llvm/test/CodeGen/AMDGPU/permute_i8.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=amdgpu10.10 < %s | FileCheck %s -check-prefixes=GFX10
; RUN: llc -mtriple=amdgpu9.08  < %s | FileCheck %s -check-prefixes=GFX9

define hidden void @shuffle6766(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle6766:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x6060706
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle6766:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x6060706
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 6, i32 7, i32 6, i32 6>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shuffle3744(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle3744:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v6, v7, 0x307
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle3744:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_movk_i32 s4, 0x307
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v6, v7, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 3, i32 7, i32 4, i32 4>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shuffle4445(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle4445:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x5040404
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle4445:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x5040404
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 4, i32 4, i32 4, i32 5>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shuffle0101(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle0101:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x5040504
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle0101:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x5040504
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shuffle1004(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle1004:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v6, v7, 0x40405
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle1004:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x40405
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v6, v7, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 1, i32 0, i32 0, i32 4>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}



define hidden void @shuffle7533(ptr addrspace(0) %in0, ptr addrspace(0) %in1, ptr addrspace(0) %out0) {
; GFX10-LABEL: shuffle7533:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    flat_load_dword v6, v[0:1]
; GFX10-NEXT:    flat_load_dword v7, v[2:3]
; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x3030507
; GFX10-NEXT:    flat_store_dword v[4:5], v0
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle7533:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    flat_load_dword v6, v[0:1]
; GFX9-NEXT:    flat_load_dword v7, v[2:3]
; GFX9-NEXT:    s_mov_b32 s4, 0x3030507
; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    flat_store_dword v[4:5], v0
; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(0) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(0) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 7, i32 5, i32 3, i32 3>
  store <4 x i8> %shuffle0_0, ptr addrspace(0) %out0, align 4
  ret void
}

define hidden void @shuffle7767(ptr addrspace(0) %in0, ptr addrspace(0) %in1, ptr addrspace(0) %out0) {
; GFX10-LABEL: shuffle7767:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    flat_load_dword v0, v[2:3]
; GFX10-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x7060707
; GFX10-NEXT:    flat_store_dword v[4:5], v0
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle7767:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    flat_load_dword v0, v[2:3]
; GFX9-NEXT:    s_mov_b32 s4, 0x7060707
; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    flat_store_dword v[4:5], v0
; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(0) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(0) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 7, i32 7, i32 6, i32 7>
  store <4 x i8> %shuffle0_0, ptr addrspace(0) %out0, align 4
  ret void
}

define hidden void @shuffle0554(ptr addrspace(3) %in0, ptr addrspace(3) %in1, ptr addrspace(3) %out0) {
; GFX10-LABEL: shuffle0554:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    ds_read_b32 v0, v0
; GFX10-NEXT:    ds_read_b32 v1, v1
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v1, 0x10104
; GFX10-NEXT:    ds_write_b32 v2, v0
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle0554:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    ds_read_b32 v0, v0
; GFX9-NEXT:    ds_read_b32 v1, v1
; GFX9-NEXT:    s_mov_b32 s4, 0x10104
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v1, s4
; GFX9-NEXT:    ds_write_b32 v2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(3) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(3) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 0, i32 5, i32 5, i32 4>
  store <4 x i8> %shuffle0_0, ptr addrspace(3) %out0, align 4
  ret void
}

define hidden void @shuffle2127(ptr addrspace(3) %in0, ptr addrspace(3) %in1, ptr addrspace(3) %out0) {
; GFX10-LABEL: shuffle2127:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    ds_read_b32 v0, v0
; GFX10-NEXT:    ds_read_b32 v1, v1
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v1, 0x3060506
; GFX10-NEXT:    ds_write_b32 v2, v0
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle2127:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    ds_read_b32 v0, v0
; GFX9-NEXT:    ds_read_b32 v1, v1
; GFX9-NEXT:    s_mov_b32 s4, 0x3060506
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v1, s4
; GFX9-NEXT:    ds_write_b32 v2, v0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(3) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(3) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 2, i32 1, i32 2, i32 7>
  store <4 x i8> %shuffle0_0, ptr addrspace(3) %out0, align 4
  ret void
}

define hidden void @shuffle5047(ptr addrspace(5) %in0, ptr addrspace(5) %in1, ptr addrspace(5) %out0) {
; GFX10-LABEL: shuffle5047:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    s_clause 0x1
; GFX10-NEXT:    buffer_load_dword v3, v0, s[0:3], 0 offen
; GFX10-NEXT:    buffer_load_dword v4, v1, s[0:3], 0 offen
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v4, v3, 0x7040005
; GFX10-NEXT:    buffer_store_dword v0, v2, s[0:3], 0 offen
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle5047:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    buffer_load_dword v3, v0, s[0:3], 0 offen
; GFX9-NEXT:    buffer_load_dword v4, v1, s[0:3], 0 offen
; GFX9-NEXT:    s_mov_b32 s4, 0x7040005
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v4, v3, s4
; GFX9-NEXT:    buffer_store_dword v0, v2, s[0:3], 0 offen
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(5) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(5) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 5, i32 0, i32 4, i32 7>
  store <4 x i8> %shuffle0_0, ptr addrspace(5) %out0, align 4
  ret void
}

define hidden void @shuffle3546(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle3546:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v6, v7, 0x2000107
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle3546:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x2000107
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v6, v7, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 3, i32 5, i32 4, i32 6>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}


