AVX1-ONLY-NEXT: vmovapd 224(%r8), %ymm11
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm11[0],ymm8[1,2,3]
+; AVX1-ONLY-NEXT: vmovapd 240(%rsi), %xmm14
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm14[1],xmm2[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 248(%rcx), %ymm14
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT: vmovapd 224(%r8), %ymm14
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm14[0],ymm8[1,2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm4[0],ymm11[1],ymm4[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm4[0],ymm14[1],ymm4[2,3]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm11[2],ymm0[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm14[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1,2],ymm11[3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0,1,2],ymm14[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm11 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT: # xmm11 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT: # xmm13 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm6 = xmm6[0],mem[0]
; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
; AVX1-ONLY-NEXT: # xmm3 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm15
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm12 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %xmm14
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm12 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm15
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm11
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm2
@@ -2006,14 +2009,14 @@
; AVX1-ONLY-NEXT: vmovaps %xmm9, 1120(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm6, 816(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm10, 800(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm11, 496(%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm13, 496(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm2, 480(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm0, 176(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm13, 160(%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm11, 160(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm1, 336(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm14, 320(%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm15, 320(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm12, 656(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm15, 640(%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm14, 640(%r9)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 1216(%r9)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
@@ -4275,14 +4278,14 @@
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm14
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm14[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm14
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm13
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm13[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vmovaps 160(%rdi), %xmm12
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm12[0],mem[0]
@@ -4311,16 +4314,16 @@
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm13
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm14
+; AVX1-ONLY-NEXT: vmovaps 384(%rdi), %xmm14
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm13
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
; AVX1-ONLY-NEXT: vmovaps %xmm12, 16(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm14, (%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm13, (%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm4, 1936(%r9)
-; AVX1-ONLY-NEXT: vmovaps %xmm13, 1920(%r9)
+; AVX1-ONLY-NEXT: vmovaps %xmm14, 1920(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm0, 2256(%r9)
; AVX1-ONLY-NEXT: vmovaps %xmm5, 2240(%r9)
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
index 4ea0da9..43d48ef 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll
@@ -196,50 +196,50 @@
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm1
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm2
-; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm3
-; AVX1-ONLY-NEXT: vmovaps (%r10), %xmm4
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm2
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm3
+; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm4
+; AVX1-ONLY-NEXT: vmovaps (%r10), %xmm5
; AVX1-ONLY-NEXT: vmovaps 16(%r10), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm5
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm5 = ymm2[1],ymm5[1],ymm2[3],ymm5[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm5
-; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm6
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm6[0],xmm5[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm8
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm7, %ymm9
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm7, %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm6
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm6 = ymm3[1],ymm6[1],ymm3[3],ymm6[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm6
+; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm7
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm8, %ymm9
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm8, %ymm8
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm9
; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm10
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm10[1],xmm9[1]
; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm12
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],mem[0],ymm3[2],mem[2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm4[0],mem[0],ymm4[2],mem[2]
; AVX1-ONLY-NEXT: vmovaps 16(%rcx), %xmm11
; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm12
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm12[0],xmm11[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm12[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm5[1],xmm4[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm11[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 24(%r9), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm8[0],xmm10[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm6[1],xmm5[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm11[1],mem[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 24(%r9), %ymm5
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm10[0]
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm5 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps %ymm1, 96(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm2, 96(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm5, 16(%rax)
-; AVX1-ONLY-NEXT: vmovaps %xmm4, (%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 128(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm2, 192(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm6, 64(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 32(%rax)
+; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 128(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 192(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm7, 64(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 32(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm0, 160(%rax)
; AVX1-ONLY-NEXT: vzeroupper
; AVX1-ONLY-NEXT: retq
@@ -532,104 +532,105 @@
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: pushq %rax
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm2
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %ymm7
-; AVX1-ONLY-NEXT: vmovaps (%r9), %ymm3
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm3
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %ymm8
+; AVX1-ONLY-NEXT: vmovaps (%r9), %ymm4
; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm5
; AVX1-ONLY-NEXT: vmovaps 16(%rax), %xmm0
-; AVX1-ONLY-NEXT: vmovapd 32(%rax), %xmm8
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 32(%rax), %xmm9
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm1
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm2
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovapd 32(%r8), %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm6
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm10
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm10[1],xmm6[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm11[4,5,6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm11
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm12
-; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm9
-; AVX1-ONLY-NEXT: vmovapd 32(%r9), %xmm14
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm4 = xmm4[0],xmm14[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm4, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm10 = xmm11[0],xmm10[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rcx), %ymm10, %ymm11
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm10, %ymm6
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0],ymm11[1],ymm6[2],ymm11[2]
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm10
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm10[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm10
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm10[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm11[0,1],ymm7[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm12
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm11
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm11[6,7]
+; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm11
+; AVX1-ONLY-NEXT: vmovapd 32(%r8), %xmm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm10
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm13
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm14 = xmm13[1],xmm10[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm15
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm14
+; AVX1-ONLY-NEXT: vmovapd 32(%r9), %xmm15
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm6 = xmm6[0],xmm15[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm7[0],xmm13[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rcx), %ymm7, %ymm13
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm7, %ymm7
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm7[0],ymm13[1],ymm7[2],ymm13[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm10
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm10[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm10
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm10[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm13
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm13[6,7]
; AVX1-ONLY-NEXT: vmovaps 16(%rcx), %xmm13
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm13[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm15
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm15 = ymm3[1],ymm15[1],ymm3[3],ymm15[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm15[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm12[0],xmm9[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm15
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm12, %ymm0
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm11 = ymm4[1],ymm11[1],ymm4[3],ymm11[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm14[0,1],ymm11[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm14
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm12 = xmm12[0],xmm14[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm12, %ymm0
+; AVX1-ONLY-NEXT: vmovaps %xmm1, %xmm7
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm12, %ymm12
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm0[2,3],ymm12[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm14[1],xmm8[1]
-; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm8
-; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm14
