llvm/test/CodeGen/AMDGPU/lds-misaligned-bug.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefix=SPLIT %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1011 -verify-machineinstrs < %s | FileCheck -check-prefix=SPLIT %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1012 -verify-machineinstrs < %s | FileCheck -check-prefix=SPLIT %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs -mattr=+cumode < %s | FileCheck -check-prefix=ALIGNED-GFX10 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs -mattr=+cumode,+unaligned-access-mode < %s | FileCheck -check-prefix=UNALIGNED-GFX10 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefix=ALIGNED-GFX11 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs -mattr=+cumode < %s | FileCheck -check-prefix=ALIGNED-GFX11 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs -mattr=+cumode -early-live-intervals < %s | FileCheck -check-prefix=ALIGNED-GFX11 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs -mattr=+cumode,+unaligned-access-mode < %s | FileCheck -check-prefix=UNALIGNED-GFX11 %s

define amdgpu_kernel void @test_local_misaligned_v2(ptr addrspace(3) %arg) {
; SPLIT-LABEL: test_local_misaligned_v2:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dword s0, s[4:5], 0x24
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_lshl_add_u32 v2, v0, 2, s0
; SPLIT-NEXT:    ds_read2_b32 v[0:1], v2 offset1:1
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    ds_write2_b32 v2, v1, v0 offset1:1
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_local_misaligned_v2:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_lshl_add_u32 v2, v0, 2, s0
; ALIGNED-GFX10-NEXT:    ds_read2_b32 v[0:1], v2 offset1:1
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    ds_write2_b32 v2, v1, v0 offset1:1
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_local_misaligned_v2:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_lshl_add_u32 v2, v0, 2, s0
; UNALIGNED-GFX10-NEXT:    ds_read2_b32 v[0:1], v2 offset1:1
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    ds_write2_b32 v2, v1, v0 offset1:1
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_local_misaligned_v2:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshl_add_u32 v2, v0, 2, s0
; ALIGNED-GFX11-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset1:1
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    ds_store_2addr_b32 v2, v1, v0 offset1:1
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_local_misaligned_v2:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshl_add_u32 v2, v0, 2, s0
; UNALIGNED-GFX11-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset1:1
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    ds_store_2addr_b32 v2, v1, v0 offset1:1
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr addrspace(3) %arg, i32 %lid
  %load = load <2 x i32>, ptr addrspace(3) %gep, align 4
  %v1 = extractelement <2 x i32> %load, i32 0
  %v2 = extractelement <2 x i32> %load, i32 1
  %v3 = insertelement <2 x i32> poison, i32 %v2, i32 0
  %v4 = insertelement <2 x i32> %v3, i32 %v1, i32 1
  store <2 x i32> %v4, ptr addrspace(3) %gep, align 4
  ret void
}

define amdgpu_kernel void @test_local_misaligned_v4(ptr addrspace(3) %arg) {
; SPLIT-LABEL: test_local_misaligned_v4:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dword s0, s[4:5], 0x24
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_lshl_add_u32 v4, v0, 2, s0
; SPLIT-NEXT:    ds_read2_b32 v[0:1], v4 offset0:2 offset1:3
; SPLIT-NEXT:    ds_read2_b32 v[2:3], v4 offset1:1
; SPLIT-NEXT:    s_waitcnt lgkmcnt(1)
; SPLIT-NEXT:    ds_write2_b32 v4, v1, v0 offset1:1
; SPLIT-NEXT:    s_waitcnt lgkmcnt(1)
; SPLIT-NEXT:    ds_write2_b32 v4, v3, v2 offset0:2 offset1:3
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_local_misaligned_v4:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_lshl_add_u32 v4, v0, 2, s0
; ALIGNED-GFX10-NEXT:    ds_read2_b32 v[0:1], v4 offset1:1
; ALIGNED-GFX10-NEXT:    ds_read2_b32 v[2:3], v4 offset0:2 offset1:3
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(1)
; ALIGNED-GFX10-NEXT:    ds_write2_b32 v4, v1, v0 offset0:2 offset1:3
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(1)
; ALIGNED-GFX10-NEXT:    ds_write2_b32 v4, v3, v2 offset1:1
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_local_misaligned_v4:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_lshl_add_u32 v4, v0, 2, s0
; UNALIGNED-GFX10-NEXT:    ds_read2_b32 v[0:1], v4 offset0:2 offset1:3