define hidden void @shuffle7330ud2(ptr addrspace(1) %in0, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle7330ud2:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x4070706
; GFX10-NEXT:    global_store_dword v[2:3], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle7330ud2:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x4070706
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 7, i32 3, i32 3, i32 0>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shuffle5341ud2(ptr addrspace(1) %in0, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle5341ud2:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_alignbit_b32 v0, v0, v0, 16
; GFX10-NEXT:    global_store_dword v[2:3], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle5341ud2:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_alignbit_b32 v0, v0, v0, 16
; GFX9-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 5, i32 3, i32 4, i32 1>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shuffle6106ud2(ptr addrspace(1) %in0, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle6106ud2:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x5040504
; GFX10-NEXT:    global_store_dword v[2:3], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle6106ud2:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x5040504
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 6, i32 1, i32 0, i32 6>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}


define hidden void @shuffle4327ud2(ptr addrspace(1) %in0, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle4327ud2:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x7060706
; GFX10-NEXT:    global_store_dword v[2:3], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle4327ud2:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x7060706
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 4, i32 3, i32 2, i32 7>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shuffle3263ud2(ptr addrspace(1) %in0, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle3263ud2:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x7060607
; GFX10-NEXT:    global_store_dword v[2:3], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle3263ud2:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x7060607
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 6, i32 3>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shuffle2763ud2(ptr addrspace(1) %in0, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle2763ud2:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x7060706
; GFX10-NEXT:    global_store_dword v[2:3], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle2763ud2:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x7060706
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 2, i32 7, i32 6, i32 3>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shuffle1327ud2(ptr addrspace(1) %in0, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle1327ud2:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x7060705
; GFX10-NEXT:    global_store_dword v[2:3], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle1327ud2:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x7060705
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 1, i32 3, i32 2, i32 7>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shuffle0605ud2(ptr addrspace(1) %in0, ptr addrspace(1) %out0) {
; GFX10-LABEL: shuffle0605ud2:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x5040504
; GFX10-NEXT:    global_store_dword v[2:3], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shuffle0605ud2:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x5040504
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 0, i32 6, i32 0, i32 5>
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @insertUsesOr(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0) {
; GFX10-LABEL: insertUsesOr:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v4, 0xc0c0005
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: insertUsesOr:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0xc0c0005
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v4, s4
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 4>
  %vecins = insertelement <4 x i8> %shuffle0_0, i8 %elt, i32 1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  ret void
}

define hidden void @addUsesOr(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0) {
; GFX10-LABEL: addUsesOr:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 24, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshrrev_b32_e32 v1, 24, v7
; GFX10-NEXT:    v_add_nc_u16 v2, v4, v7 op_sel:[1,1,0]
; GFX10-NEXT:    v_add_nc_u16 v0, v0, v1
; GFX10-NEXT:    v_lshrrev_b16 v1, 8, v7
; GFX10-NEXT:    v_lshlrev_b16 v0, 8, v0
; GFX10-NEXT:    v_add_nc_u16 v1, v4, v1
; GFX10-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_lshlrev_b16 v1, 8, v1
; GFX10-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: addUsesOr:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_u16_sdwa v0, v4, v7 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
; GFX9-NEXT:    v_add_u16_sdwa v1, v4, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
; GFX9-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_add_u16_sdwa v1, v4, v7 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1
; GFX9-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 7, i32 0, i32 6, i32 3>
  %added = add <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %added, ptr addrspace(1) %out0
  ret void
}


define amdgpu_kernel void @shuffle8i8(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out1) #0 {
; GFX10-LABEL: shuffle8i8:
; GFX10:       ; %bb.0: ; %bb
; GFX10-NEXT:    s_clause 0x1
; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX10-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
; GFX10-NEXT:    v_mov_b32_e32 v0, 0xc0c0106
; GFX10-NEXT:    v_mov_b32_e32 v1, 0x6050400
; GFX10-NEXT:    v_mov_b32_e32 v2, 0xc0c0400
; GFX10-NEXT:    v_mov_b32_e32 v5, 0
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
; GFX10-NEXT:    s_load_dwordx2 s[8:9], s[2:3], 0x0
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    s_lshr_b32 s0, s5, 8
; GFX10-NEXT:    v_perm_b32 v0, s4, s8, v0
; GFX10-NEXT:    v_perm_b32 v1, s9, s0, v1
; GFX10-NEXT:    s_lshr_b32 s0, s9, 16
; GFX10-NEXT:    s_lshr_b32 s1, s4, 8
; GFX10-NEXT:    v_perm_b32 v3, s0, s4, v2
; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
; GFX10-NEXT:    v_perm_b32 v2, s8, s1, v2
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 16, v1
; GFX10-NEXT:    v_or_b32_e32 v1, v3, v0
; GFX10-NEXT:    v_or_b32_e32 v0, v2, v4
; GFX10-NEXT:    global_store_dwordx2 v5, v[0:1], s[6:7]
; GFX10-NEXT:    s_endpgm
;
; GFX9-LABEL: shuffle8i8:
; GFX9:       ; %bb.0: ; %bb
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
; GFX9-NEXT:    v_mov_b32_e32 v0, 0x6050400
; GFX9-NEXT:    v_mov_b32_e32 v5, 0xc0c0106
; GFX9-NEXT:    v_mov_b32_e32 v3, 0xc0c0400
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x0
; GFX9-NEXT:    v_mov_b32_e32 v2, 0
; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    s_lshr_b32 s1, s1, 8
; GFX9-NEXT:    v_mov_b32_e32 v4, s1
; GFX9-NEXT:    v_perm_b32 v0, s3, v4, v0
; GFX9-NEXT:    v_mov_b32_e32 v4, s2
; GFX9-NEXT:    s_lshr_b32 s5, s3, 16
; GFX9-NEXT:    v_mov_b32_e32 v1, s0
; GFX9-NEXT:    v_perm_b32 v4, s0, v4, v5
; GFX9-NEXT:    s_lshr_b32 s4, s0, 8
; GFX9-NEXT:    v_perm_b32 v1, s5, v1, v3
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 16, v4
; GFX9-NEXT:    v_or_b32_e32 v1, v1, v4
; GFX9-NEXT:    v_mov_b32_e32 v4, s4
; GFX9-NEXT:    v_perm_b32 v3, s2, v4, v3
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
; GFX9-NEXT:    v_or_b32_e32 v0, v3, v0
; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[6:7]
; GFX9-NEXT:    s_endpgm
bb:
  %vec0 = load <8 x i8>, ptr addrspace(1) %in0
  %vec1 = load <8 x i8>, ptr addrspace(1) %in1
  %shuffle0 = shufflevector <8 x i8> %vec0, <8 x i8> %vec1, <8 x i32> <i32 1, i32 8, i32 5, i32 12, i32 0, i32 14, i32 2, i32 9>
  store <8 x i8> %shuffle0, ptr addrspace(1) %out1
  ret void
}

declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
declare i32 @llvm.amdgcn.workitem.id.y() nounwind readnone

; Not combined to perm due to non-vectorized use, non-divergent
define hidden void @add(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0) {
; GFX10-LABEL: add:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 24, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshrrev_b32_e32 v1, 24, v7
; GFX10-NEXT:    v_lshrrev_b16 v2, 8, v7
; GFX10-NEXT:    v_lshrrev_b16 v3, 8, v4
; GFX10-NEXT:    v_add_nc_u16 v1, v7, v1
; GFX10-NEXT:    v_add_nc_u16 v0, v0, v2
; GFX10-NEXT:    v_add_nc_u16 v2, v2, v7 op_sel:[0,1,0]
; GFX10-NEXT:    v_add_nc_u16 v3, v3, v7
; GFX10-NEXT:    v_lshlrev_b16 v1, 8, v1
; GFX10-NEXT:    v_lshlrev_b16 v0, 8, v0
; GFX10-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: add:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v4, v[2:3], off
; GFX9-NEXT:    global_load_dword v7, v[0:1], off
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_add_u16_sdwa v0, v4, v4 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
; GFX9-NEXT:    v_add_u16_sdwa v1, v4, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:WORD_1
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_u16_sdwa v2, v7, v4 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_1
; GFX9-NEXT:    v_add_u16_sdwa v3, v7, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 1, i32 3, i32 5, i32 4>
  %vecins = add <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  ret void
}

; Not combined to perm due to non-vectorized use
define hidden void @add_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0) {
; GFX10-LABEL: add_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 24, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshrrev_b16 v1, 8, v7
; GFX10-NEXT:    v_add_nc_u16 v0, v0, v1
; GFX10-NEXT:    v_lshrrev_b16 v1, 8, v4
; GFX10-NEXT:    v_lshlrev_b16 v0, 8, v0
; GFX10-NEXT:    v_add_nc_u16 v1, v1, v7
; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: add_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_u16_sdwa v0, v4, v7 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_1
; GFX9-NEXT:    v_add_u16_sdwa v1, v4, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 4>
  %vecins = add <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  ret void
}

; Not combined to perm due to non-divergent use
define hidden void @add_store(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: add_store:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v9, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 24, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshrrev_b16 v1, 8, v9
; GFX10-NEXT:    v_lshrrev_b16 v2, 8, v4
; GFX10-NEXT:    v_add_nc_u16 v0, v0, v1
; GFX10-NEXT:    v_mov_b32_e32 v1, 0xffffff00
; GFX10-NEXT:    v_add_nc_u16 v3, v2, v9
; GFX10-NEXT:    v_lshlrev_b16 v0, 8, v0
; GFX10-NEXT:    v_and_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_e32 v1, v2, v1
; GFX10-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; GFX10-NEXT:    v_and_b32_e32 v1, 0xffff, v1
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: add_store:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_movk_i32 s4, 0xff00
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_and_b32_sdwa v0, v4, s4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_u16_sdwa v1, v4, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_1
; GFX9-NEXT:    v_add_u16_sdwa v2, v4, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX9-NEXT:    v_and_b32_e32 v1, 0xffff, v1
; GFX9-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT:    global_store_dword v[5:6], v1, off
; GFX9-NEXT:    global_store_dword v[7:8], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec0 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 4>
  %vecins = add <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}

; Not combined to perm due to 16 bit or
define hidden void @add_store_div_16(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: add_store_div_16:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v9, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 24, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshrrev_b16 v1, 8, v9
; GFX10-NEXT:    v_lshrrev_b16 v2, 8, v4
; GFX10-NEXT:    v_add_nc_u16 v0, v0, v1
; GFX10-NEXT:    v_mov_b32_e32 v1, 0xffffff00
; GFX10-NEXT:    v_add_nc_u16 v3, v2, v9
; GFX10-NEXT:    v_lshlrev_b16 v0, 8, v0
; GFX10-NEXT:    v_and_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_e32 v1, v2, v1
; GFX10-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; GFX10-NEXT:    v_and_b32_e32 v1, 0xffff, v1
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: add_store_div_16:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_movk_i32 s4, 0xff00
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_and_b32_sdwa v0, v4, s4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_add_u16_sdwa v1, v4, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_1
; GFX9-NEXT:    v_add_u16_sdwa v2, v4, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX9-NEXT:    v_and_b32_e32 v1, 0xffff, v1
; GFX9-NEXT:    v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT:    global_store_dword v[5:6], v1, off
; GFX9-NEXT:    global_store_dword v[7:8], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 4>
  %vecins = add <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}