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm8 = ymm8[0],ymm14[0],ymm8[2],ymm14[2]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm0[0,1],ymm8[2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm15[1],xmm9[1]
+; AVX1-ONLY-NEXT: vmovapd 32(%rdi), %ymm9
+; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm15
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm9[0],ymm15[0],ymm9[2],ymm15[2]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm0[0,1],ymm9[2,3]
; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm14[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm15[0,0,3,2]
; AVX1-ONLY-NEXT: vmovapd 32(%rax), %ymm1
-; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm1[2,3],ymm14[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm14[0],ymm0[1],ymm14[2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm14
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm14 = ymm14[0],mem[0],ymm14[2],mem[2]
+; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm1[2,3],ymm15[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm15[0],ymm0[1],ymm15[2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovapd 32(%r8), %ymm15
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm15[0],mem[0],ymm15[2],mem[2]
; AVX1-ONLY-NEXT: vmovapd 48(%rcx), %xmm0
-; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm4
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm4 = xmm4[0],xmm0[0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm14[2,3]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm14
-; AVX1-ONLY-NEXT: vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT: # ymm14 = ymm14[0],mem[0],ymm14[2],mem[2]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm9[1],xmm5[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm14[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm9[0],ymm3[0],ymm9[2],ymm3[2]
-; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm9
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm9 = xmm9[0],xmm13[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm9[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovapd 48(%rdx), %xmm6
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm6 = xmm6[0],xmm0[0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm15
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm15[0],ymm3[0],ymm15[2],ymm3[2]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm14[1],xmm5[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm5
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm4 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm5
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm13[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT: vbroadcastsd 56(%r9), %ymm9
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm9[2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 56(%r9), %ymm5
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm15[0],xmm10[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm9
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT: vmovaps %xmm9, 16(%rax)
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm1 = xmm7[0],xmm10[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %xmm5
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
+; AVX1-ONLY-NEXT: vmovaps %xmm5, 16(%rax)
; AVX1-ONLY-NEXT: vmovaps %xmm1, (%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm3, 128(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm5, 96(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm4, 352(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm4, 128(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm3, 96(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm6, 352(%rax)
; AVX1-ONLY-NEXT: vmovapd %ymm2, 384(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm8, 320(%rax)
+; AVX1-ONLY-NEXT: vmovapd %ymm9, 320(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm12, 32(%rax)
; AVX1-ONLY-NEXT: vmovaps %ymm11, 192(%rax)
-; AVX1-ONLY-NEXT: vmovaps %ymm7, 64(%rax)
-; AVX1-ONLY-NEXT: vmovapd %ymm6, 224(%rax)
+; AVX1-ONLY-NEXT: vmovaps %ymm8, 64(%rax)
+; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT: vmovaps %ymm1, 224(%rax)
; AVX1-ONLY-NEXT: vmovapd %ymm0, 416(%rax)
; AVX1-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX1-ONLY-NEXT: vmovaps %ymm0, 256(%rax)
@@ -2020,63 +2021,63 @@
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: subq $520, %rsp # imm = 0x208
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm5
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%rdx), %ymm0
-; AVX1-ONLY-NEXT: vmovaps (%r9), %ymm6
-; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm2[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm3
-; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm7
-; AVX1-ONLY-NEXT: vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rax), %xmm4
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm1[1],ymm5[3],ymm1[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rcx), %xmm0
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm8
+; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%rdx), %ymm1
+; AVX1-ONLY-NEXT: vmovaps (%r9), %ymm9
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm6[1],ymm1[1],ymm6[3],ymm1[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm3
+; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm2
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm4 = xmm3[0],xmm0[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm4, %ymm6
+; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm10
+; AVX1-ONLY-NEXT: vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rax), %xmm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm4, %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm6[2,3],ymm4[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm4
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm6
+; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm6[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm4[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm8[1],ymm5[1],ymm8[3],ymm5[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rcx), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm9[1],ymm2[1],ymm9[3],ymm2[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm1
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm0[0],xmm1[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rcx), %ymm2, %ymm3
; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm4
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm12
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm11
; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm2
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm12
+; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm11
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm11[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm4[1]
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
@@ -4772,63 +4773,63 @@
; AVX1-ONLY-NEXT: vmovaps (%rdi), %ymm3
; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovaps (%rdx), %ymm2
-; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm5
+; AVX1-ONLY-NEXT: vmovaps (%r8), %ymm6
; AVX1-ONLY-NEXT: vmovaps (%r9), %ymm4
-; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm6
-; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm7
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm7[0],xmm6[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm8, %ymm9
-; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm10
-; AVX1-ONLY-NEXT: vmovaps 16(%rax), %xmm11
-; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm0
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm8, %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1],ymm9[2,3],ymm8[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm8
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm2[0,1,2,3],ymm8[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm9
-; AVX1-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm9[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1],ymm8[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm8[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm8
+; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm9
+; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm5
+; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm0
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm7 = xmm9[0],xmm8[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm11
+; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm10
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm7, %ymm11
+; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm12
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm7, %ymm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm11[2,3],ymm7[4,5],ymm11[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm7
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1,2,3],ymm7[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm11
+; AVX1-ONLY-NEXT: vmovaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm11[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm7
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rax), %xmm9
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm6[1],xmm10[1]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm12[1]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
-; AVX1-ONLY-NEXT: vmovaps 16(%rcx), %xmm5
-; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm6
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm6[0],ymm4[0],ymm6[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovaps 16(%rcx), %xmm6
+; AVX1-ONLY-NEXT: vmovaps 16(%rdx), %xmm8
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm8[0],xmm6[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm3
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
+; AVX1-ONLY-NEXT: vmovaps 32(%rax), %xmm3
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm10[1],ymm1[3],ymm10[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm5[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm2
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm4[1],ymm2[1],ymm4[3],ymm2[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm6[2,3,2,3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rcx), %ymm3, %ymm4
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm7[0],xmm1[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rcx), %ymm2, %ymm4
; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
-; AVX1-ONLY-NEXT: vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm3
-; AVX1-ONLY-NEXT: vmovaps 32(%r8), %xmm4
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm4[0],xmm3[0]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm6, %ymm6
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[2]
+; AVX1-ONLY-NEXT: vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vmovaps 32(%r9), %xmm4
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm6 = xmm0[0],xmm4[0]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm6, %ymm6
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %ymm2
@@ -4836,32 +4837,32 @@
; AVX1-ONLY-NEXT: vbroadcastsd 40(%rcx), %ymm5
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %ymm5
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 32(%r8), %ymm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm0[1]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 32(%r8), %ymm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm4[1],xmm3[1]
; AVX1-ONLY-NEXT: vmovaps 32(%r9), %ymm3
; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm5[0],ymm2[2],ymm5[2]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rcx), %xmm0
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%rcx), %xmm1
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
; AVX1-ONLY-NEXT: vmovaps 48(%rdx), %xmm2
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm1[1],ymm5[3],ymm1[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%rdi), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm0[1],ymm5[3],ymm0[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