; UNALIGNED-GFX10-NEXT:    ds_read2_b32 v[2:3], v4 offset1:1
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(1)
; UNALIGNED-GFX10-NEXT:    ds_write2_b32 v4, v1, v0 offset1:1
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(1)
; UNALIGNED-GFX10-NEXT:    ds_write2_b32 v4, v3, v2 offset0:2 offset1:3
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_local_misaligned_v4:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshl_add_u32 v4, v0, 2, s0
; ALIGNED-GFX11-NEXT:    ds_load_2addr_b32 v[0:1], v4 offset1:1
; ALIGNED-GFX11-NEXT:    ds_load_2addr_b32 v[2:3], v4 offset0:2 offset1:3
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(1)
; ALIGNED-GFX11-NEXT:    ds_store_2addr_b32 v4, v1, v0 offset0:2 offset1:3
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(1)
; ALIGNED-GFX11-NEXT:    ds_store_2addr_b32 v4, v3, v2 offset1:1
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_local_misaligned_v4:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshl_add_u32 v4, v0, 2, s0
; UNALIGNED-GFX11-NEXT:    ds_load_2addr_b32 v[0:1], v4 offset0:2 offset1:3
; UNALIGNED-GFX11-NEXT:    ds_load_2addr_b32 v[2:3], v4 offset1:1
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(1)
; UNALIGNED-GFX11-NEXT:    ds_store_2addr_b32 v4, v1, v0 offset1:1
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(1)
; UNALIGNED-GFX11-NEXT:    ds_store_2addr_b32 v4, v3, v2 offset0:2 offset1:3
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr addrspace(3) %arg, i32 %lid
  %load = load <4 x i32>, ptr addrspace(3) %gep, align 4
  %v1 = extractelement <4 x i32> %load, i32 0
  %v2 = extractelement <4 x i32> %load, i32 1
  %v3 = extractelement <4 x i32> %load, i32 2
  %v4 = extractelement <4 x i32> %load, i32 3
  %v5 = insertelement <4 x i32> poison, i32 %v4, i32 0
  %v6 = insertelement <4 x i32> %v5, i32 %v3, i32 1
  %v7 = insertelement <4 x i32> %v6, i32 %v2, i32 2
  %v8 = insertelement <4 x i32> %v7, i32 %v1, i32 3
  store <4 x i32> %v8, ptr addrspace(3) %gep, align 4
  ret void
}

define amdgpu_kernel void @test_local_misaligned_v3(ptr addrspace(3) %arg) {
; SPLIT-LABEL: test_local_misaligned_v3:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dword s0, s[4:5], 0x24
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_lshl_add_u32 v2, v0, 2, s0
; SPLIT-NEXT:    ds_read2_b32 v[0:1], v2 offset1:1
; SPLIT-NEXT:    ds_read_b32 v3, v2 offset:8
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    ds_write2_b32 v2, v3, v0 offset1:1
; SPLIT-NEXT:    ds_write_b32 v2, v1 offset:8
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_local_misaligned_v3:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_lshl_add_u32 v2, v0, 2, s0
; ALIGNED-GFX10-NEXT:    ds_read2_b32 v[0:1], v2 offset1:1
; ALIGNED-GFX10-NEXT:    ds_read_b32 v3, v2 offset:8
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    ds_write2_b32 v2, v3, v0 offset1:1
; ALIGNED-GFX10-NEXT:    ds_write_b32 v2, v1 offset:8
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_local_misaligned_v3:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_lshl_add_u32 v2, v0, 2, s0
; UNALIGNED-GFX10-NEXT:    ds_read2_b32 v[0:1], v2 offset1:1
; UNALIGNED-GFX10-NEXT:    ds_read_b32 v3, v2 offset:8
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    ds_write2_b32 v2, v3, v0 offset1:1
; UNALIGNED-GFX10-NEXT:    ds_write_b32 v2, v1 offset:8
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_local_misaligned_v3:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshl_add_u32 v2, v0, 2, s0
; ALIGNED-GFX11-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset1:1
; ALIGNED-GFX11-NEXT:    ds_load_b32 v3, v2 offset:8
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    ds_store_2addr_b32 v2, v3, v0 offset1:1
; ALIGNED-GFX11-NEXT:    ds_store_b32 v2, v1 offset:8
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_local_misaligned_v3:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshl_add_u32 v2, v0, 2, s0
; UNALIGNED-GFX11-NEXT:    ds_load_2addr_b32 v[0:1], v2 offset1:1
; UNALIGNED-GFX11-NEXT:    ds_load_b32 v3, v2 offset:8
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    ds_store_2addr_b32 v2, v3, v0 offset1:1
; UNALIGNED-GFX11-NEXT:    ds_store_b32 v2, v1 offset:8
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr addrspace(3) %arg, i32 %lid
  %load = load <3 x i32>, ptr addrspace(3) %gep, align 4
  %v1 = extractelement <3 x i32> %load, i32 0
  %v2 = extractelement <3 x i32> %load, i32 1