; Vectorized use, divergent, 32 bit or
define hidden void @add_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: add_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v9, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 24, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshrrev_b32_e32 v1, 24, v9
; GFX10-NEXT:    v_lshrrev_b16 v2, 8, v9
; GFX10-NEXT:    v_lshrrev_b16 v3, 8, v4
; GFX10-NEXT:    v_add_nc_u16 v1, v9, v1
; GFX10-NEXT:    v_add_nc_u16 v0, v0, v2
; GFX10-NEXT:    v_add_nc_u16 v2, v2, v9 op_sel:[0,1,0]
; GFX10-NEXT:    v_add_nc_u16 v3, v3, v9
; GFX10-NEXT:    v_lshlrev_b16 v1, 8, v1
; GFX10-NEXT:    v_lshlrev_b16 v0, 8, v0
; GFX10-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    v_perm_b32 v1, v4, v9, 0x10705
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: add_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x10705
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v4, v9, s4
; GFX9-NEXT:    v_add_u16_sdwa v1, v9, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
; GFX9-NEXT:    v_add_u16_sdwa v2, v9, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:WORD_1
; GFX9-NEXT:    v_add_u16_sdwa v3, v4, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_1
; GFX9-NEXT:    v_add_u16_sdwa v4, v4, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v2, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v1, off
; GFX9-NEXT:    global_store_dword v[7:8], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 1, i32 3, i32 5, i32 4>
  %vecins = add <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}

define hidden void @and_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: and_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v9, v[2:3], off
; GFX10-NEXT:    v_mov_b32_e32 v0, 2
; GFX10-NEXT:    v_mov_b32_e32 v1, 1
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_and_b32_e32 v2, 0x100, v4
; GFX10-NEXT:    v_and_b32_sdwa v3, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:DWORD
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_and_b32_sdwa v0, v9, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT:    v_and_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_e32 v0, v1, v0
; GFX10-NEXT:    v_perm_b32 v1, v4, v9, 0x5070006
; GFX10-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: and_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    v_mov_b32_e32 v0, 2
; GFX9-NEXT:    v_mov_b32_e32 v1, 1
; GFX9-NEXT:    s_mov_b32 s4, 0x5070006
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_and_b32_e32 v3, 0x100, v4
; GFX9-NEXT:    v_and_b32_sdwa v10, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:DWORD
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_and_b32_sdwa v0, v9, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT:    v_and_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v3, v10, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_e32 v0, v1, v0
; GFX9-NEXT:    v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    v_perm_b32 v2, v4, v9, s4
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    global_store_dword v[7:8], v2, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 2, i32 4, i32 3, i32 1>
  %vecins = and <4 x i8> %shuffle0_0, <i8 1, i8 2, i8 2, i8 1>
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}

define hidden void @ashr_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: ashr_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v9, v[0:1], off
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_mov_b32_e32 v2, 26
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_bfe_i32 v1, v9, 0, 8
; GFX10-NEXT:    v_ashrrev_i32_sdwa v2, v2, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT:    v_ashrrev_i32_e32 v3, 25, v9
; GFX10-NEXT:    v_lshlrev_b16 v1, 7, v1
; GFX10-NEXT:    v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_ashrrev_i16 v4, 10, v0
; GFX10-NEXT:    v_perm_b32 v0, v9, v0, 0x4010707
; GFX10-NEXT:    v_and_b32_e32 v1, 0xffffff00, v1
; GFX10-NEXT:    v_or_b32_sdwa v1, v4, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    global_store_dword v[5:6], v1, off
; GFX10-NEXT:    global_store_dword v[7:8], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: ashr_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    v_mov_b32_e32 v1, 7
; GFX9-NEXT:    s_mov_b32 s4, 0x4010707
; GFX9-NEXT:    v_mov_b32_e32 v0, 26
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_lshlrev_b16_sdwa v1, v1, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v2, v4, v9, s4
; GFX9-NEXT:    v_ashrrev_i32_sdwa v0, v0, v4 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT:    v_ashrrev_i32_e32 v3, 25, v4
; GFX9-NEXT:    v_ashrrev_i16_e32 v9, 10, v9
; GFX9-NEXT:    v_and_b32_e32 v1, 0xffffff00, v1
; GFX9-NEXT:    v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v9, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    global_store_dword v[7:8], v2, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 3, i32 3, i32 5, i32 0>
  %vecins = ashr <4 x i8> %shuffle0_0, <i8 1, i8 2, i8 2, i8 1>
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}

define hidden void @bc_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: bc_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v9, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v9, v4, 0x7060104
; GFX10-NEXT:    global_store_dword v[7:8], v0, off
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: bc_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x7060104
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v9, v4, s4
; GFX9-NEXT:    global_store_dword v[7:8], v0, off
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 4, i32 1, i32 6, i32 7>
  %insvec = bitcast <4 x i8> %shuffle0_0 to i32
  store i32 %insvec, ptr addrspace(1) %out1
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0
  ret void
}


define hidden void @eve_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1, ptr addrspace(1) %out2) {
; GFX10-LABEL: eve_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v5, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 24, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v1, v5, v4, 0x1020305
; GFX10-NEXT:    global_store_byte v[9:10], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: eve_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v5, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x1020305
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_lshrrev_b32_e32 v1, 24, v4
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v5, v4, s4
; GFX9-NEXT:    global_store_byte v[9:10], v1, off
; GFX9-NEXT:    global_store_dword v[7:8], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 5, i32 3, i32 2, i32 1>
  %tmp = extractelement <4 x i8> %shuffle0_0, i32 1
  store i8 %tmp, ptr addrspace(1) %out2
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}