; AVX1-ONLY-NEXT: vmovaps 48(%rax), %xmm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 48(%r8), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm1[1],ymm3[3],ymm1[3]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 48(%r8), %xmm0
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%r9), %xmm0
; AVX1-ONLY-NEXT: vmovaps 64(%r8), %xmm1
@@ -4967,8 +4968,8 @@
; AVX1-ONLY-NEXT: vmovaps 128(%rdx), %ymm2
; AVX1-ONLY-NEXT: vbroadcastsd 136(%rcx), %ymm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps 128(%rsi), %xmm13
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm13[2,3,2,3]
+; AVX1-ONLY-NEXT: vmovaps 128(%rsi), %xmm14
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm14[2,3,2,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
@@ -5097,9 +5098,9 @@
; AVX1-ONLY-NEXT: vmovapd 224(%r8), %ymm2
; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%rax), %ymm2, %ymm5
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 232(%rcx), %ymm4
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 232(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 224(%r8), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovapd 224(%rdi), %ymm0
@@ -5117,17 +5118,17 @@
; AVX1-ONLY-NEXT: vbroadcastsd 216(%r9), %ymm15
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm15[4,5,6,7]
; AVX1-ONLY-NEXT: vmovapd 240(%rcx), %xmm15
-; AVX1-ONLY-NEXT: vmovapd 240(%rdx), %xmm14
-; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm14 = xmm14[0],xmm15[0]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm14[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 240(%r9), %ymm14
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm14[3]
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm14 = xmm15[1],mem[1]
+; AVX1-ONLY-NEXT: vmovapd 240(%rdx), %xmm13
+; AVX1-ONLY-NEXT: vunpcklpd {{.*#+}} xmm13 = xmm13[0],xmm15[0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm13[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT: vbroadcastsd 240(%r9), %ymm13
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm13[3]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm13 = xmm15[1],mem[1]
; AVX1-ONLY-NEXT: vbroadcastsd 248(%r9), %ymm15
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm13[0,1],ymm15[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm0[1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm14[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm12[0],xmm13[0]
+; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm13[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm13 = xmm12[0],xmm14[0]
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm14 = xmm10[0],xmm11[0]
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm8 = xmm9[0],xmm8[0]
; AVX1-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
@@ -10432,48 +10433,48 @@
; AVX1-ONLY: # %bb.0:
; AVX1-ONLY-NEXT: subq $3832, %rsp # imm = 0xEF8
; AVX1-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm4
-; AVX1-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %ymm6
+; AVX1-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps (%rdx), %ymm0
-; AVX1-ONLY-NEXT: vmovaps (%r9), %ymm5
-; AVX1-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm2
-; AVX1-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm1
-; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm2[0]
-; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm3
-; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm6
-; AVX1-ONLY-NEXT: vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm2, %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT: vmovaps (%r9), %ymm7
+; AVX1-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%r9), %xmm1
+; AVX1-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps (%r8), %xmm2
+; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
+; AVX1-ONLY-NEXT: vmovaps (%rdi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm4
+; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm3, %ymm5
+; AVX1-ONLY-NEXT: vmovaps (%rax), %xmm8
+; AVX1-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm3, %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3],ymm3[4,5],ymm5[6,7]
+; AVX1-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm3
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm5
+; AVX1-ONLY-NEXT: vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[2,3,2,3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1],ymm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vbroadcastsd 8(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT: vmovaps (%rsi), %xmm3
-; AVX1-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vmovaps 16(%rax), %xmm1
-; AVX1-ONLY-NEXT: vmovaps 16(%rdi), %xmm2
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm4[1],ymm2[1],ymm4[3],ymm2[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 16(%rax), %xmm2
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm6[1],ymm4[1],ymm6[3],ymm4[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 16(%rcx), %xmm0
; AVX1-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm1
-; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm5[1],ymm1[1],ymm5[3],ymm1[3]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT: vmovaps 16(%r8), %xmm2
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm7[1],ymm2[1],ymm7[3],ymm2[3]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 32(%rsi), %xmm0
-; AVX1-ONLY-NEXT: vmovaps 32(%rdi), %xmm1
; AVX1-ONLY-NEXT: vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
; AVX1-ONLY-NEXT: vinsertf128 $1, 32(%rcx), %ymm2, %ymm3
; AVX1-ONLY-NEXT: vmovaps 32(%rdx), %xmm4
@@ -10995,9 +10996,9 @@
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2],ymm3[2]
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT: vbroadcastsd 488(%rcx), %ymm2
-; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm0[1]
+; AVX1-ONLY-NEXT: vbroadcastsd 488(%rcx), %ymm1
+; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vinsertf128 $1, 480(%r8), %ymm0, %ymm0
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
index cbd0e20..4667d2e 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
@@ -1415,73 +1415,73 @@
; AVX2-FAST-PERLANE-NEXT: movq {{[0-9]+}}(%rsp), %r10
; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdi), %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rsi), %xmm1
-; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdx), %xmm4
-; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rcx), %xmm5
-; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r8), %xmm2
-; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r9), %xmm3
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm5, %ymm4, %ymm7
-; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm3, %ymm2, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm8 = zero,zero,zero,ymm6[5],zero,zero,zero,zero,zero,zero,ymm6[6],zero,zero,zero,zero,zero,zero,zero,ymm6[23],zero,zero,zero,zero,zero,zero,ymm6[24],zero,zero,zero,zero,zero,zero
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm9 = ymm6[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,zero,ymm9[5],zero,zero,zero,zero,zero,zero,ymm9[6],zero,zero,zero,zero,zero,ymm9[23],zero,zero,zero,zero,zero,zero,ymm9[24],zero,zero,zero,zero,zero,zero,ymm9[25]
-; AVX2-FAST-PERLANE-NEXT: vpor %ymm8, %ymm9, %ymm8
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,zero,zero,ymm7[5],zero,zero,zero,zero,zero,zero,ymm7[6],zero,zero,zero,zero,zero,zero,zero,ymm7[23],zero,zero,zero,zero,zero,zero,ymm7[24],zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rdx), %xmm5
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%rcx), %xmm6
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r8), %xmm3
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r9), %xmm4
+; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r10), %xmm2
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm6, %ymm5, %ymm8
+; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm4, %ymm3, %ymm10
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm9 = zero,zero,zero,ymm7[5],zero,zero,zero,zero,zero,zero,ymm7[6],zero,zero,zero,zero,zero,zero,zero,ymm7[23],zero,zero,zero,zero,zero,zero,ymm7[24],zero,zero,zero,zero,zero,zero
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm11 = ymm7[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm11 = zero,zero,zero,zero,zero,zero,ymm11[5],zero,zero,zero,zero,zero,zero,ymm11[6],zero,zero,zero,zero,zero,ymm11[23],zero,zero,zero,zero,zero,zero,ymm11[24],zero,zero,zero,zero,zero
-; AVX2-FAST-PERLANE-NEXT: vpor %ymm11, %ymm9, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm11 = <u,u,u,255,255,0,0,u,u,u,255,255,0,0,u,u,u,255,255,0,0,u,u,u,255,255,0,0,u,u,u,255>
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm11, %ymm8, %ymm9, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm8 = ymm10[4],zero,zero,zero,zero,zero,zero,ymm10[5],zero,zero,zero,zero,zero,zero,ymm10[6],zero,zero,zero,zero,zero,zero,zero,ymm10[23],zero,zero,zero,zero,zero,zero,ymm10[24],zero,zero
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm11 = ymm10[2,3,0,1]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm11 = zero,ymm11[4],zero,zero,zero,zero,zero,zero,ymm11[5],zero,zero,zero,zero,zero,zero,ymm11[6],zero,zero,zero,zero,zero,ymm11[23],zero,zero,zero,zero,zero,zero,ymm11[24],zero,zero,zero
-; AVX2-FAST-PERLANE-NEXT: vpor %ymm11, %ymm8, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vmovdqa (%r10), %xmm8
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm12 = xmm8[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm11 = zero,zero,zero,zero,ymm11[5],zero,zero,zero,zero,zero,zero,ymm11[6],zero,zero,zero,zero,zero,ymm11[23],zero,zero,zero,zero,zero,zero,ymm11[24],zero,zero,zero,zero,zero,zero,ymm11[25]
+; AVX2-FAST-PERLANE-NEXT: vpor %ymm9, %ymm11, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm11 = zero,zero,zero,zero,zero,ymm8[5],zero,zero,zero,zero,zero,zero,ymm8[6],zero,zero,zero,zero,zero,zero,zero,ymm8[23],zero,zero,zero,zero,zero,zero,ymm8[24],zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm8[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,zero,zero,zero,zero,ymm12[5],zero,zero,zero,zero,zero,zero,ymm12[6],zero,zero,zero,zero,zero,ymm12[23],zero,zero,zero,zero,zero,zero,ymm12[24],zero,zero,zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpor %ymm12, %ymm11, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm12 = <u,u,u,255,255,0,0,u,u,u,255,255,0,0,u,u,u,255,255,0,0,u,u,u,255,255,0,0,u,u,u,255>
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm12, %ymm9, %ymm11, %ymm9
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm11 = ymm10[4],zero,zero,zero,zero,zero,zero,ymm10[5],zero,zero,zero,zero,zero,zero,ymm10[6],zero,zero,zero,zero,zero,zero,zero,ymm10[23],zero,zero,zero,zero,zero,zero,ymm10[24],zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm10[2,3,0,1]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm12 = zero,ymm12[4],zero,zero,zero,zero,zero,zero,ymm12[5],zero,zero,zero,zero,zero,zero,ymm12[6],zero,zero,zero,zero,zero,ymm12[23],zero,zero,zero,zero,zero,zero,ymm12[24],zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpor %ymm12, %ymm11, %ymm11
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm12 = xmm2[4,5,4,5,4,5,8,9,6,7,6,7,6,7,6,7]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm12[0,0,1,0]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm13 = <255,255,0,u,u,u,u,255,255,0,u,u,u,u,255,255,0,u,u,u,u,255,255,0,u,u,u,u,255,255,0,u>
; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm11, %ymm12, %ymm11
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm12 = [0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255]
; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm12, %ymm9, %ymm11, %ymm9