  %v3 = extractelement <3 x i32> %load, i32 2
  %v5 = insertelement <3 x i32> poison, i32 %v3, i32 0
  %v6 = insertelement <3 x i32> %v5, i32 %v1, i32 1
  %v7 = insertelement <3 x i32> %v6, i32 %v2, i32 2
  store <3 x i32> %v7, ptr addrspace(3) %gep, align 4
  ret void
}

define amdgpu_kernel void @test_flat_misaligned_v2(ptr %arg) {
; SPLIT-LABEL: test_flat_misaligned_v2:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; SPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_add_co_u32 v0, s0, s0, v0
; SPLIT-NEXT:    v_add_co_ci_u32_e64 v1, s0, s1, 0, s0
; SPLIT-NEXT:    v_add_co_u32 v2, vcc_lo, v0, 4
; SPLIT-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v1, vcc_lo
; SPLIT-NEXT:    s_clause 0x1
; SPLIT-NEXT:    flat_load_dword v4, v[2:3]
; SPLIT-NEXT:    flat_load_dword v5, v[0:1]
; SPLIT-NEXT:    s_waitcnt vmcnt(1) lgkmcnt(1)
; SPLIT-NEXT:    flat_store_dword v[0:1], v4
; SPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(1)
; SPLIT-NEXT:    flat_store_dword v[2:3], v5
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_flat_misaligned_v2:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_add_co_u32 v0, s0, s0, v0
; ALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v1, s0, s1, 0, s0
; ALIGNED-GFX10-NEXT:    flat_load_dwordx2 v[2:3], v[0:1]
; ALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; ALIGNED-GFX10-NEXT:    flat_store_dwordx2 v[0:1], v[3:4]
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_flat_misaligned_v2:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_add_co_u32 v0, s0, s0, v0
; UNALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v1, s0, s1, 0, s0
; UNALIGNED-GFX10-NEXT:    flat_load_dwordx2 v[2:3], v[0:1]
; UNALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; UNALIGNED-GFX10-NEXT:    flat_store_dwordx2 v[0:1], v[3:4]
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_flat_misaligned_v2:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_add_co_u32 v0, s0, s0, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v1, null, s1, 0, s0
; ALIGNED-GFX11-NEXT:    flat_load_b64 v[2:3], v[0:1]
; ALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_mov_b32_e32 v4, v2
; ALIGNED-GFX11-NEXT:    flat_store_b64 v[0:1], v[3:4]
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_flat_misaligned_v2:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_add_co_u32 v0, s0, s0, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v1, null, s1, 0, s0
; UNALIGNED-GFX11-NEXT:    flat_load_b64 v[2:3], v[0:1]
; UNALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_mov_b32_e32 v4, v2
; UNALIGNED-GFX11-NEXT:    flat_store_b64 v[0:1], v[3:4]
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr %arg, i32 %lid
  %load = load <2 x i32>, ptr %gep, align 4
  %v1 = extractelement <2 x i32> %load, i32 0
  %v2 = extractelement <2 x i32> %load, i32 1
  %v3 = insertelement <2 x i32> poison, i32 %v2, i32 0
  %v4 = insertelement <2 x i32> %v3, i32 %v1, i32 1
  store <2 x i32> %v4, ptr %gep, align 4
  ret void
}

define amdgpu_kernel void @test_flat_misaligned_v4(ptr %arg) {
; SPLIT-LABEL: test_flat_misaligned_v4:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; SPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_add_co_u32 v0, s0, s0, v0
; SPLIT-NEXT:    v_add_co_ci_u32_e64 v1, s0, s1, 0, s0
; SPLIT-NEXT:    v_add_co_u32 v2, vcc_lo, v0, 12
; SPLIT-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v1, vcc_lo
; SPLIT-NEXT:    v_add_co_u32 v4, vcc_lo, v0, 4
; SPLIT-NEXT:    v_add_co_ci_u32_e32 v5, vcc_lo, 0, v1, vcc_lo
; SPLIT-NEXT:    v_add_co_u32 v6, vcc_lo, v0, 8
; SPLIT-NEXT:    v_add_co_ci_u32_e32 v7, vcc_lo, 0, v1, vcc_lo
; SPLIT-NEXT:    s_clause 0x3
; SPLIT-NEXT:    flat_load_dword v8, v[2:3]
; SPLIT-NEXT:    flat_load_dword v9, v[4:5]
; SPLIT-NEXT:    flat_load_dword v10, v[0:1]
; SPLIT-NEXT:    flat_load_dword v11, v[6:7]
; SPLIT-NEXT:    s_waitcnt vmcnt(2) lgkmcnt(2)
; SPLIT-NEXT:    flat_store_dword v[6:7], v9
; SPLIT-NEXT:    s_waitcnt vmcnt(1) lgkmcnt(2)
; SPLIT-NEXT:    flat_store_dword v[2:3], v10
; SPLIT-NEXT:    flat_store_dword v[0:1], v8
; SPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(3)
; SPLIT-NEXT:    flat_store_dword v[4:5], v11
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_flat_misaligned_v4:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_add_co_u32 v7, s0, s0, v0
; ALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v8, s0, s1, 0, s0
; ALIGNED-GFX10-NEXT:    flat_load_dwordx4 v[0:3], v[7:8]
; ALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v5, v1
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v6, v0
; ALIGNED-GFX10-NEXT:    flat_store_dwordx4 v[7:8], v[3:6]
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_flat_misaligned_v4:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_add_co_u32 v7, s0, s0, v0
; UNALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v8, s0, s1, 0, s0
; UNALIGNED-GFX10-NEXT:    flat_load_dwordx4 v[0:3], v[7:8]
; UNALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v5, v1
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v6, v0
; UNALIGNED-GFX10-NEXT:    flat_store_dwordx4 v[7:8], v[3:6]
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_flat_misaligned_v4:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_add_co_u32 v7, s0, s0, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v8, null, s1, 0, s0
; ALIGNED-GFX11-NEXT:    flat_load_b128 v[0:3], v[7:8]
; ALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v2
; ALIGNED-GFX11-NEXT:    v_mov_b32_e32 v6, v0
; ALIGNED-GFX11-NEXT:    flat_store_b128 v[7:8], v[3:6]
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_flat_misaligned_v4:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_add_co_u32 v7, s0, s0, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v8, null, s1, 0, s0
; UNALIGNED-GFX11-NEXT:    flat_load_b128 v[0:3], v[7:8]
; UNALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v2
; UNALIGNED-GFX11-NEXT:    v_mov_b32_e32 v6, v0
; UNALIGNED-GFX11-NEXT:    flat_store_b128 v[7:8], v[3:6]
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr %arg, i32 %lid
  %load = load <4 x i32>, ptr %gep, align 4
  %v1 = extractelement <4 x i32> %load, i32 0
  %v2 = extractelement <4 x i32> %load, i32 1
  %v3 = extractelement <4 x i32> %load, i32 2
  %v4 = extractelement <4 x i32> %load, i32 3
  %v5 = insertelement <4 x i32> poison, i32 %v4, i32 0
  %v6 = insertelement <4 x i32> %v5, i32 %v3, i32 1
  %v7 = insertelement <4 x i32> %v6, i32 %v2, i32 2
  %v8 = insertelement <4 x i32> %v7, i32 %v1, i32 3
  store <4 x i32> %v8, ptr %gep, align 4
  ret void
}

define amdgpu_kernel void @test_flat_misaligned_v3(ptr %arg) {
; SPLIT-LABEL: test_flat_misaligned_v3:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; SPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_add_co_u32 v0, s0, s0, v0
; SPLIT-NEXT:    v_add_co_ci_u32_e64 v1, s0, s1, 0, s0
; SPLIT-NEXT:    v_add_co_u32 v2, vcc_lo, v0, 4
; SPLIT-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v1, vcc_lo
; SPLIT-NEXT:    v_add_co_u32 v4, vcc_lo, v0, 8
; SPLIT-NEXT:    v_add_co_ci_u32_e32 v5, vcc_lo, 0, v1, vcc_lo
; SPLIT-NEXT:    s_clause 0x2
; SPLIT-NEXT:    flat_load_dword v6, v[2:3]
; SPLIT-NEXT:    flat_load_dword v7, v[4:5]
; SPLIT-NEXT:    flat_load_dword v8, v[0:1]
; SPLIT-NEXT:    s_waitcnt vmcnt(2) lgkmcnt(2)
; SPLIT-NEXT:    flat_store_dword v[4:5], v6
; SPLIT-NEXT:    s_waitcnt vmcnt(1) lgkmcnt(2)
; SPLIT-NEXT:    flat_store_dword v[0:1], v7
; SPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(2)
; SPLIT-NEXT:    flat_store_dword v[2:3], v8
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_flat_misaligned_v3:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_add_co_u32 v5, s0, s0, v0
; ALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v6, s0, s1, 0, s0
; ALIGNED-GFX10-NEXT:    flat_load_dwordx3 v[0:2], v[5:6]
; ALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v3, v0
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v1
; ALIGNED-GFX10-NEXT:    flat_store_dwordx3 v[5:6], v[2:4]
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_flat_misaligned_v3:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_add_co_u32 v5, s0, s0, v0
; UNALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v6, s0, s1, 0, s0
; UNALIGNED-GFX10-NEXT:    flat_load_dwordx3 v[0:2], v[5:6]
; UNALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v3, v0
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v1
; UNALIGNED-GFX10-NEXT:    flat_store_dwordx3 v[5:6], v[2:4]
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_flat_misaligned_v3:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_add_co_u32 v5, s0, s0, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v6, null, s1, 0, s0
; ALIGNED-GFX11-NEXT:    flat_load_b96 v[0:2], v[5:6]
; ALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_dual_mov_b32 v3, v0 :: v_dual_mov_b32 v4, v1
; ALIGNED-GFX11-NEXT:    flat_store_b96 v[5:6], v[2:4]