; Not combined to perm due to multi use of or operands (introduced by insert op)
define hidden void @ive_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: ive_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v9, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v9, 2, v9
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v9
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v9
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v9, v[0:1], off
; GFX10-NEXT:    global_load_dword v10, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshlrev_b16 v0, 8, v9 op_sel:[0,1,0]
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v1, v10, v4, 0xc0c0006
; GFX10-NEXT:    v_or_b32_sdwa v0, v9, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
; GFX10-NEXT:    v_perm_b32 v1, v10, v9, 0x2000706
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: ive_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v9, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v9, 2, v9
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v9
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v9
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v9, v[0:1], off
; GFX9-NEXT:    global_load_dword v10, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s5, 0xc0c0006
; GFX9-NEXT:    s_mov_b32 s4, 0x2000706
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_lshrrev_b32_e32 v0, 16, v9
; GFX9-NEXT:    v_lshlrev_b16_e32 v0, 8, v0
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v2, v10, v4, s5
; GFX9-NEXT:    v_or_b32_sdwa v0, v9, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_lshl_or_b32 v0, v0, 16, v2
; GFX9-NEXT:    v_perm_b32 v1, v10, v9, s4
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    global_store_dword v[7:8], v1, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 6, i32 7, i32 0, i32 2>
  %vecins = insertelement <4 x i8> %shuffle0_0, i8 %elt, i32 1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @lhsr_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: lhsr_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v9, v[2:3], off
; GFX10-NEXT:    v_mov_b32_e32 v0, 26
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b16 v1, 1, v4
; GFX10-NEXT:    v_lshrrev_b32_e32 v2, 26, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshrrev_b32_sdwa v0, v0, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT:    v_lshrrev_b32_e32 v3, 25, v9
; GFX10-NEXT:    v_and_b32_e32 v1, 0x7f00, v1
; GFX10-NEXT:    v_or_b32_e32 v0, v3, v0
; GFX10-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    v_perm_b32 v1, v9, v4, 0x1030707
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: lhsr_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x1030707
; GFX9-NEXT:    v_mov_b32_e32 v0, 26
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_lshrrev_b16_e32 v2, 1, v4
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v1, v9, v4, s4
; GFX9-NEXT:    v_lshrrev_b32_e32 v3, 26, v4
; GFX9-NEXT:    v_lshrrev_b32_sdwa v0, v0, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT:    v_lshrrev_b32_e32 v4, 25, v9
; GFX9-NEXT:    v_and_b32_e32 v2, 0x7f00, v2
; GFX9-NEXT:    v_or_b32_e32 v0, v4, v0
; GFX9-NEXT:    v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    global_store_dword v[7:8], v1, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 7, i32 7, i32 3, i32 1>
  %vecins = lshr <4 x i8> %shuffle0_0, <i8 1, i8 2, i8 2, i8 1>
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @mul_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: mul_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v9, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b16 v0, 8, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshrrev_b32_e32 v1, 24, v9
; GFX10-NEXT:    v_lshrrev_b16 v2, 8, v9
; GFX10-NEXT:    v_mul_lo_u16 v3, v4, v9
; GFX10-NEXT:    v_mul_lo_u16 v1, v9, v1 op_sel:[1,0,0]
; GFX10-NEXT:    v_mul_lo_u16 v0, v0, v2
; GFX10-NEXT:    v_mul_lo_u16 v2, v9, v9 op_sel:[0,1,0]
; GFX10-NEXT:    v_lshlrev_b16 v1, 8, v1
; GFX10-NEXT:    v_lshlrev_b16 v0, 8, v0
; GFX10-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    v_perm_b32 v1, v4, v9, 0x2000504
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: mul_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x2000504
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v4, v9, s4
; GFX9-NEXT:    v_mul_lo_u16_sdwa v1, v9, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:BYTE_3
; GFX9-NEXT:    v_mul_lo_u16_sdwa v2, v9, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX9-NEXT:    v_mul_lo_u16_e32 v3, v4, v9
; GFX9-NEXT:    v_mul_lo_u16_sdwa v4, v4, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
; GFX9-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v2, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v1, off
; GFX9-NEXT:    global_store_dword v[7:8], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 0, i32 1, i32 4, i32 6>
  %vecins = mul <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @or_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: or_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v2, v[2:3], off
; GFX10-NEXT:    v_bfrev_b32_e32 v4, 4.0
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshlrev_b16 v1, 8, v2 op_sel:[0,1,0]
; GFX10-NEXT:    v_lshlrev_b16 v3, 8, v2
; GFX10-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_or_b32_sdwa v3, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX10-NEXT:    v_perm_b32 v0, v0, v2, 0x2010005
; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_e32 v3, 0x201, v3
; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    global_store_dword v[5:6], v1, off
; GFX10-NEXT:    global_store_dword v[7:8], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: or_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v2, v[2:3], off
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    global_load_dword v0, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x2010005
; GFX9-NEXT:    s_movk_i32 s5, 0x102
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_lshrrev_b32_e32 v1, 16, v2
; GFX9-NEXT:    v_lshlrev_b16_e32 v3, 8, v2
; GFX9-NEXT:    v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v1, s5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v4, v0, v2, s4
; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_e32 v0, 0x201, v0
; GFX9-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    global_store_dword v[7:8], v4, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 1, i32 4, i32 5, i32 6>
  %vecins = or <4 x i8> %shuffle0_0, <i8 1, i8 2, i8 2, i8 1>
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}