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm11 = xmm8[2,3,2,3,0,1,0,1,8,9,10,11,2,3,2,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm11 = xmm2[2,3,2,3,0,1,0,1,8,9,10,11,2,3,2,3]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm11 = ymm11[0,0,1,0]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm10[0,2,0,2]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm12 = ymm12[u,u,u,u,0,8,u,u,u,u,u,1,9,u,u,u,u,u,18,26,u,u,u,u,u,19,27,u,u,u,u,u]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm13 = <u,u,u,u,255,255,0,u,u,u,u,255,255,0,u,u,u,u,255,255,0,u,u,u,u,255,255,0,u,u,u,u>
; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm12, %ymm11, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm7[0,2,0,2]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm8[0,2,0,2]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm12 = zero,zero,ymm12[0,8],zero,zero,zero,zero,zero,ymm12[1,9],zero,zero,zero,zero,zero,ymm12[18,26],zero,zero,zero,zero,zero,ymm12[19,27],zero,zero,zero,zero,zero,ymm12[20,28]
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm13 = ymm6[0,2,0,2]
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm13 = ymm7[0,2,0,2]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm13 = ymm13[0,8],zero,zero,zero,zero,zero,ymm13[1,9],zero,zero,zero,zero,zero,ymm13[2,10],zero,zero,zero,zero,zero,ymm13[19,27],zero,zero,zero,zero,zero,ymm13[20,28],zero,zero
; AVX2-FAST-PERLANE-NEXT: vpor %ymm12, %ymm13, %ymm12
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm13 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm12, %ymm11, %ymm11
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm12 = xmm8[8,9,10,11,8,9,10,11,10,11,12,13,10,11,12,13]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm12 = xmm2[8,9,10,11,8,9,10,11,10,11,12,13,10,11,12,13]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm12 = ymm12[0,1,0,1]
; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm10 = ymm10[1,3,1,3]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm10 = ymm10[u,u,u,1,9,u,u,u,u,u,2,10,u,u,u,u,u,19,27,u,u,u,u,u,20,28,u,u,u,u,u,21]
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm13 = <u,u,u,255,255,0,u,u,u,u,255,255,0,u,u,u,u,255,255,0,u,u,u,u,255,255,0,u,u,u,u,255>
; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm13, %ymm10, %ymm12, %ymm10
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm6 = ymm6[3,1,1,3]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm6 = ymm6[1],zero,zero,zero,zero,zero,ymm6[10,2],zero,zero,zero,zero,zero,ymm6[11,3],zero,zero,zero,zero,zero,ymm6[20,28],zero,zero,zero,zero,zero,ymm6[21,29],zero,zero,zero
-; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm7[1,3,3,1]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm7 = zero,ymm7[1,9],zero,zero,zero,zero,zero,ymm7[2,10],zero,zero,zero,zero,zero,ymm7[3,19],zero,zero,zero,zero,zero,ymm7[28,20],zero,zero,zero,zero,zero,ymm7[29,21],zero
-; AVX2-FAST-PERLANE-NEXT: vpor %ymm6, %ymm7, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm7 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0]
-; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm7, %ymm6, %ymm10, %ymm6
-; AVX2-FAST-PERLANE-NEXT: vpunpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm4 = zero,zero,zero,zero,xmm4[12,13],zero,zero,zero,zero,zero,xmm4[14,15],zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm7 = ymm7[3,1,1,3]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm7 = ymm7[1],zero,zero,zero,zero,zero,ymm7[10,2],zero,zero,zero,zero,zero,ymm7[11,3],zero,zero,zero,zero,zero,ymm7[20,28],zero,zero,zero,zero,zero,ymm7[21,29],zero,zero,zero
+; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm8 = ymm8[1,3,3,1]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm8 = zero,ymm8[1,9],zero,zero,zero,zero,zero,ymm8[2,10],zero,zero,zero,zero,zero,ymm8[3,19],zero,zero,zero,zero,zero,ymm8[28,20],zero,zero,zero,zero,zero,ymm8[29,21],zero
+; AVX2-FAST-PERLANE-NEXT: vpor %ymm7, %ymm8, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm8 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0]
+; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm8, %ymm7, %ymm10, %ymm7
+; AVX2-FAST-PERLANE-NEXT: vpunpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm5 = zero,zero,zero,zero,xmm5[12,13],zero,zero,zero,zero,zero,xmm5[14,15],zero,zero,zero
; AVX2-FAST-PERLANE-NEXT: vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm0 = zero,zero,xmm0[12,13],zero,zero,zero,zero,zero,xmm0[14,15],zero,zero,zero,zero,zero
-; AVX2-FAST-PERLANE-NEXT: vpor %xmm4, %xmm0, %xmm0
-; AVX2-FAST-PERLANE-NEXT: vpunpckhbw {{.*#+}} xmm1 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
+; AVX2-FAST-PERLANE-NEXT: vpor %xmm5, %xmm0, %xmm0
+; AVX2-FAST-PERLANE-NEXT: vpunpckhbw {{.*#+}} xmm1 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[10],zero,zero,zero,zero,zero,xmm1[13,12],zero,zero,zero,zero,zero,xmm1[15,14],zero
-; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm2 = zero,xmm8[13,14,15,4,5],zero,zero,xmm8[14,15,14,15,12],zero,zero,xmm8[15]
+; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm2 = zero,xmm2[13,14,15,4,5],zero,zero,xmm2[14,15,14,15,12],zero,zero,xmm2[15]
; AVX2-FAST-PERLANE-NEXT: vpor %xmm2, %xmm1, %xmm1
; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} xmm2 = [0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0]
; AVX2-FAST-PERLANE-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0
; AVX2-FAST-PERLANE-NEXT: vmovdqa %xmm0, 96(%rax)
-; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm6, 64(%rax)
+; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm7, 64(%rax)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm11, (%rax)
; AVX2-FAST-PERLANE-NEXT: vmovdqa %ymm9, 32(%rax)
; AVX2-FAST-PERLANE-NEXT: vzeroupper
diff --git a/llvm/test/CodeGen/X86/vector-reduce-and-bool.ll b/llvm/test/CodeGen/X86/vector-reduce-and-bool.ll
index 600326b..bcb3f4d 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-and-bool.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-and-bool.ll
@@ -70,30 +70,25 @@
; SSE41-NEXT: setb %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: trunc_v4i32_v4i1:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: setb %al
-; AVX1OR2-NEXT: retq
+; AVX1-LABEL: trunc_v4i32_v4i1:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX1-NEXT: setb %al
+; AVX1-NEXT: retq
;
-; AVX512F-LABEL: trunc_v4i32_v4i1:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512F-NEXT: setb %al
-; AVX512F-NEXT: retq
+; AVX2-LABEL: trunc_v4i32_v4i1:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; AVX2-NEXT: vptest %xmm1, %xmm0
+; AVX2-NEXT: setb %al
+; AVX2-NEXT: retq
;
-; AVX512BW-LABEL: trunc_v4i32_v4i1:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512BW-NEXT: setb %al
-; AVX512BW-NEXT: retq
-;
-; AVX512VL-LABEL: trunc_v4i32_v4i1:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [4294967297,4294967297]
-; AVX512VL-NEXT: vptest %xmm1, %xmm0
-; AVX512VL-NEXT: setb %al
-; AVX512VL-NEXT: retq
+; AVX512-LABEL: trunc_v4i32_v4i1:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; AVX512-NEXT: vptest %xmm1, %xmm0
+; AVX512-NEXT: setb %al
+; AVX512-NEXT: retq
%a = trunc <4 x i32> %0 to <4 x i1>
%b = call i1 @llvm.vector.reduce.and.v4i1(<4 x i1> %a)
ret i1 %b
@@ -115,30 +110,11 @@
; SSE41-NEXT: setb %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: trunc_v8i16_v8i1:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: setb %al
-; AVX1OR2-NEXT: retq
-;
-; AVX512F-LABEL: trunc_v8i16_v8i1:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512F-NEXT: setb %al
-; AVX512F-NEXT: retq
-;
-; AVX512BW-LABEL: trunc_v8i16_v8i1:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512BW-NEXT: setb %al
-; AVX512BW-NEXT: retq
-;
-; AVX512VL-LABEL: trunc_v8i16_v8i1:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [281479271743489,281479271743489]
-; AVX512VL-NEXT: vptest %xmm1, %xmm0
-; AVX512VL-NEXT: setb %al
-; AVX512VL-NEXT: retq
+; AVX-LABEL: trunc_v8i16_v8i1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX-NEXT: setb %al
+; AVX-NEXT: retq
%a = trunc <8 x i16> %0 to <8 x i1>
%b = call i1 @llvm.vector.reduce.and.v8i1(<8 x i1> %a)
ret i1 %b
@@ -159,30 +135,11 @@
; SSE41-NEXT: setb %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: trunc_v16i8_v16i1:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: setb %al
-; AVX1OR2-NEXT: retq
-;
-; AVX512F-LABEL: trunc_v16i8_v16i1:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512F-NEXT: setb %al
-; AVX512F-NEXT: retq
-;
-; AVX512BW-LABEL: trunc_v16i8_v16i1:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512BW-NEXT: setb %al
-; AVX512BW-NEXT: retq
-;
-; AVX512VL-LABEL: trunc_v16i8_v16i1:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [72340172838076673,72340172838076673]
-; AVX512VL-NEXT: vptest %xmm1, %xmm0
-; AVX512VL-NEXT: setb %al
-; AVX512VL-NEXT: retq
+; AVX-LABEL: trunc_v16i8_v16i1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX-NEXT: setb %al
+; AVX-NEXT: retq
%a = trunc <16 x i8> %0 to <16 x i1>
%b = call i1 @llvm.vector.reduce.and.v16i1(<16 x i1> %a)
ret i1 %b
@@ -258,7 +215,7 @@
;
; AVX2-LABEL: trunc_v8i32_v8i1:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setb %al
; AVX2-NEXT: vzeroupper
@@ -266,7 +223,7 @@
;
; AVX512-LABEL: trunc_v8i32_v8i1:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX512-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX512-NEXT: vptest %ymm1, %ymm0
; AVX512-NEXT: setb %al
; AVX512-NEXT: vzeroupper
@@ -294,28 +251,12 @@
; SSE41-NEXT: setb %al
; SSE41-NEXT: retq
;
-; AVX1-LABEL: trunc_v16i16_v16i1:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
-; AVX1-NEXT: setb %al
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: trunc_v16i16_v16i1:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX2-NEXT: vptest %ymm1, %ymm0
-; AVX2-NEXT: setb %al
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
-;
-; AVX512-LABEL: trunc_v16i16_v16i1:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX512-NEXT: vptest %ymm1, %ymm0
-; AVX512-NEXT: setb %al
-; AVX512-NEXT: vzeroupper
-; AVX512-NEXT: retq
+; AVX-LABEL: trunc_v16i16_v16i1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
+; AVX-NEXT: setb %al
+; AVX-NEXT: vzeroupper
+; AVX-NEXT: retq
%a = trunc <16 x i16> %0 to <16 x i1>
%b = call i1 @llvm.vector.reduce.and.v16i1(<16 x i1> %a)
ret i1 %b
@@ -338,28 +279,12 @@
; SSE41-NEXT: setb %al
; SSE41-NEXT: retq
;
-; AVX1-LABEL: trunc_v32i8_v32i1:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
-; AVX1-NEXT: setb %al
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: trunc_v32i8_v32i1:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX2-NEXT: vptest %ymm1, %ymm0
-; AVX2-NEXT: setb %al
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
-;
-; AVX512-LABEL: trunc_v32i8_v32i1:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX512-NEXT: vptest %ymm1, %ymm0
-; AVX512-NEXT: setb %al
-; AVX512-NEXT: vzeroupper
-; AVX512-NEXT: retq
+; AVX-LABEL: trunc_v32i8_v32i1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
+; AVX-NEXT: setb %al
+; AVX-NEXT: vzeroupper
+; AVX-NEXT: retq
%a = trunc <32 x i8> %0 to <32 x i1>
%b = call i1 @llvm.vector.reduce.and.v32i1(<32 x i1> %a)
ret i1 %b
@@ -458,7 +383,7 @@
; AVX2-LABEL: trunc_v16i32_v16i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setb %al
; AVX2-NEXT: vzeroupper
@@ -509,8 +434,7 @@
; AVX2-LABEL: trunc_v32i16_v32i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX2-NEXT: vptest %ymm1, %ymm0
+; AVX2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX2-NEXT: setb %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
@@ -581,8 +505,7 @@
; AVX2-LABEL: trunc_v64i8_v64i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX2-NEXT: vptest %ymm1, %ymm0
+; AVX2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX2-NEXT: setb %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
@@ -1050,9 +973,8 @@
; SSE2: # %bb.0:
; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
; SSE2-NEXT: psllw $15, %xmm0
-; SSE2-NEXT: psraw $15, %xmm0