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_flat_misaligned_v3:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_add_co_u32 v5, s0, s0, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v6, null, s1, 0, s0
; UNALIGNED-GFX11-NEXT:    flat_load_b96 v[0:2], v[5:6]
; UNALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_dual_mov_b32 v3, v0 :: v_dual_mov_b32 v4, v1
; UNALIGNED-GFX11-NEXT:    flat_store_b96 v[5:6], v[2:4]
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr %arg, i32 %lid
  %load = load <3 x i32>, ptr %gep, align 4
  %v1 = extractelement <3 x i32> %load, i32 0
  %v2 = extractelement <3 x i32> %load, i32 1
  %v3 = extractelement <3 x i32> %load, i32 2
  %v5 = insertelement <3 x i32> poison, i32 %v3, i32 0
  %v6 = insertelement <3 x i32> %v5, i32 %v1, i32 1
  %v7 = insertelement <3 x i32> %v6, i32 %v2, i32 2
  store <3 x i32> %v7, ptr %gep, align 4
  ret void
}

define amdgpu_kernel void @test_local_aligned_v2(ptr addrspace(3) %arg) {
; SPLIT-LABEL: test_local_aligned_v2:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dword s0, s[4:5], 0x24
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_lshl_add_u32 v3, v0, 2, s0
; SPLIT-NEXT:    ds_read_b64 v[0:1], v3
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_mov_b32_e32 v2, v0
; SPLIT-NEXT:    ds_write_b64 v3, v[1:2]
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_local_aligned_v2:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_lshl_add_u32 v3, v0, 2, s0
; ALIGNED-GFX10-NEXT:    ds_read_b64 v[0:1], v3
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v2, v0
; ALIGNED-GFX10-NEXT:    ds_write_b64 v3, v[1:2]
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_local_aligned_v2:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_lshl_add_u32 v3, v0, 2, s0
; UNALIGNED-GFX10-NEXT:    ds_read_b64 v[0:1], v3
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v2, v0
; UNALIGNED-GFX10-NEXT:    ds_write_b64 v3, v[1:2]
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_local_aligned_v2:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshl_add_u32 v3, v0, 2, s0
; ALIGNED-GFX11-NEXT:    ds_load_b64 v[0:1], v3
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_mov_b32_e32 v2, v0
; ALIGNED-GFX11-NEXT:    ds_store_b64 v3, v[1:2]
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_local_aligned_v2:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshl_add_u32 v3, v0, 2, s0
; UNALIGNED-GFX11-NEXT:    ds_load_b64 v[0:1], v3
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_mov_b32_e32 v2, v0
; UNALIGNED-GFX11-NEXT:    ds_store_b64 v3, v[1:2]
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr addrspace(3) %arg, i32 %lid
  %load = load <2 x i32>, ptr addrspace(3) %gep, align 8
  %v1 = extractelement <2 x i32> %load, i32 0
  %v2 = extractelement <2 x i32> %load, i32 1
  %v3 = insertelement <2 x i32> poison, i32 %v2, i32 0
  %v4 = insertelement <2 x i32> %v3, i32 %v1, i32 1
  store <2 x i32> %v4, ptr addrspace(3) %gep, align 8
  ret void
}

define amdgpu_kernel void @test_local_aligned_v3(ptr addrspace(3) %arg) {
; SPLIT-LABEL: test_local_aligned_v3:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dword s0, s[4:5], 0x24
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_lshl_add_u32 v5, v0, 2, s0
; SPLIT-NEXT:    ds_read_b96 v[0:2], v5
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_mov_b32_e32 v3, v0
; SPLIT-NEXT:    v_mov_b32_e32 v4, v1
; SPLIT-NEXT:    ds_write_b96 v5, v[2:4]
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_local_aligned_v3:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_lshl_add_u32 v5, v0, 2, s0
; ALIGNED-GFX10-NEXT:    ds_read_b96 v[0:2], v5
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v3, v0
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v1
; ALIGNED-GFX10-NEXT:    ds_write_b96 v5, v[2:4]
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_local_aligned_v3:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_lshl_add_u32 v5, v0, 2, s0
; UNALIGNED-GFX10-NEXT:    ds_read_b96 v[0:2], v5
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v3, v0
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v1
; UNALIGNED-GFX10-NEXT:    ds_write_b96 v5, v[2:4]
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_local_aligned_v3:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshl_add_u32 v5, v0, 2, s0
; ALIGNED-GFX11-NEXT:    ds_load_b96 v[0:2], v5
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_dual_mov_b32 v3, v0 :: v_dual_mov_b32 v4, v1