define hidden void @sdiv_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: sdiv_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[2:3], off
; GFX10-NEXT:    global_load_dword v9, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v0, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v2, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v3, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v10, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v1, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v11, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3
; GFX10-NEXT:    v_rcp_f32_e32 v12, v0
; GFX10-NEXT:    v_rcp_f32_e32 v2, v2
; GFX10-NEXT:    v_rcp_f32_e32 v3, v3
; GFX10-NEXT:    v_rcp_f32_e32 v10, v10
; GFX10-NEXT:    v_add_nc_u32_e32 v1, 1, v1
; GFX10-NEXT:    v_add_nc_u32_e32 v0, 1, v0
; GFX10-NEXT:    v_add_nc_u32_e32 v11, 1, v11
; GFX10-NEXT:    v_mul_f32_e32 v12, v1, v12
; GFX10-NEXT:    v_mul_f32_e32 v1, v1, v2
; GFX10-NEXT:    v_mul_f32_e32 v0, v0, v3
; GFX10-NEXT:    v_mul_f32_e32 v2, v11, v10
; GFX10-NEXT:    v_trunc_f32_e32 v3, v12
; GFX10-NEXT:    v_trunc_f32_e32 v1, v1
; GFX10-NEXT:    v_trunc_f32_e32 v0, v0
; GFX10-NEXT:    v_trunc_f32_e32 v2, v2
; GFX10-NEXT:    v_cvt_i32_f32_e32 v3, v3
; GFX10-NEXT:    v_cvt_i32_f32_e32 v1, v1
; GFX10-NEXT:    v_cvt_i32_f32_sdwa v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD
; GFX10-NEXT:    v_cvt_i32_f32_sdwa v2, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v1, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    v_perm_b32 v1, v9, v4, 0x60706
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: sdiv_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[2:3], off
; GFX9-NEXT:    global_load_dword v9, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x60706
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v0, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v3, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v10, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v11, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v1, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v2, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3
; GFX9-NEXT:    v_rcp_f32_e32 v12, v0
; GFX9-NEXT:    v_rcp_f32_e32 v3, v3
; GFX9-NEXT:    v_rcp_f32_e32 v10, v10
; GFX9-NEXT:    v_rcp_f32_e32 v11, v11
; GFX9-NEXT:    v_add_u32_e32 v1, 1, v1
; GFX9-NEXT:    v_add_u32_e32 v2, 1, v2
; GFX9-NEXT:    v_add_u32_e32 v0, 1, v0
; GFX9-NEXT:    v_mul_f32_e32 v12, v1, v12
; GFX9-NEXT:    v_mul_f32_e32 v2, v2, v3
; GFX9-NEXT:    v_mul_f32_e32 v1, v1, v10
; GFX9-NEXT:    v_mul_f32_e32 v0, v0, v11
; GFX9-NEXT:    v_trunc_f32_e32 v3, v12
; GFX9-NEXT:    v_trunc_f32_e32 v2, v2
; GFX9-NEXT:    v_trunc_f32_e32 v1, v1
; GFX9-NEXT:    v_trunc_f32_e32 v0, v0
; GFX9-NEXT:    v_cvt_i32_f32_e32 v3, v3
; GFX9-NEXT:    v_cvt_i32_f32_e32 v1, v1
; GFX9-NEXT:    v_cvt_i32_f32_sdwa v0, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD
; GFX9-NEXT:    v_cvt_i32_f32_sdwa v2, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD
; GFX9-NEXT:    v_perm_b32 v4, v9, v4, s4
; GFX9-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    global_store_dword v[7:8], v4, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 2, i32 3, i32 2, i32 4>
  %vecins = sdiv <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @sext_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: sext_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[2:3], off
; GFX10-NEXT:    global_load_dword v9, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_ashrrev_i16 v0, 8, v4 op_sel:[0,1,0]
; GFX10-NEXT:    v_ashrrev_i16 v1, 8, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_ashrrev_i16 v2, 8, v9 op_sel:[0,1,0]
; GFX10-NEXT:    v_perm_b32 v1, v0, v1, 0x5040100
; GFX10-NEXT:    v_perm_b32 v0, v2, v2, 0x5040100
; GFX10-NEXT:    v_perm_b32 v2, v9, v4, 0x3010707
; GFX10-NEXT:    global_store_dwordx2 v[7:8], v[0:1], off
; GFX10-NEXT:    global_store_dword v[5:6], v2, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: sext_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    v_mov_b32_e32 v0, 8
; GFX9-NEXT:    s_mov_b32 s5, 0x5040100
; GFX9-NEXT:    s_mov_b32 s4, 0x3010707
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_ashrrev_i16_e32 v1, 8, v9
; GFX9-NEXT:    v_ashrrev_i16_sdwa v3, v0, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX9-NEXT:    v_ashrrev_i16_sdwa v0, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
; GFX9-NEXT:    v_perm_b32 v1, v3, v1, s5
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s5
; GFX9-NEXT:    v_perm_b32 v2, v4, v9, s4
; GFX9-NEXT:    global_store_dwordx2 v[7:8], v[0:1], off
; GFX9-NEXT:    global_store_dword v[5:6], v2, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 3, i32 3, i32 5, i32 7>
  %insvec = sext <4 x i8> %shuffle0_0 to <4 x i16>
  store <4 x i16> %insvec, ptr addrspace(1) %out1
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0
  ret void
}