; SSE2-NEXT: pmovmskb %xmm0, %eax
-; SSE2-NEXT: testl %eax, %eax
+; SSE2-NEXT: testl $43690, %eax # imm = 0xAAAA
; SSE2-NEXT: sete %al
; SSE2-NEXT: retq
;
@@ -1060,9 +982,8 @@
; SSE41: # %bb.0:
; SSE41-NEXT: pmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; SSE41-NEXT: psllw $15, %xmm0
-; SSE41-NEXT: psraw $15, %xmm0
; SSE41-NEXT: pmovmskb %xmm0, %eax
-; SSE41-NEXT: testl %eax, %eax
+; SSE41-NEXT: testl $43690, %eax # imm = 0xAAAA
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
@@ -1070,9 +991,8 @@
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX1OR2-NEXT: vpsllw $15, %xmm0, %xmm0
-; AVX1OR2-NEXT: vpsraw $15, %xmm0, %xmm0
; AVX1OR2-NEXT: vpmovmskb %xmm0, %eax
-; AVX1OR2-NEXT: testl %eax, %eax
+; AVX1OR2-NEXT: testl $43690, %eax # imm = 0xAAAA
; AVX1OR2-NEXT: sete %al
; AVX1OR2-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-reduce-and.ll b/llvm/test/CodeGen/X86/vector-reduce-and.ll
index b85696a..a18f58a 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-and.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-and.ll
@@ -489,11 +489,11 @@
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX1-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-NEXT: vzeroupper
@@ -504,11 +504,11 @@
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX2-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
@@ -519,11 +519,11 @@
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX512-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
@@ -555,11 +555,11 @@
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-NEXT: vandps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX1-NEXT: vandps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-NEXT: vzeroupper
@@ -571,11 +571,11 @@
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX2-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
@@ -588,11 +588,11 @@
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX512-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
@@ -630,11 +630,11 @@
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-NEXT: vandps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX1-NEXT: vandps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-NEXT: vzeroupper
@@ -648,11 +648,11 @@
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX2-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
@@ -666,11 +666,11 @@
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX512-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpand %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
diff --git a/llvm/test/CodeGen/X86/vector-reduce-or-bool.ll b/llvm/test/CodeGen/X86/vector-reduce-or-bool.ll
index 2ccb6ac..feb194e7 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-or-bool.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-or-bool.ll
@@ -70,30 +70,25 @@
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: trunc_v4i32_v4i1:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: setne %al
-; AVX1OR2-NEXT: retq
+; AVX1-LABEL: trunc_v4i32_v4i1:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX1-NEXT: setne %al
+; AVX1-NEXT: retq
;
-; AVX512F-LABEL: trunc_v4i32_v4i1:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512F-NEXT: setne %al
-; AVX512F-NEXT: retq
+; AVX2-LABEL: trunc_v4i32_v4i1:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; AVX2-NEXT: vptest %xmm1, %xmm0
+; AVX2-NEXT: setne %al
+; AVX2-NEXT: retq
;
-; AVX512BW-LABEL: trunc_v4i32_v4i1:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512BW-NEXT: setne %al
-; AVX512BW-NEXT: retq
-;
-; AVX512VL-LABEL: trunc_v4i32_v4i1:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [4294967297,4294967297]
-; AVX512VL-NEXT: vptest %xmm1, %xmm0
-; AVX512VL-NEXT: setne %al
-; AVX512VL-NEXT: retq
+; AVX512-LABEL: trunc_v4i32_v4i1:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; AVX512-NEXT: vptest %xmm1, %xmm0
+; AVX512-NEXT: setne %al
+; AVX512-NEXT: retq
%a = trunc <4 x i32> %0 to <4 x i1>
%b = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> %a)
ret i1 %b
@@ -114,30 +109,11 @@
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: trunc_v8i16_v8i1:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: setne %al
-; AVX1OR2-NEXT: retq
-;
-; AVX512F-LABEL: trunc_v8i16_v8i1:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512F-NEXT: setne %al
-; AVX512F-NEXT: retq
-;
-; AVX512BW-LABEL: trunc_v8i16_v8i1:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512BW-NEXT: setne %al
-; AVX512BW-NEXT: retq
-;
-; AVX512VL-LABEL: trunc_v8i16_v8i1:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [281479271743489,281479271743489]
-; AVX512VL-NEXT: vptest %xmm1, %xmm0
-; AVX512VL-NEXT: setne %al
-; AVX512VL-NEXT: retq
+; AVX-LABEL: trunc_v8i16_v8i1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX-NEXT: setne %al
+; AVX-NEXT: retq
%a = trunc <8 x i16> %0 to <8 x i1>
%b = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> %a)
ret i1 %b
@@ -158,30 +134,11 @@
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: trunc_v16i8_v16i1:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: setne %al
-; AVX1OR2-NEXT: retq
-;
-; AVX512F-LABEL: trunc_v16i8_v16i1:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512F-NEXT: setne %al
-; AVX512F-NEXT: retq
-;
-; AVX512BW-LABEL: trunc_v16i8_v16i1:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512BW-NEXT: setne %al
-; AVX512BW-NEXT: retq
-;
-; AVX512VL-LABEL: trunc_v16i8_v16i1:
-; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [72340172838076673,72340172838076673]
-; AVX512VL-NEXT: vptest %xmm1, %xmm0
-; AVX512VL-NEXT: setne %al
-; AVX512VL-NEXT: retq
+; AVX-LABEL: trunc_v16i8_v16i1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX-NEXT: setne %al
+; AVX-NEXT: retq
%a = trunc <16 x i8> %0 to <16 x i1>
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
@@ -257,7 +214,7 @@
;
; AVX2-LABEL: trunc_v8i32_v8i1:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
@@ -265,7 +222,7 @@
;
; AVX512-LABEL: trunc_v8i32_v8i1:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX512-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX512-NEXT: vptest %ymm1, %ymm0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
@@ -292,28 +249,12 @@
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
-; AVX1-LABEL: trunc_v16i16_v16i1:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
-; AVX1-NEXT: setne %al
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: trunc_v16i16_v16i1:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX2-NEXT: vptest %ymm1, %ymm0
-; AVX2-NEXT: setne %al
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
-;
-; AVX512-LABEL: trunc_v16i16_v16i1:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX512-NEXT: vptest %ymm1, %ymm0
-; AVX512-NEXT: setne %al
-; AVX512-NEXT: vzeroupper
-; AVX512-NEXT: retq
+; AVX-LABEL: trunc_v16i16_v16i1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
+; AVX-NEXT: setne %al
+; AVX-NEXT: vzeroupper
+; AVX-NEXT: retq
%a = trunc <16 x i16> %0 to <16 x i1>
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
@@ -336,28 +277,12 @@
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
-; AVX1-LABEL: trunc_v32i8_v32i1:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
-; AVX1-NEXT: setne %al
-; AVX1-NEXT: vzeroupper
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: trunc_v32i8_v32i1:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX2-NEXT: vptest %ymm1, %ymm0
-; AVX2-NEXT: setne %al
-; AVX2-NEXT: vzeroupper
-; AVX2-NEXT: retq
-;
-; AVX512-LABEL: trunc_v32i8_v32i1:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX512-NEXT: vptest %ymm1, %ymm0
-; AVX512-NEXT: setne %al
-; AVX512-NEXT: vzeroupper
-; AVX512-NEXT: retq
+; AVX-LABEL: trunc_v32i8_v32i1:
+; AVX: # %bb.0:
+; AVX-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
+; AVX-NEXT: setne %al
+; AVX-NEXT: vzeroupper
+; AVX-NEXT: retq
%a = trunc <32 x i8> %0 to <32 x i1>
%b = call i1 @llvm.vector.reduce.or.v32i1(<32 x i1> %a)
ret i1 %b
@@ -454,7 +379,7 @@
; AVX2-LABEL: trunc_v16i32_v16i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
@@ -504,8 +429,7 @@
; AVX2-LABEL: trunc_v32i16_v32i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
-; AVX2-NEXT: vptest %ymm1, %ymm0
+; AVX2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
@@ -554,8 +478,7 @@
; AVX2-LABEL: trunc_v64i8_v64i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
-; AVX2-NEXT: vptest %ymm1, %ymm0
+; AVX2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
@@ -2182,5 +2105,3 @@
declare i1 @llvm.vector.reduce.or.v16i1(<16 x i1>)
declare i1 @llvm.vector.reduce.or.v32i1(<32 x i1>)
declare i1 @llvm.vector.reduce.or.v64i1(<64 x i1>)
-;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
-; AVX: {{.*}}
diff --git a/llvm/test/CodeGen/X86/vector-reduce-or-cmp.ll b/llvm/test/CodeGen/X86/vector-reduce-or-cmp.ll
index f8ba00b..dfe163a 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-or-cmp.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-or-cmp.ll
@@ -884,7 +884,7 @@
;
; AVX2-LABEL: mask_v8i32:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [9223372039002259456,9223372039002259456,9223372039002259456,9223372039002259456]
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm1 = [2147483648,2147483648,2147483648,2147483648,2147483648,2147483648,2147483648,2147483648]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: sete %al
; AVX2-NEXT: vzeroupper
@@ -892,7 +892,7 @@
;
; AVX512-LABEL: mask_v8i32:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [9223372039002259456,9223372039002259456,9223372039002259456,9223372039002259456]
+; AVX512-NEXT: vpbroadcastd {{.*#+}} ymm1 = [2147483648,2147483648,2147483648,2147483648,2147483648,2147483648,2147483648,2147483648]
; AVX512-NEXT: vptest %ymm1, %ymm0
; AVX512-NEXT: sete %al
; AVX512-NEXT: vzeroupper
@@ -1018,34 +1018,28 @@
; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
-; AVX1OR2-LABEL: PR44781:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vmovdqu (%rdi), %xmm0
-; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX1OR2-NEXT: sete %al
-; AVX1OR2-NEXT: retq
+; AVX1-LABEL: PR44781:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vmovdqu (%rdi), %xmm0
+; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; AVX1-NEXT: sete %al
+; AVX1-NEXT: retq
;
-; AVX512F-LABEL: PR44781:
-; AVX512F: # %bb.0:
-; AVX512F-NEXT: vmovdqu (%rdi), %xmm0
-; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512F-NEXT: sete %al
-; AVX512F-NEXT: retq
+; AVX2-LABEL: PR44781:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vmovdqu (%rdi), %xmm0
+; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [15,15,15,15]
+; AVX2-NEXT: vptest %xmm1, %xmm0
+; AVX2-NEXT: sete %al
+; AVX2-NEXT: retq
;
-; AVX512BW-LABEL: PR44781:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vmovdqu (%rdi), %xmm0
-; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; AVX512BW-NEXT: sete %al
-; AVX512BW-NEXT: retq
-;
-; AVX512BWVL-LABEL: PR44781:
-; AVX512BWVL: # %bb.0:
-; AVX512BWVL-NEXT: vmovdqu (%rdi), %xmm0
-; AVX512BWVL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [64424509455,64424509455]
-; AVX512BWVL-NEXT: vptest %xmm1, %xmm0
-; AVX512BWVL-NEXT: sete %al
-; AVX512BWVL-NEXT: retq
+; AVX512-LABEL: PR44781:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vmovdqu (%rdi), %xmm0
+; AVX512-NEXT: vpbroadcastd {{.*#+}} xmm1 = [15,15,15,15]
+; AVX512-NEXT: vptest %xmm1, %xmm0
+; AVX512-NEXT: sete %al
+; AVX512-NEXT: retq
%2 = load <4 x i32>, ptr %0, align 4
%3 = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> %2)
%4 = and i32 %3, 15
diff --git a/llvm/test/CodeGen/X86/vector-reduce-or.ll b/llvm/test/CodeGen/X86/vector-reduce-or.ll
index 076392a..ffd87aa 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-or.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-or.ll
@@ -489,11 +489,11 @@
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX1-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-NEXT: vzeroupper