; ALIGNED-GFX11-NEXT:    ds_store_b96 v5, v[2:4]
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_local_aligned_v3:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshl_add_u32 v5, v0, 2, s0
; UNALIGNED-GFX11-NEXT:    ds_load_b96 v[0:2], v5
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_dual_mov_b32 v3, v0 :: v_dual_mov_b32 v4, v1
; UNALIGNED-GFX11-NEXT:    ds_store_b96 v5, v[2:4]
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr addrspace(3) %arg, i32 %lid
  %load = load <3 x i32>, ptr addrspace(3) %gep, align 16
  %v1 = extractelement <3 x i32> %load, i32 0
  %v2 = extractelement <3 x i32> %load, i32 1
  %v3 = extractelement <3 x i32> %load, i32 2
  %v5 = insertelement <3 x i32> poison, i32 %v3, i32 0
  %v6 = insertelement <3 x i32> %v5, i32 %v1, i32 1
  %v7 = insertelement <3 x i32> %v6, i32 %v2, i32 2
  store <3 x i32> %v7, ptr addrspace(3) %gep, align 16
  ret void
}

define amdgpu_kernel void @test_flat_aligned_v2(ptr %arg) {
; SPLIT-LABEL: test_flat_aligned_v2:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; SPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_add_co_u32 v0, s0, s0, v0
; SPLIT-NEXT:    v_add_co_ci_u32_e64 v1, s0, s1, 0, s0
; SPLIT-NEXT:    flat_load_dwordx2 v[2:3], v[0:1]
; SPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; SPLIT-NEXT:    v_mov_b32_e32 v4, v2
; SPLIT-NEXT:    flat_store_dwordx2 v[0:1], v[3:4]
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_flat_aligned_v2:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_add_co_u32 v0, s0, s0, v0
; ALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v1, s0, s1, 0, s0
; ALIGNED-GFX10-NEXT:    flat_load_dwordx2 v[2:3], v[0:1]
; ALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; ALIGNED-GFX10-NEXT:    flat_store_dwordx2 v[0:1], v[3:4]
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_flat_aligned_v2:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_add_co_u32 v0, s0, s0, v0
; UNALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v1, s0, s1, 0, s0
; UNALIGNED-GFX10-NEXT:    flat_load_dwordx2 v[2:3], v[0:1]
; UNALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; UNALIGNED-GFX10-NEXT:    flat_store_dwordx2 v[0:1], v[3:4]
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_flat_aligned_v2:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_add_co_u32 v0, s0, s0, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v1, null, s1, 0, s0
; ALIGNED-GFX11-NEXT:    flat_load_b64 v[2:3], v[0:1]
; ALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_mov_b32_e32 v4, v2
; ALIGNED-GFX11-NEXT:    flat_store_b64 v[0:1], v[3:4]
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_flat_aligned_v2:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_add_co_u32 v0, s0, s0, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v1, null, s1, 0, s0
; UNALIGNED-GFX11-NEXT:    flat_load_b64 v[2:3], v[0:1]
; UNALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_mov_b32_e32 v4, v2
; UNALIGNED-GFX11-NEXT:    flat_store_b64 v[0:1], v[3:4]
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr %arg, i32 %lid
  %load = load <2 x i32>, ptr %gep, align 8
  %v1 = extractelement <2 x i32> %load, i32 0
  %v2 = extractelement <2 x i32> %load, i32 1
  %v3 = insertelement <2 x i32> poison, i32 %v2, i32 0
  %v4 = insertelement <2 x i32> %v3, i32 %v1, i32 1
  store <2 x i32> %v4, ptr %gep, align 8
  ret void
}

define amdgpu_kernel void @test_flat_aligned_v4(ptr %arg) {
; SPLIT-LABEL: test_flat_aligned_v4:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; SPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_add_co_u32 v7, s0, s0, v0
; SPLIT-NEXT:    v_add_co_ci_u32_e64 v8, s0, s1, 0, s0
; SPLIT-NEXT:    flat_load_dwordx4 v[0:3], v[7:8]
; SPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; SPLIT-NEXT:    v_mov_b32_e32 v4, v2
; SPLIT-NEXT:    v_mov_b32_e32 v5, v1
; SPLIT-NEXT:    v_mov_b32_e32 v6, v0
; SPLIT-NEXT:    flat_store_dwordx4 v[7:8], v[3:6]
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_flat_aligned_v4:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_add_co_u32 v7, s0, s0, v0
; ALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v8, s0, s1, 0, s0
; ALIGNED-GFX10-NEXT:    flat_load_dwordx4 v[0:3], v[7:8]
; ALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v5, v1
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v6, v0
; ALIGNED-GFX10-NEXT:    flat_store_dwordx4 v[7:8], v[3:6]
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_flat_aligned_v4:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_add_co_u32 v7, s0, s0, v0
; UNALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v8, s0, s1, 0, s0
; UNALIGNED-GFX10-NEXT:    flat_load_dwordx4 v[0:3], v[7:8]
; UNALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v5, v1
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v6, v0
; UNALIGNED-GFX10-NEXT:    flat_store_dwordx4 v[7:8], v[3:6]
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_flat_aligned_v4:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_add_co_u32 v7, s0, s0, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v8, null, s1, 0, s0
; ALIGNED-GFX11-NEXT:    flat_load_b128 v[0:3], v[7:8]
; ALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v2
; ALIGNED-GFX11-NEXT:    v_mov_b32_e32 v6, v0
; ALIGNED-GFX11-NEXT:    flat_store_b128 v[7:8], v[3:6]
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_flat_aligned_v4:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_add_co_u32 v7, s0, s0, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v8, null, s1, 0, s0
; UNALIGNED-GFX11-NEXT:    flat_load_b128 v[0:3], v[7:8]
; UNALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v2
; UNALIGNED-GFX11-NEXT:    v_mov_b32_e32 v6, v0
; UNALIGNED-GFX11-NEXT:    flat_store_b128 v[7:8], v[3:6]
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr %arg, i32 %lid
  %load = load <4 x i32>, ptr %gep, align 16
  %v1 = extractelement <4 x i32> %load, i32 0
  %v2 = extractelement <4 x i32> %load, i32 1
  %v3 = extractelement <4 x i32> %load, i32 2
  %v4 = extractelement <4 x i32> %load, i32 3
  %v5 = insertelement <4 x i32> poison, i32 %v4, i32 0
  %v6 = insertelement <4 x i32> %v5, i32 %v3, i32 1
  %v7 = insertelement <4 x i32> %v6, i32 %v2, i32 2
  %v8 = insertelement <4 x i32> %v7, i32 %v1, i32 3
  store <4 x i32> %v8, ptr %gep, align 16
  ret void
}

define amdgpu_kernel void @test_local_v4_aligned8(ptr addrspace(3) %arg) {
; SPLIT-LABEL: test_local_v4_aligned8:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dword s0, s[4:5], 0x24
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_lshl_add_u32 v6, v0, 2, s0
; SPLIT-NEXT:    ds_read2_b64 v[0:3], v6 offset1:1
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_mov_b32_e32 v4, v1
; SPLIT-NEXT:    v_mov_b32_e32 v5, v0
; SPLIT-NEXT:    v_mov_b32_e32 v1, v3
; SPLIT-NEXT:    ds_write2_b64 v6, v[1:2], v[4:5] offset1:1
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_local_v4_aligned8:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_lshl_add_u32 v5, v0, 2, s0
; ALIGNED-GFX10-NEXT:    ds_read2_b64 v[0:3], v5 offset1:1
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v2, v0
; ALIGNED-GFX10-NEXT:    ds_write2_b64 v5, v[3:4], v[1:2] offset1:1
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_local_v4_aligned8:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dword s0, s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_lshl_add_u32 v5, v0, 2, s0
; UNALIGNED-GFX10-NEXT:    ds_read2_b64 v[0:3], v5 offset1:1
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v2, v0
; UNALIGNED-GFX10-NEXT:    ds_write2_b64 v5, v[3:4], v[1:2] offset1:1
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_local_v4_aligned8:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshl_add_u32 v5, v0, 2, s0
; ALIGNED-GFX11-NEXT:    ds_load_2addr_b64 v[0:3], v5 offset1:1
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_mov_b32_e32 v4, v2
; ALIGNED-GFX11-NEXT:    v_mov_b32_e32 v2, v0
; ALIGNED-GFX11-NEXT:    ds_store_2addr_b64 v5, v[3:4], v[1:2] offset1:1
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_local_v4_aligned8:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b32 s0, s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshl_add_u32 v5, v0, 2, s0
; UNALIGNED-GFX11-NEXT:    ds_load_2addr_b64 v[0:3], v5 offset1:1
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_mov_b32_e32 v4, v2
; UNALIGNED-GFX11-NEXT:    v_mov_b32_e32 v2, v0
; UNALIGNED-GFX11-NEXT:    ds_store_2addr_b64 v5, v[3:4], v[1:2] offset1:1
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr addrspace(3) %arg, i32 %lid
  %load = load <4 x i32>, ptr addrspace(3) %gep, align 8
  %v1 = extractelement <4 x i32> %load, i32 0
  %v2 = extractelement <4 x i32> %load, i32 1
  %v3 = extractelement <4 x i32> %load, i32 2
  %v4 = extractelement <4 x i32> %load, i32 3
  %v5 = insertelement <4 x i32> poison, i32 %v4, i32 0
  %v6 = insertelement <4 x i32> %v5, i32 %v3, i32 1
  %v7 = insertelement <4 x i32> %v6, i32 %v2, i32 2