define hidden void @shl_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: shl_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[2:3], off
; GFX10-NEXT:    global_load_dword v9, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshlrev_b16 v0, 1, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshlrev_b16 v1, 2, v9
; GFX10-NEXT:    v_and_b32_e32 v2, 0xfffffe00, v0
; GFX10-NEXT:    v_and_b32_e32 v3, 0xfc, v1
; GFX10-NEXT:    v_and_b32_e32 v1, 0xfffffc00, v1
; GFX10-NEXT:    v_and_b32_e32 v0, 0xfe, v0
; GFX10-NEXT:    v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_e32 v0, v0, v1
; GFX10-NEXT:    v_perm_b32 v1, v4, v9, 0x5000104
; GFX10-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shl_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x5000104
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_lshlrev_b16_e32 v2, 2, v4
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_lshlrev_b16_e32 v1, 1, v9
; GFX9-NEXT:    v_perm_b32 v0, v9, v4, s4
; GFX9-NEXT:    v_and_b32_e32 v3, 0xfffffe00, v1
; GFX9-NEXT:    v_and_b32_e32 v4, 0xfc, v2
; GFX9-NEXT:    v_and_b32_e32 v2, 0xfffffc00, v2
; GFX9-NEXT:    v_and_b32_e32 v1, 0xfe, v1
; GFX9-NEXT:    v_or_b32_sdwa v3, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_e32 v1, v1, v2
; GFX9-NEXT:    v_or_b32_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v1, off
; GFX9-NEXT:    global_store_dword v[7:8], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 4, i32 1, i32 0, i32 5>
  %vecins = shl <4 x i8> %shuffle0_0, <i8 1, i8 2, i8 2, i8 1>
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @sitofp_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: sitofp_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[2:3], off
; GFX10-NEXT:    global_load_dword v9, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v3, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v2, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v1, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v0, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1
; GFX10-NEXT:    v_perm_b32 v4, v4, v9, 0x6010205
; GFX10-NEXT:    global_store_dwordx4 v[7:8], v[0:3], off
; GFX10-NEXT:    global_store_dword v[5:6], v4, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: sitofp_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[2:3], off
; GFX9-NEXT:    global_load_dword v9, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x6010205
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v3, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v2, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v1, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v0, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1
; GFX9-NEXT:    v_perm_b32 v4, v4, v9, s4
; GFX9-NEXT:    global_store_dwordx4 v[7:8], v[0:3], off
; GFX9-NEXT:    global_store_dword v[5:6], v4, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 5, i32 2, i32 1, i32 6>
  %insvec = sitofp <4 x i8> %shuffle0_0 to <4 x float>
  store <4 x float> %insvec, ptr addrspace(1) %out1
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0
  ret void
}