@@ -504,11 +504,11 @@
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX2-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
@@ -519,11 +519,11 @@
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
@@ -555,11 +555,11 @@
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-NEXT: vorps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX1-NEXT: vorps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-NEXT: vzeroupper
@@ -571,11 +571,11 @@
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX2-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
@@ -588,11 +588,11 @@
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
@@ -630,11 +630,11 @@
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-NEXT: vorps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX1-NEXT: vorps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-NEXT: vzeroupper
@@ -648,11 +648,11 @@
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX2-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
@@ -666,11 +666,11 @@
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
diff --git a/llvm/test/CodeGen/X86/vector-reduce-xor.ll b/llvm/test/CodeGen/X86/vector-reduce-xor.ll
index 737c992..1546282 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-xor.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-xor.ll
@@ -489,11 +489,11 @@
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX1-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-NEXT: vzeroupper
@@ -504,11 +504,11 @@
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX2-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
@@ -519,11 +519,11 @@
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX512-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
@@ -555,11 +555,11 @@
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vxorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-NEXT: vxorps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vxorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX1-NEXT: vxorps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vxorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-NEXT: vzeroupper
@@ -571,11 +571,11 @@
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX2-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
@@ -588,11 +588,11 @@
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX512-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
@@ -630,11 +630,11 @@
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vxorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-NEXT: vxorps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vxorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX1-NEXT: vxorps %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vxorps %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX1-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-NEXT: vzeroupper
@@ -648,11 +648,11 @@
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX2-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX2-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX2-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
@@ -666,11 +666,11 @@
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX512-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
-; AVX512-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpxor %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
diff --git a/llvm/test/CodeGen/X86/vector-replicaton-i1-mask.ll b/llvm/test/CodeGen/X86/vector-replicaton-i1-mask.ll
index 7d3c0ab..5676836 100644
--- a/llvm/test/CodeGen/X86/vector-replicaton-i1-mask.ll
+++ b/llvm/test/CodeGen/X86/vector-replicaton-i1-mask.ll
@@ -790,27 +790,47 @@
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
-; AVX512BW-LABEL: mask_replication_factor3_vf16:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: kmovw (%rdi), %k1
-; AVX512BW-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,0,0,1,1,1,2,2,2,3,3,3,4,4,4,5]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm1, %zmm1
-; AVX512BW-NEXT: vptestmd %zmm1, %zmm1, %k1
-; AVX512BW-NEXT: vmovdqa32 (%rsi), %zmm1 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm2 = [10,11,11,11,12,12,12,13,13,13,14,14,14,15,15,15]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm2, %zmm2
-; AVX512BW-NEXT: vptestmd %zmm2, %zmm2, %k1
-; AVX512BW-NEXT: vmovdqa32 128(%rsi), %zmm2 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm3 = [5,5,6,6,6,7,7,7,8,8,8,9,9,9,10,10]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm3, %zmm0
-; AVX512BW-NEXT: vptestmd %zmm0, %zmm0, %k1
-; AVX512BW-NEXT: vmovdqa32 64(%rsi), %zmm0 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 %zmm0, 64(%rdx)
-; AVX512BW-NEXT: vmovdqa64 %zmm2, 128(%rdx)
-; AVX512BW-NEXT: vmovdqa64 %zmm1, (%rdx)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
+; AVX512BW-ONLY-LABEL: mask_replication_factor3_vf16:
+; AVX512BW-ONLY: # %bb.0:
+; AVX512BW-ONLY-NEXT: kmovw (%rdi), %k1
+; AVX512BW-ONLY-NEXT: vpmovm2b %k1, %zmm0
+; AVX512BW-ONLY-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,0,1,0,1,0,1]
+; AVX512BW-ONLY-NEXT: vpshufb {{.*#+}} zmm0 = zmm0[0,0,0,1,1,1,2,2,2,3,3,3,4,4,4,5,21,21,22,22,22,23,23,23,24,24,24,25,25,25,26,26,42,43,43,43,44,44,44,45,45,45,46,46,46,47,47,47,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512BW-ONLY-NEXT: vpmovb2m %zmm0, %k2
+; AVX512BW-ONLY-NEXT: vmovdqa32 (%rsi), %zmm0 {%k2} {z}
+; AVX512BW-ONLY-NEXT: vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm2 = [10,11,11,11,12,12,12,13,13,13,14,14,14,15,15,15]
+; AVX512BW-ONLY-NEXT: vpermd %zmm1, %zmm2, %zmm1
+; AVX512BW-ONLY-NEXT: vptestmd %zmm1, %zmm1, %k1
+; AVX512BW-ONLY-NEXT: vmovdqa32 128(%rsi), %zmm1 {%k1} {z}
+; AVX512BW-ONLY-NEXT: kshiftrd $16, %k2, %k1
+; AVX512BW-ONLY-NEXT: vmovdqa32 64(%rsi), %zmm2 {%k1} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm2, 64(%rdx)
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm1, 128(%rdx)
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm0, (%rdx)
+; AVX512BW-ONLY-NEXT: vzeroupper
+; AVX512BW-ONLY-NEXT: retq
+;
+; AVX512VBMI-ONLY-LABEL: mask_replication_factor3_vf16:
+; AVX512VBMI-ONLY: # %bb.0:
+; AVX512VBMI-ONLY-NEXT: kmovw (%rdi), %k1
+; AVX512VBMI-ONLY-NEXT: vpmovm2b %k1, %zmm0
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm1 = <0,0,0,1,1,1,2,2,2,3,3,3,4,4,4,5,5,5,6,6,6,7,7,7,8,8,8,9,9,9,10,10,10,11,11,11,12,12,12,13,13,13,14,14,14,15,15,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512VBMI-ONLY-NEXT: vpermb %zmm0, %zmm1, %zmm0
+; AVX512VBMI-ONLY-NEXT: vpmovb2m %zmm0, %k2
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 (%rsi), %zmm0 {%k2} {z}
+; AVX512VBMI-ONLY-NEXT: vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm2 = [10,11,11,11,12,12,12,13,13,13,14,14,14,15,15,15]
+; AVX512VBMI-ONLY-NEXT: vpermd %zmm1, %zmm2, %zmm1
+; AVX512VBMI-ONLY-NEXT: vptestmd %zmm1, %zmm1, %k1
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 128(%rsi), %zmm1 {%k1} {z}
+; AVX512VBMI-ONLY-NEXT: kshiftrd $16, %k2, %k1
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 64(%rsi), %zmm2 {%k1} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm2, 64(%rdx)
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm1, 128(%rdx)
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm0, (%rdx)
+; AVX512VBMI-ONLY-NEXT: vzeroupper
+; AVX512VBMI-ONLY-NEXT: retq
%src.mask.padded = load <64 x i1>, ptr %in.maskvec, align 64
%src.mask = shufflevector <64 x i1> %src.mask.padded, <64 x i1> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
%tgt.mask = shufflevector <16 x i1> %src.mask, <16 x i1> poison, <48 x i32> <i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 2, i32 2, i32 2, i32 3, i32 3, i32 3, i32 4, i32 4, i32 4, i32 5, i32 5, i32 5, i32 6, i32 6, i32 6, i32 7, i32 7, i32 7, i32 8, i32 8, i32 8, i32 9, i32 9, i32 9, i32 10, i32 10, i32 10, i32 11, i32 11, i32 11, i32 12, i32 12, i32 12, i32 13, i32 13, i32 13, i32 14, i32 14, i32 14, i32 15, i32 15, i32 15>
@@ -6135,27 +6155,47 @@
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
-; AVX512BW-LABEL: mask_replication_factor6_vf8:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: kmovw (%rdi), %k1
-; AVX512BW-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,0,0,0,0,0,1,1,1,1,1,1,2,2,2,2]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm1, %zmm1
-; AVX512BW-NEXT: vptestmd %zmm1, %zmm1, %k1
-; AVX512BW-NEXT: vmovdqa32 (%rsi), %zmm1 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm2 = [5,5,5,5,6,6,6,6,6,6,7,7,7,7,7,7]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm2, %zmm2
-; AVX512BW-NEXT: vptestmd %zmm2, %zmm2, %k1
-; AVX512BW-NEXT: vmovdqa32 128(%rsi), %zmm2 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm3 = [2,2,3,3,3,3,3,3,4,4,4,4,4,4,5,5]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm3, %zmm0
-; AVX512BW-NEXT: vptestmd %zmm0, %zmm0, %k1
-; AVX512BW-NEXT: vmovdqa32 64(%rsi), %zmm0 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 %zmm0, 64(%rdx)
-; AVX512BW-NEXT: vmovdqa64 %zmm2, 128(%rdx)
-; AVX512BW-NEXT: vmovdqa64 %zmm1, (%rdx)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
+; AVX512BW-ONLY-LABEL: mask_replication_factor6_vf8:
+; AVX512BW-ONLY: # %bb.0:
+; AVX512BW-ONLY-NEXT: kmovw (%rdi), %k1
+; AVX512BW-ONLY-NEXT: vpmovm2b %k1, %zmm0
+; AVX512BW-ONLY-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,0,1,0,1,0,1]
+; AVX512BW-ONLY-NEXT: vpshufb {{.*#+}} zmm0 = zmm0[0,0,0,0,0,0,1,1,1,1,1,1,2,2,2,2,18,18,19,19,19,19,19,19,20,20,20,20,20,20,21,21,37,37,37,37,38,38,38,38,38,38,39,39,39,39,39,39,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512BW-ONLY-NEXT: vpmovb2m %zmm0, %k2
+; AVX512BW-ONLY-NEXT: vmovdqa32 (%rsi), %zmm0 {%k2} {z}
+; AVX512BW-ONLY-NEXT: vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm2 = [5,5,5,5,6,6,6,6,6,6,7,7,7,7,7,7]
+; AVX512BW-ONLY-NEXT: vpermd %zmm1, %zmm2, %zmm1
+; AVX512BW-ONLY-NEXT: vptestmd %zmm1, %zmm1, %k1
+; AVX512BW-ONLY-NEXT: vmovdqa32 128(%rsi), %zmm1 {%k1} {z}
+; AVX512BW-ONLY-NEXT: kshiftrd $16, %k2, %k1
+; AVX512BW-ONLY-NEXT: vmovdqa32 64(%rsi), %zmm2 {%k1} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm2, 64(%rdx)
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm1, 128(%rdx)
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm0, (%rdx)
+; AVX512BW-ONLY-NEXT: vzeroupper
+; AVX512BW-ONLY-NEXT: retq
+;
+; AVX512VBMI-ONLY-LABEL: mask_replication_factor6_vf8:
+; AVX512VBMI-ONLY: # %bb.0:
+; AVX512VBMI-ONLY-NEXT: kmovw (%rdi), %k1
+; AVX512VBMI-ONLY-NEXT: vpmovm2b %k1, %zmm0
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm1 = <0,0,0,0,0,0,1,1,1,1,1,1,2,2,2,2,2,2,3,3,3,3,3,3,4,4,4,4,4,4,5,5,5,5,5,5,6,6,6,6,6,6,7,7,7,7,7,7,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512VBMI-ONLY-NEXT: vpermb %zmm0, %zmm1, %zmm0
+; AVX512VBMI-ONLY-NEXT: vpmovb2m %zmm0, %k2
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 (%rsi), %zmm0 {%k2} {z}
+; AVX512VBMI-ONLY-NEXT: vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm2 = [5,5,5,5,6,6,6,6,6,6,7,7,7,7,7,7]
+; AVX512VBMI-ONLY-NEXT: vpermd %zmm1, %zmm2, %zmm1
+; AVX512VBMI-ONLY-NEXT: vptestmd %zmm1, %zmm1, %k1
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 128(%rsi), %zmm1 {%k1} {z}
+; AVX512VBMI-ONLY-NEXT: kshiftrd $16, %k2, %k1
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 64(%rsi), %zmm2 {%k1} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm2, 64(%rdx)
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm1, 128(%rdx)
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm0, (%rdx)
+; AVX512VBMI-ONLY-NEXT: vzeroupper
+; AVX512VBMI-ONLY-NEXT: retq
%src.mask.padded = load <64 x i1>, ptr %in.maskvec, align 64
%src.mask = shufflevector <64 x i1> %src.mask.padded, <64 x i1> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
%tgt.mask = shufflevector <8 x i1> %src.mask, <8 x i1> poison, <48 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