  %v8 = insertelement <4 x i32> %v7, i32 %v1, i32 3
  store <4 x i32> %v8, ptr addrspace(3) %gep, align 8
  ret void
}

define amdgpu_kernel void @test_flat_v4_aligned8(ptr %arg) {
; SPLIT-LABEL: test_flat_v4_aligned8:
; SPLIT:       ; %bb.0: ; %bb
; SPLIT-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; SPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; SPLIT-NEXT:    s_waitcnt lgkmcnt(0)
; SPLIT-NEXT:    v_add_co_u32 v0, s0, s0, v0
; SPLIT-NEXT:    v_add_co_ci_u32_e64 v1, s0, s1, 0, s0
; SPLIT-NEXT:    v_add_co_u32 v2, vcc_lo, v0, 8
; SPLIT-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v1, vcc_lo
; SPLIT-NEXT:    s_clause 0x1
; SPLIT-NEXT:    flat_load_dwordx2 v[4:5], v[0:1]
; SPLIT-NEXT:    flat_load_dwordx2 v[6:7], v[2:3]
; SPLIT-NEXT:    s_waitcnt vmcnt(1) lgkmcnt(1)
; SPLIT-NEXT:    v_mov_b32_e32 v8, v5
; SPLIT-NEXT:    v_mov_b32_e32 v9, v4
; SPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; SPLIT-NEXT:    v_mov_b32_e32 v4, v7
; SPLIT-NEXT:    v_mov_b32_e32 v5, v6
; SPLIT-NEXT:    flat_store_dwordx2 v[2:3], v[8:9]
; SPLIT-NEXT:    flat_store_dwordx2 v[0:1], v[4:5]
; SPLIT-NEXT:    s_endpgm
;
; ALIGNED-GFX10-LABEL: test_flat_v4_aligned8:
; ALIGNED-GFX10:       ; %bb.0: ; %bb
; ALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; ALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_add_co_u32 v7, s0, s0, v0
; ALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v8, s0, s1, 0, s0
; ALIGNED-GFX10-NEXT:    flat_load_dwordx4 v[0:3], v[7:8]
; ALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v5, v1
; ALIGNED-GFX10-NEXT:    v_mov_b32_e32 v6, v0
; ALIGNED-GFX10-NEXT:    flat_store_dwordx4 v[7:8], v[3:6]
; ALIGNED-GFX10-NEXT:    s_endpgm
;
; UNALIGNED-GFX10-LABEL: test_flat_v4_aligned8:
; UNALIGNED-GFX10:       ; %bb.0: ; %bb
; UNALIGNED-GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_add_co_u32 v7, s0, s0, v0
; UNALIGNED-GFX10-NEXT:    v_add_co_ci_u32_e64 v8, s0, s1, 0, s0
; UNALIGNED-GFX10-NEXT:    flat_load_dwordx4 v[0:3], v[7:8]
; UNALIGNED-GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v4, v2
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v5, v1
; UNALIGNED-GFX10-NEXT:    v_mov_b32_e32 v6, v0
; UNALIGNED-GFX10-NEXT:    flat_store_dwordx4 v[7:8], v[3:6]
; UNALIGNED-GFX10-NEXT:    s_endpgm
;
; ALIGNED-GFX11-LABEL: test_flat_v4_aligned8:
; ALIGNED-GFX11:       ; %bb.0: ; %bb
; ALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; ALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; ALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_add_co_u32 v7, s0, s0, v0
; ALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; ALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v8, null, s1, 0, s0
; ALIGNED-GFX11-NEXT:    flat_load_b128 v[0:3], v[7:8]
; ALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; ALIGNED-GFX11-NEXT:    v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v2
; ALIGNED-GFX11-NEXT:    v_mov_b32_e32 v6, v0
; ALIGNED-GFX11-NEXT:    flat_store_b128 v[7:8], v[3:6]
; ALIGNED-GFX11-NEXT:    s_endpgm
;
; UNALIGNED-GFX11-LABEL: test_flat_v4_aligned8:
; UNALIGNED-GFX11:       ; %bb.0: ; %bb
; UNALIGNED-GFX11-NEXT:    s_load_b64 s[0:1], s[4:5], 0x24
; UNALIGNED-GFX11-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; UNALIGNED-GFX11-NEXT:    s_waitcnt lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_add_co_u32 v7, s0, s0, v0
; UNALIGNED-GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
; UNALIGNED-GFX11-NEXT:    v_add_co_ci_u32_e64 v8, null, s1, 0, s0
; UNALIGNED-GFX11-NEXT:    flat_load_b128 v[0:3], v[7:8]
; UNALIGNED-GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; UNALIGNED-GFX11-NEXT:    v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v2
; UNALIGNED-GFX11-NEXT:    v_mov_b32_e32 v6, v0
; UNALIGNED-GFX11-NEXT:    flat_store_b128 v[7:8], v[3:6]
; UNALIGNED-GFX11-NEXT:    s_endpgm
bb:
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr %arg, i32 %lid
  %load = load <4 x i32>, ptr %gep, align 8
  %v1 = extractelement <4 x i32> %load, i32 0
  %v2 = extractelement <4 x i32> %load, i32 1
  %v3 = extractelement <4 x i32> %load, i32 2
  %v4 = extractelement <4 x i32> %load, i32 3
  %v5 = insertelement <4 x i32> poison, i32 %v4, i32 0
  %v6 = insertelement <4 x i32> %v5, i32 %v3, i32 1
  %v7 = insertelement <4 x i32> %v6, i32 %v2, i32 2
  %v8 = insertelement <4 x i32> %v7, i32 %v1, i32 3
  store <4 x i32> %v8, ptr %gep, align 8
  ret void
}

declare i32 @llvm.amdgcn.workitem.id.x()