@@ -9545,47 +9585,87 @@
; AVX512DQ-NEXT: vzeroupper
; AVX512DQ-NEXT: retq
;
-; AVX512BW-LABEL: mask_replication_factor7_vf16:
-; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: kmovw (%rdi), %k1
-; AVX512BW-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,0,0,0,0,0,0,1,1,1,1,1,1,1,2,2]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm1, %zmm1
-; AVX512BW-NEXT: vptestmd %zmm1, %zmm1, %k1
-; AVX512BW-NEXT: vmovdqa32 (%rsi), %zmm1 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm2 = [13,13,14,14,14,14,14,14,14,15,15,15,15,15,15,15]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm2, %zmm2
-; AVX512BW-NEXT: vptestmd %zmm2, %zmm2, %k1
-; AVX512BW-NEXT: vmovdqa32 384(%rsi), %zmm2 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm3 = [11,11,11,11,12,12,12,12,12,12,12,13,13,13,13,13]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm3, %zmm3
-; AVX512BW-NEXT: vptestmd %zmm3, %zmm3, %k1
-; AVX512BW-NEXT: vmovdqa32 320(%rsi), %zmm3 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm4 = [9,9,9,9,9,9,10,10,10,10,10,10,10,11,11,11]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm4, %zmm4
-; AVX512BW-NEXT: vptestmd %zmm4, %zmm4, %k1
-; AVX512BW-NEXT: vmovdqa32 256(%rsi), %zmm4 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm5 = [6,7,7,7,7,7,7,7,8,8,8,8,8,8,8,9]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm5, %zmm5
-; AVX512BW-NEXT: vptestmd %zmm5, %zmm5, %k1
-; AVX512BW-NEXT: vmovdqa32 192(%rsi), %zmm5 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm6 = [4,4,4,5,5,5,5,5,5,5,6,6,6,6,6,6]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm6, %zmm6
-; AVX512BW-NEXT: vptestmd %zmm6, %zmm6, %k1
-; AVX512BW-NEXT: vmovdqa32 128(%rsi), %zmm6 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm7 = [2,2,2,2,2,3,3,3,3,3,3,3,4,4,4,4]
-; AVX512BW-NEXT: vpermd %zmm0, %zmm7, %zmm0
-; AVX512BW-NEXT: vptestmd %zmm0, %zmm0, %k1
-; AVX512BW-NEXT: vmovdqa32 64(%rsi), %zmm0 {%k1} {z}
-; AVX512BW-NEXT: vmovdqa64 %zmm0, 64(%rdx)
-; AVX512BW-NEXT: vmovdqa64 %zmm6, 128(%rdx)
-; AVX512BW-NEXT: vmovdqa64 %zmm5, 192(%rdx)
-; AVX512BW-NEXT: vmovdqa64 %zmm4, 256(%rdx)
-; AVX512BW-NEXT: vmovdqa64 %zmm3, 320(%rdx)
-; AVX512BW-NEXT: vmovdqa64 %zmm2, 384(%rdx)
-; AVX512BW-NEXT: vmovdqa64 %zmm1, (%rdx)
-; AVX512BW-NEXT: vzeroupper
-; AVX512BW-NEXT: retq
+; AVX512BW-ONLY-LABEL: mask_replication_factor7_vf16:
+; AVX512BW-ONLY: # %bb.0:
+; AVX512BW-ONLY-NEXT: kmovw (%rdi), %k2
+; AVX512BW-ONLY-NEXT: vpmovm2b %k2, %zmm0
+; AVX512BW-ONLY-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,0,1,0,1,0,1]
+; AVX512BW-ONLY-NEXT: vpshufb {{.*#+}} zmm0 = zmm0[9,9,9,9,9,9,10,10,10,10,10,10,10,11,11,11,27,27,27,27,28,28,28,28,28,28,28,29,29,29,29,29,45,45,46,46,46,46,46,46,46,47,47,47,47,47,47,47,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; AVX512BW-ONLY-NEXT: vpmovb2m %zmm0, %k1
+; AVX512BW-ONLY-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k2} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,0,0,0,0,0,0,1,1,1,1,1,1,1,2,2]
+; AVX512BW-ONLY-NEXT: vpermd %zmm0, %zmm1, %zmm1
+; AVX512BW-ONLY-NEXT: vptestmd %zmm1, %zmm1, %k2
+; AVX512BW-ONLY-NEXT: vmovdqa32 (%rsi), %zmm1 {%k2} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm2 = [13,13,14,14,14,14,14,14,14,15,15,15,15,15,15,15]
+; AVX512BW-ONLY-NEXT: vpermd %zmm0, %zmm2, %zmm2
+; AVX512BW-ONLY-NEXT: vptestmd %zmm2, %zmm2, %k2
+; AVX512BW-ONLY-NEXT: vmovdqa32 384(%rsi), %zmm2 {%k2} {z}
+; AVX512BW-ONLY-NEXT: kshiftrd $16, %k1, %k2
+; AVX512BW-ONLY-NEXT: vmovdqa32 320(%rsi), %zmm3 {%k2} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa32 256(%rsi), %zmm4 {%k1} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm5 = [6,7,7,7,7,7,7,7,8,8,8,8,8,8,8,9]
+; AVX512BW-ONLY-NEXT: vpermd %zmm0, %zmm5, %zmm5
+; AVX512BW-ONLY-NEXT: vptestmd %zmm5, %zmm5, %k1
+; AVX512BW-ONLY-NEXT: vmovdqa32 192(%rsi), %zmm5 {%k1} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm6 = [4,4,4,5,5,5,5,5,5,5,6,6,6,6,6,6]
+; AVX512BW-ONLY-NEXT: vpermd %zmm0, %zmm6, %zmm6
+; AVX512BW-ONLY-NEXT: vptestmd %zmm6, %zmm6, %k1
+; AVX512BW-ONLY-NEXT: vmovdqa32 128(%rsi), %zmm6 {%k1} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm7 = [2,2,2,2,2,3,3,3,3,3,3,3,4,4,4,4]
+; AVX512BW-ONLY-NEXT: vpermd %zmm0, %zmm7, %zmm0
+; AVX512BW-ONLY-NEXT: vptestmd %zmm0, %zmm0, %k1
+; AVX512BW-ONLY-NEXT: vmovdqa32 64(%rsi), %zmm0 {%k1} {z}
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm0, 64(%rdx)
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm6, 128(%rdx)
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm5, 192(%rdx)
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm4, 256(%rdx)
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm3, 320(%rdx)
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm2, 384(%rdx)
+; AVX512BW-ONLY-NEXT: vmovdqa64 %zmm1, (%rdx)
+; AVX512BW-ONLY-NEXT: vzeroupper
+; AVX512BW-ONLY-NEXT: retq
+;
+; AVX512VBMI-ONLY-LABEL: mask_replication_factor7_vf16:
+; AVX512VBMI-ONLY: # %bb.0:
+; AVX512VBMI-ONLY-NEXT: kmovw (%rdi), %k2
+; AVX512VBMI-ONLY-NEXT: vpmovm2b %k2, %zmm0
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm1 = <9,9,9,9,9,9,10,10,10,10,10,10,10,11,11,11,11,11,11,11,12,12,12,12,12,12,12,13,13,13,13,13,13,13,14,14,14,14,14,14,14,15,15,15,15,15,15,15,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u>
+; AVX512VBMI-ONLY-NEXT: vpermb %zmm0, %zmm1, %zmm0
+; AVX512VBMI-ONLY-NEXT: vpmovb2m %zmm0, %k1
+; AVX512VBMI-ONLY-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k2} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,0,0,0,0,0,0,1,1,1,1,1,1,1,2,2]
+; AVX512VBMI-ONLY-NEXT: vpermd %zmm0, %zmm1, %zmm1
+; AVX512VBMI-ONLY-NEXT: vptestmd %zmm1, %zmm1, %k2
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 (%rsi), %zmm1 {%k2} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm2 = [13,13,14,14,14,14,14,14,14,15,15,15,15,15,15,15]
+; AVX512VBMI-ONLY-NEXT: vpermd %zmm0, %zmm2, %zmm2
+; AVX512VBMI-ONLY-NEXT: vptestmd %zmm2, %zmm2, %k2
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 384(%rsi), %zmm2 {%k2} {z}
+; AVX512VBMI-ONLY-NEXT: kshiftrd $16, %k1, %k2
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 320(%rsi), %zmm3 {%k2} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 256(%rsi), %zmm4 {%k1} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm5 = [6,7,7,7,7,7,7,7,8,8,8,8,8,8,8,9]
+; AVX512VBMI-ONLY-NEXT: vpermd %zmm0, %zmm5, %zmm5
+; AVX512VBMI-ONLY-NEXT: vptestmd %zmm5, %zmm5, %k1
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 192(%rsi), %zmm5 {%k1} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm6 = [4,4,4,5,5,5,5,5,5,5,6,6,6,6,6,6]
+; AVX512VBMI-ONLY-NEXT: vpermd %zmm0, %zmm6, %zmm6
+; AVX512VBMI-ONLY-NEXT: vptestmd %zmm6, %zmm6, %k1
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 128(%rsi), %zmm6 {%k1} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 {{.*#+}} zmm7 = [2,2,2,2,2,3,3,3,3,3,3,3,4,4,4,4]
+; AVX512VBMI-ONLY-NEXT: vpermd %zmm0, %zmm7, %zmm0
+; AVX512VBMI-ONLY-NEXT: vptestmd %zmm0, %zmm0, %k1
+; AVX512VBMI-ONLY-NEXT: vmovdqa32 64(%rsi), %zmm0 {%k1} {z}
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm0, 64(%rdx)
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm6, 128(%rdx)
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm5, 192(%rdx)
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm4, 256(%rdx)
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm3, 320(%rdx)
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm2, 384(%rdx)
+; AVX512VBMI-ONLY-NEXT: vmovdqa64 %zmm1, (%rdx)
+; AVX512VBMI-ONLY-NEXT: vzeroupper
+; AVX512VBMI-ONLY-NEXT: retq
%src.mask.padded = load <64 x i1>, ptr %in.maskvec, align 64
%src.mask = shufflevector <64 x i1> %src.mask.padded, <64 x i1> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
%tgt.mask = shufflevector <16 x i1> %src.mask, <16 x i1> poison, <112 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
diff --git a/llvm/test/CodeGen/X86/vector-rotate-256.ll b/llvm/test/CodeGen/X86/vector-rotate-256.ll
index 7047d5e..6c854b7 100644
--- a/llvm/test/CodeGen/X86/vector-rotate-256.ll
+++ b/llvm/test/CodeGen/X86/vector-rotate-256.ll
@@ -666,7 +666,7 @@
;
; AVX2-LABEL: splatvar_rotate_v16i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX2-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX2-NEXT: vpsrlw $1, %ymm0, %ymm4
; AVX2-NEXT: vpsrlw %xmm3, %ymm4, %ymm3
@@ -677,7 +677,7 @@
;
; AVX512F-LABEL: splatvar_rotate_v16i16:
; AVX512F: # %bb.0:
-; AVX512F-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512F-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512F-NEXT: vpsrlw $1, %ymm0, %ymm4
; AVX512F-NEXT: vpsrlw %xmm3, %ymm4, %ymm3
@@ -688,7 +688,7 @@
;
; AVX512VL-LABEL: splatvar_rotate_v16i16:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512VL-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512VL-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512VL-NEXT: vpsrlw $1, %ymm0, %ymm4
; AVX512VL-NEXT: vpsrlw %xmm3, %ymm4, %ymm3
@@ -699,7 +699,7 @@
;
; AVX512BW-LABEL: splatvar_rotate_v16i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512BW-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512BW-NEXT: vpsrlw $1, %ymm0, %ymm4
; AVX512BW-NEXT: vpsrlw %xmm3, %ymm4, %ymm3
@@ -710,7 +710,7 @@
;
; AVX512VLBW-LABEL: splatvar_rotate_v16i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512VLBW-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512VLBW-NEXT: vpsrlw $1, %ymm0, %ymm4
; AVX512VLBW-NEXT: vpsrlw %xmm3, %ymm4, %ymm3
diff --git a/llvm/test/CodeGen/X86/vector-rotate-512.ll b/llvm/test/CodeGen/X86/vector-rotate-512.ll
index 6504c3e..d3bb0e1 100644
--- a/llvm/test/CodeGen/X86/vector-rotate-512.ll
+++ b/llvm/test/CodeGen/X86/vector-rotate-512.ll
@@ -347,7 +347,7 @@
;
; AVX512BW-LABEL: splatvar_rotate_v32i16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512BW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512BW-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512BW-NEXT: vpsrlw $1, %zmm0, %zmm4
; AVX512BW-NEXT: vpsrlw %xmm3, %zmm4, %zmm3
@@ -358,7 +358,7 @@
;
; AVX512VLBW-LABEL: splatvar_rotate_v32i16:
; AVX512VLBW: # %bb.0:
-; AVX512VLBW-NEXT: vpbroadcastq {{.*#+}} xmm2 = [15,0,0,0,15,0,0,0]
+; AVX512VLBW-NEXT: vmovdqa {{.*#+}} xmm2 = [15,0,0,0]
; AVX512VLBW-NEXT: vpandn %xmm2, %xmm1, %xmm3
; AVX512VLBW-NEXT: vpsrlw $1, %zmm0, %zmm4
; AVX512VLBW-NEXT: vpsrlw %xmm3, %zmm4, %zmm3
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-concatenation.ll b/llvm/test/CodeGen/X86/vector-shuffle-concatenation.ll
index 2812bf3..7b9881f 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-concatenation.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-concatenation.ll
@@ -568,7 +568,7 @@
;
; AVX-LABEL: concat_shuf_of_a_to_itself:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = mem[1,0]
+; AVX-NEXT: vpermilps {{.*#+}} xmm0 = mem[2,3,0,1]
; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX-NEXT: vmovaps %ymm0, (%rsi)
; AVX-NEXT: vzeroupper
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-sse4a.ll b/llvm/test/CodeGen/X86/vector-shuffle-sse4a.ll
index b8db14c..8a64d1a 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-sse4a.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-sse4a.ll
@@ -364,8 +364,8 @@
; AMD10H: # %bb.0:
; AMD10H-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; AMD10H-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; AMD10H-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AMD10H-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
+; AMD10H-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AMD10H-NEXT: packuswb %xmm0, %xmm0
; AMD10H-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-zext.ll b/llvm/test/CodeGen/X86/vector-zext.ll
index 7abd288..9c0615b 100644
--- a/llvm/test/CodeGen/X86/vector-zext.ll
+++ b/llvm/test/CodeGen/X86/vector-zext.ll
@@ -2571,9 +2571,10 @@
;
; SSE41-LABEL: splatshuf_zext_v4i64:
; SSE41: # %bb.0:
-; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
-; SSE41-NEXT: pmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
-; SSE41-NEXT: movdqa %xmm0, %xmm1
+; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; SSE41-NEXT: pxor %xmm2, %xmm2
+; SSE41-NEXT: pmovzxdq {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero
+; SSE41-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; SSE41-NEXT: retq
;
; AVX1-LABEL: splatshuf_zext_v4i64:
@@ -2710,8 +2711,11 @@
;
; SSE41-LABEL: splatshuf_zext_v16i16:
; SSE41: # %bb.0:
-; SSE41-NEXT: pshufb {{.*#+}} xmm0 = xmm0[14],zero,xmm0[14],zero,xmm0[14],zero,xmm0[14],zero,xmm0[14],zero,xmm0[14],zero,xmm0[14],zero,xmm0[14],zero
; SSE41-NEXT: movdqa %xmm0, %xmm1
+; SSE41-NEXT: pshufb {{.*#+}} xmm1 = xmm1[14,14,14,14,14,14,14,14,14,14,14,14,14,14,14,14]
+; SSE41-NEXT: pxor %xmm2, %xmm2
+; SSE41-NEXT: pmovzxbw {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; SSE41-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm2[8],xmm1[9],xmm2[9],xmm1[10],xmm2[10],xmm1[11],xmm2[11],xmm1[12],xmm2[12],xmm1[13],xmm2[13],xmm1[14],xmm2[14],xmm1[15],xmm2[15]
; SSE41-NEXT: retq
;
; AVX1-LABEL: splatshuf_zext_v16i16:
diff --git a/llvm/test/CodeGen/X86/wide-integer-cmp.ll b/llvm/test/CodeGen/X86/wide-integer-cmp.ll
index a15d633..189f516 100644
--- a/llvm/test/CodeGen/X86/wide-integer-cmp.ll
+++ b/llvm/test/CodeGen/X86/wide-integer-cmp.ll
@@ -99,8 +99,8 @@
; CHECK-NEXT: movl {{[0-9]+}}(%esp), %esi
; CHECK-NEXT: cmpl {{[0-9]+}}(%esp), %edx
; CHECK-NEXT: sbbl {{[0-9]+}}(%esp), %esi
-; CHECK-NEXT: sbbl {{[0-9]+}}(%esp), %ecx
; CHECK-NEXT: sbbl {{[0-9]+}}(%esp), %eax
+; CHECK-NEXT: sbbl {{[0-9]+}}(%esp), %ecx
; CHECK-NEXT: jge .LBB4_2
; CHECK-NEXT: # %bb.1: # %bb1
; CHECK-NEXT: movl $1, %eax
diff --git a/llvm/test/CodeGen/X86/widen-load-of-small-alloca-with-zero-upper-half.ll b/llvm/test/CodeGen/X86/widen-load-of-small-alloca-with-zero-upper-half.ll
index 691ca40..f1524f7 100644
--- a/llvm/test/CodeGen/X86/widen-load-of-small-alloca-with-zero-upper-half.ll
+++ b/llvm/test/CodeGen/X86/widen-load-of-small-alloca-with-zero-upper-half.ll
@@ -65,6 +65,7 @@
; X64-NO-BMI2-LABEL: load_1byte_chunk_of_4byte_alloca_with_zero_upper_half:
; X64-NO-BMI2: # %bb.0:
; X64-NO-BMI2-NEXT: movzwl (%rdi), %eax
+; X64-NO-BMI2-NEXT: movzwl %ax, %eax
; X64-NO-BMI2-NEXT: leal (,%rsi,8), %ecx
; X64-NO-BMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NO-BMI2-NEXT: shrl %cl, %eax
@@ -74,6 +75,7 @@
; X64-BMI2-LABEL: load_1byte_chunk_of_4byte_alloca_with_zero_upper_half:
; X64-BMI2: # %bb.0:
; X64-BMI2-NEXT: movzwl (%rdi), %eax
+; X64-BMI2-NEXT: movzwl %ax, %eax
; X64-BMI2-NEXT: shll $3, %esi
; X64-BMI2-NEXT: shrxl %esi, %eax, %eax
; X64-BMI2-NEXT: movb %al, (%rdx)
@@ -81,14 +83,15 @@
;
; X86-NO-BMI2-LABEL: load_1byte_chunk_of_4byte_alloca_with_zero_upper_half:
; X86-NO-BMI2: # %bb.0:
-; X86-NO-BMI2-NEXT: movl {{[0-9]+}}(%esp), %edx
-; X86-NO-BMI2-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NO-BMI2-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NO-BMI2-NEXT: movzwl (%eax), %eax
+; X86-NO-BMI2-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NO-BMI2-NEXT: movl {{[0-9]+}}(%esp), %edx
+; X86-NO-BMI2-NEXT: movzwl (%edx), %edx
+; X86-NO-BMI2-NEXT: movzwl %dx, %edx
; X86-NO-BMI2-NEXT: shll $3, %ecx
; X86-NO-BMI2-NEXT: # kill: def $cl killed $cl killed $ecx
-; X86-NO-BMI2-NEXT: shrl %cl, %eax
-; X86-NO-BMI2-NEXT: movb %al, (%edx)
+; X86-NO-BMI2-NEXT: shrl %cl, %edx
+; X86-NO-BMI2-NEXT: movb %dl, (%eax)
; X86-NO-BMI2-NEXT: retl
;
; X86-BMI2-LABEL: load_1byte_chunk_of_4byte_alloca_with_zero_upper_half:
@@ -97,6 +100,7 @@
; X86-BMI2-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-BMI2-NEXT: movl {{[0-9]+}}(%esp), %edx
; X86-BMI2-NEXT: movzwl (%edx), %edx
+; X86-BMI2-NEXT: movzwl %dx, %edx
; X86-BMI2-NEXT: shll $3, %ecx
; X86-BMI2-NEXT: shrxl %ecx, %edx, %ecx
; X86-BMI2-NEXT: movb %cl, (%eax)
@@ -119,6 +123,7 @@
; X64-NO-BMI2-LABEL: load_2byte_chunk_of_4byte_alloca_with_zero_upper_half:
; X64-NO-BMI2: # %bb.0:
; X64-NO-BMI2-NEXT: movzwl (%rdi), %eax
+; X64-NO-BMI2-NEXT: movzwl %ax, %eax
; X64-NO-BMI2-NEXT: leal (,%rsi,8), %ecx
; X64-NO-BMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-NO-BMI2-NEXT: shrl %cl, %eax
@@ -128,6 +133,7 @@
; X64-BMI2-LABEL: load_2byte_chunk_of_4byte_alloca_with_zero_upper_half:
; X64-BMI2: # %bb.0:
; X64-BMI2-NEXT: movzwl (%rdi), %eax
+; X64-BMI2-NEXT: movzwl %ax, %eax
; X64-BMI2-NEXT: shll $3, %esi
; X64-BMI2-NEXT: shrxl %esi, %eax, %eax
; X64-BMI2-NEXT: movw %ax, (%rdx)
@@ -139,6 +145,7 @@
; X86-NO-BMI2-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NO-BMI2-NEXT: movl {{[0-9]+}}(%esp), %edx
; X86-NO-BMI2-NEXT: movzwl (%edx), %edx
+; X86-NO-BMI2-NEXT: movzwl %dx, %edx
; X86-NO-BMI2-NEXT: shll $3, %ecx
; X86-NO-BMI2-NEXT: # kill: def $cl killed $cl killed $ecx
; X86-NO-BMI2-NEXT: shrl %cl, %edx
@@ -151,6 +158,7 @@
; X86-BMI2-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-BMI2-NEXT: movl {{[0-9]+}}(%esp), %edx
; X86-BMI2-NEXT: movzwl (%edx), %edx
+; X86-BMI2-NEXT: movzwl %dx, %edx
; X86-BMI2-NEXT: shll $3, %ecx
; X86-BMI2-NEXT: shrxl %ecx, %edx, %ecx
; X86-BMI2-NEXT: movw %cx, (%eax)
@@ -1933,7 +1941,7 @@
}
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; ALL: {{.*}}
-; X64-NO-SHLD: {{.*}}
-; X64-SHLD: {{.*}}
; X86-NO-SHLD: {{.*}}
; X86-SHLD: {{.*}}
+; X64-NO-SHLD: {{.*}}
+; X64-SHLD: {{.*}}
diff --git a/llvm/test/CodeGen/X86/xaluo128.ll b/llvm/test/CodeGen/X86/xaluo128.ll
index 977df0f..740a2dd 100644
--- a/llvm/test/CodeGen/X86/xaluo128.ll
+++ b/llvm/test/CodeGen/X86/xaluo128.ll
@@ -24,13 +24,13 @@
; X86-NEXT: movl {{[0-9]+}}(%esp), %ebx
; X86-NEXT: addl {{[0-9]+}}(%esp), %edi
; X86-NEXT: adcl {{[0-9]+}}(%esp), %ebx
-; X86-NEXT: adcl {{[0-9]+}}(%esp), %esi
; X86-NEXT: adcl {{[0-9]+}}(%esp), %edx
+; X86-NEXT: adcl {{[0-9]+}}(%esp), %esi
; X86-NEXT: seto %al
; X86-NEXT: movl %edi, (%ecx)
; X86-NEXT: movl %ebx, 4(%ecx)
-; X86-NEXT: movl %esi, 8(%ecx)
-; X86-NEXT: movl %edx, 12(%ecx)
+; X86-NEXT: movl %edx, 8(%ecx)
+; X86-NEXT: movl %esi, 12(%ecx)
; X86-NEXT: popl %esi
; X86-NEXT: popl %edi
; X86-NEXT: popl %ebx
@@ -64,13 +64,13 @@
; X86-NEXT: movl {{[0-9]+}}(%esp), %ebx
; X86-NEXT: addl {{[0-9]+}}(%esp), %edi
; X86-NEXT: adcl {{[0-9]+}}(%esp), %ebx
-; X86-NEXT: adcl {{[0-9]+}}(%esp), %esi
; X86-NEXT: adcl {{[0-9]+}}(%esp), %edx
+; X86-NEXT: adcl {{[0-9]+}}(%esp), %esi
; X86-NEXT: setb %al
; X86-NEXT: movl %edi, (%ecx)
; X86-NEXT: movl %ebx, 4(%ecx)
-; X86-NEXT: movl %esi, 8(%ecx)
-; X86-NEXT: movl %edx, 12(%ecx)
+; X86-NEXT: movl %edx, 8(%ecx)
+; X86-NEXT: movl %esi, 12(%ecx)
; X86-NEXT: popl %esi
; X86-NEXT: popl %edi
; X86-NEXT: popl %ebx
@@ -105,13 +105,13 @@
; X86-NEXT: movl {{[0-9]+}}(%esp), %ebx
; X86-NEXT: subl {{[0-9]+}}(%esp), %edi
; X86-NEXT: sbbl {{[0-9]+}}(%esp), %ebx
-; X86-NEXT: sbbl {{[0-9]+}}(%esp), %esi
; X86-NEXT: sbbl {{[0-9]+}}(%esp), %edx
+; X86-NEXT: sbbl {{[0-9]+}}(%esp), %esi
; X86-NEXT: seto %al
; X86-NEXT: movl %edi, (%ecx)
; X86-NEXT: movl %ebx, 4(%ecx)
-; X86-NEXT: movl %esi, 8(%ecx)
-; X86-NEXT: movl %edx, 12(%ecx)
+; X86-NEXT: movl %edx, 8(%ecx)
+; X86-NEXT: movl %esi, 12(%ecx)
; X86-NEXT: popl %esi
; X86-NEXT: popl %edi
; X86-NEXT: popl %ebx
@@ -145,13 +145,13 @@
; X86-NEXT: movl {{[0-9]+}}(%esp), %ebx
; X86-NEXT: subl {{[0-9]+}}(%esp), %edi
; X86-NEXT: sbbl {{[0-9]+}}(%esp), %ebx
-; X86-NEXT: sbbl {{[0-9]+}}(%esp), %esi
; X86-NEXT: sbbl {{[0-9]+}}(%esp), %edx
+; X86-NEXT: sbbl {{[0-9]+}}(%esp), %esi
; X86-NEXT: setb %al
; X86-NEXT: movl %edi, (%ecx)
; X86-NEXT: movl %ebx, 4(%ecx)
-; X86-NEXT: movl %esi, 8(%ecx)
-; X86-NEXT: movl %edx, 12(%ecx)
+; X86-NEXT: movl %edx, 8(%ecx)
+; X86-NEXT: movl %esi, 12(%ecx)
; X86-NEXT: popl %esi
; X86-NEXT: popl %edi
; X86-NEXT: popl %ebx
diff --git a/llvm/test/CodeGen/X86/xor.ll b/llvm/test/CodeGen/X86/xor.ll
index 2072568..c392dd6 100644
--- a/llvm/test/CodeGen/X86/xor.ll
+++ b/llvm/test/CodeGen/X86/xor.ll
@@ -409,8 +409,8 @@
;
; X64-WIN-LABEL: PR17487:
; X64-WIN: # %bb.0:
-; X64-WIN-NEXT: andb $1, %cl
; X64-WIN-NEXT: movzbl %cl, %eax
+; X64-WIN-NEXT: andl $1, %eax
; X64-WIN-NEXT: retq
%tmp = insertelement <2 x i1> undef, i1 %tobool, i32 1
%tmp1 = zext <2 x i1> %tmp to <2 x i64>
diff --git a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
index 8f9d216..b5f9c26 100644
--- a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
@@ -4855,20 +4855,20 @@
; AVX-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3,4,5,6,7]
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; AVX-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3,4,5,6,7]
-; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,0,1,1]
-; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
-; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,0,1]
+; AVX-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX-NEXT: vxorps %xmm3, %xmm3, %xmm3
-; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6],ymm3[7]
+; AVX-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6],ymm3[7]
+; AVX-NEXT: vxorps %xmm3, %xmm3, %xmm3
+; AVX-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3]
+; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,0,1,1]
; AVX-NEXT: vpaddb (%rdx), %xmm1, %xmm1
-; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX-NEXT: vpaddb 16(%rdx), %xmm0, %xmm0
-; AVX-NEXT: vpaddb 32(%rdx), %xmm2, %xmm2
+; AVX-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX-NEXT: vpaddb 16(%rdx), %xmm2, %xmm2
+; AVX-NEXT: vpaddb 32(%rdx), %xmm0, %xmm0
; AVX-NEXT: vmovdqa %xmm1, (%rcx)
-; AVX-NEXT: vmovdqa %xmm2, 32(%rcx)
-; AVX-NEXT: vmovdqa %xmm0, 16(%rcx)
+; AVX-NEXT: vmovdqa %xmm0, 32(%rcx)
+; AVX-NEXT: vmovdqa %xmm2, 16(%rcx)
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
@@ -5055,18 +5055,18 @@
; AVX-NEXT: vpaddb (%rsi), %xmm0, %xmm0
; AVX-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3,4,5,6,7]
-; AVX-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vpblendw {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3,4,5,6,7]
-; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm2
; AVX-NEXT: vxorps %xmm3, %xmm3, %xmm3
-; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm3[5,6,7]
+; AVX-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm3[5,6,7]
+; AVX-NEXT: vxorps %xmm3, %xmm3, %xmm3
+; AVX-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm3[1,2,3]
; AVX-NEXT: vpaddb (%rdx), %xmm1, %xmm1
-; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX-NEXT: vpaddb 16(%rdx), %xmm0, %xmm0
-; AVX-NEXT: vpaddb 32(%rdx), %xmm2, %xmm2
+; AVX-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX-NEXT: vpaddb 16(%rdx), %xmm2, %xmm2
+; AVX-NEXT: vpaddb 32(%rdx), %xmm0, %xmm0
; AVX-NEXT: vmovdqa %xmm1, (%rcx)
-; AVX-NEXT: vmovdqa %xmm2, 32(%rcx)
-; AVX-NEXT: vmovdqa %xmm0, 16(%rcx)
+; AVX-NEXT: vmovdqa %xmm0, 32(%rcx)
+; AVX-NEXT: vmovdqa %xmm2, 16(%rcx)
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
@@ -5332,17 +5332,17 @@
; AVX-NEXT: vpaddb 48(%rsi), %xmm1, %xmm1
; AVX-NEXT: vpaddb (%rsi), %xmm0, %xmm0
; AVX-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3],xmm1[4,5,6,7]
-; AVX-NEXT: vmovq {{.*#+}} xmm2 = xmm0[0],zero
-; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm2
; AVX-NEXT: vxorps %xmm3, %xmm3, %xmm3
-; AVX-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX-NEXT: vmovq {{.*#+}} xmm0 = xmm0[0],zero
; AVX-NEXT: vpaddb (%rdx), %xmm1, %xmm1
-; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX-NEXT: vpaddb 16(%rdx), %xmm0, %xmm0
-; AVX-NEXT: vpaddb 32(%rdx), %xmm2, %xmm2
+; AVX-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX-NEXT: vpaddb 16(%rdx), %xmm2, %xmm2
+; AVX-NEXT: vpaddb 32(%rdx), %xmm0, %xmm0
; AVX-NEXT: vmovdqa %xmm1, (%rcx)
-; AVX-NEXT: vmovdqa %xmm2, 32(%rcx)
-; AVX-NEXT: vmovdqa %xmm0, 16(%rcx)
+; AVX-NEXT: vmovdqa %xmm0, 32(%rcx)
+; AVX-NEXT: vmovdqa %xmm2, 16(%rcx)
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