define hidden void @srem_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: srem_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[2:3], off
; GFX10-NEXT:    global_load_dword v9, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v0, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v1, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v3, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v10, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v2, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3
; GFX10-NEXT:    v_cvt_f32_i32_sdwa v11, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX10-NEXT:    v_rcp_f32_e32 v0, v0
; GFX10-NEXT:    v_rcp_f32_e32 v12, v1
; GFX10-NEXT:    v_rcp_f32_e32 v13, v3
; GFX10-NEXT:    v_rcp_f32_e32 v10, v10
; GFX10-NEXT:    v_add_nc_u32_e32 v1, 1, v1
; GFX10-NEXT:    v_add_nc_u32_e32 v3, 1, v3
; GFX10-NEXT:    v_add_nc_u32_e32 v11, 1, v11
; GFX10-NEXT:    v_add_nc_u32_e32 v2, 1, v2
; GFX10-NEXT:    v_lshrrev_b32_e32 v14, 16, v4
; GFX10-NEXT:    v_mul_f32_e32 v0, v1, v0
; GFX10-NEXT:    v_mul_f32_e32 v1, v3, v12
; GFX10-NEXT:    v_mul_f32_e32 v3, v11, v13
; GFX10-NEXT:    v_mul_f32_e32 v2, v2, v10
; GFX10-NEXT:    v_lshrrev_b32_e32 v10, 8, v4
; GFX10-NEXT:    v_trunc_f32_e32 v0, v0
; GFX10-NEXT:    v_trunc_f32_e32 v1, v1
; GFX10-NEXT:    v_trunc_f32_e32 v3, v3
; GFX10-NEXT:    v_trunc_f32_e32 v2, v2
; GFX10-NEXT:    v_lshrrev_b32_e32 v11, 24, v4
; GFX10-NEXT:    v_cvt_i32_f32_e32 v0, v0
; GFX10-NEXT:    v_cvt_i32_f32_e32 v1, v1
; GFX10-NEXT:    v_cvt_i32_f32_e32 v3, v3
; GFX10-NEXT:    v_cvt_i32_f32_e32 v2, v2
; GFX10-NEXT:    v_mul_lo_u32 v0, v0, v4
; GFX10-NEXT:    v_mul_lo_u32 v1, v1, v14
; GFX10-NEXT:    v_mul_lo_u32 v3, v3, v11
; GFX10-NEXT:    v_mul_lo_u32 v2, v2, v10
; GFX10-NEXT:    v_sub_nc_u32_e32 v0, v14, v0
; GFX10-NEXT:    v_sub_nc_u32_e32 v1, v11, v1
; GFX10-NEXT:    v_sub_nc_u32_sdwa v3, v9, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT:    v_sub_nc_u32_sdwa v2, v9, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    v_perm_b32 v1, v4, v9, 0x2070306
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: srem_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[2:3], off
; GFX9-NEXT:    global_load_dword v9, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x2070306
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v2, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v3, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v12, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v13, sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v11, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3
; GFX9-NEXT:    v_cvt_f32_i32_sdwa v15, sext(v9) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2
; GFX9-NEXT:    v_rcp_f32_e32 v2, v2
; GFX9-NEXT:    v_add_u32_e32 v16, 1, v3
; GFX9-NEXT:    v_rcp_f32_e32 v12, v12
; GFX9-NEXT:    v_rcp_f32_e32 v3, v3
; GFX9-NEXT:    v_add_u32_e32 v17, 1, v13
; GFX9-NEXT:    v_rcp_f32_e32 v13, v13
; GFX9-NEXT:    v_add_u32_e32 v11, 1, v11
; GFX9-NEXT:    v_add_u32_e32 v15, 1, v15
; GFX9-NEXT:    v_mul_f32_e32 v2, v16, v2
; GFX9-NEXT:    v_mul_f32_e32 v11, v11, v12
; GFX9-NEXT:    v_mul_f32_e32 v3, v17, v3
; GFX9-NEXT:    v_mul_f32_e32 v12, v15, v13
; GFX9-NEXT:    v_trunc_f32_e32 v2, v2
; GFX9-NEXT:    v_trunc_f32_e32 v11, v11
; GFX9-NEXT:    v_trunc_f32_e32 v3, v3
; GFX9-NEXT:    v_trunc_f32_e32 v12, v12
; GFX9-NEXT:    v_cvt_i32_f32_e32 v2, v2
; GFX9-NEXT:    v_cvt_i32_f32_e32 v11, v11
; GFX9-NEXT:    v_cvt_i32_f32_e32 v3, v3
; GFX9-NEXT:    v_cvt_i32_f32_e32 v12, v12
; GFX9-NEXT:    v_lshrrev_b32_e32 v0, 16, v4
; GFX9-NEXT:    v_lshrrev_b32_e32 v10, 8, v4
; GFX9-NEXT:    v_lshrrev_b32_e32 v14, 24, v4
; GFX9-NEXT:    v_perm_b32 v1, v4, v9, s4
; GFX9-NEXT:    v_mul_lo_u32 v2, v2, v4
; GFX9-NEXT:    v_mul_lo_u32 v4, v11, v10
; GFX9-NEXT:    v_mul_lo_u32 v3, v3, v0
; GFX9-NEXT:    v_mul_lo_u32 v10, v12, v14
; GFX9-NEXT:    v_sub_u32_e32 v0, v0, v2
; GFX9-NEXT:    v_sub_u32_sdwa v2, v9, v4 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:DWORD
; GFX9-NEXT:    v_sub_u32_e32 v3, v14, v3
; GFX9-NEXT:    v_sub_u32_sdwa v4, v9, v10 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v3, v3, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    global_store_dword v[7:8], v1, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 6, i32 3, i32 7, i32 2>
  %vecins = srem <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @sub_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: sub_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v2, v[2:3], off
; GFX10-NEXT:    global_load_dword v0, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b32_e32 v1, 24, v2
; GFX10-NEXT:    v_lshrrev_b16 v3, 8, v2
; GFX10-NEXT:    v_sub_nc_u16 v4, v2, v1 op_sel:[1,0,0]
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_sub_nc_u16 v3, v0, v3
; GFX10-NEXT:    v_sub_nc_u16 v9, v1, v2 op_sel:[0,1,0]
; GFX10-NEXT:    v_sub_nc_u16 v1, v1, v2
; GFX10-NEXT:    v_perm_b32 v0, v2, v0, 0x6070007
; GFX10-NEXT:    v_lshlrev_b16 v4, 8, v4
; GFX10-NEXT:    v_lshlrev_b16 v3, 8, v3
; GFX10-NEXT:    v_or_b32_sdwa v4, v9, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    global_store_dword v[5:6], v1, off
; GFX10-NEXT:    global_store_dword v[7:8], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: sub_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x6070007
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v9, v4, s4
; GFX9-NEXT:    v_sub_u16_sdwa v1, v9, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:BYTE_3
; GFX9-NEXT:    v_sub_u16_sdwa v2, v9, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:WORD_1
; GFX9-NEXT:    v_sub_u16_sdwa v3, v4, v9 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1
; GFX9-NEXT:    v_sub_u16_sdwa v4, v9, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v2, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v1, off
; GFX9-NEXT:    global_store_dword v[7:8], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 7, i32 0, i32 7, i32 6>
  %vecins = sub <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @sv_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1, ptr addrspace(1) %out2) {
; GFX10-LABEL: sv_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v5, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v4, v5, 0x50705
; GFX10-NEXT:    global_store_dword v[7:8], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: sv_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v5, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x50705
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v4, v5, s4
; GFX9-NEXT:    global_store_dword v[7:8], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 1, i32 3, i32 1, i32 4>
  %insvec = shufflevector <4 x i8> %shuffle0_0, <4 x i8> %vec1, <4 x i32> <i32 6, i32 3, i32 7, i32 0>
  store <4 x i8> %insvec, ptr addrspace(1) %out1
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @trunc_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: trunc_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v9, v[2:3], off
; GFX10-NEXT:    v_mov_b32_e32 v0, 1
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_and_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_bfe_u32 v1, v9, 8, 1
; GFX10-NEXT:    v_lshlrev_b16 v0, 1, v0
; GFX10-NEXT:    v_lshlrev_b16 v2, 2, v1
; GFX10-NEXT:    v_or_b32_e32 v0, v1, v0
; GFX10-NEXT:    v_lshlrev_b16 v1, 3, v4
; GFX10-NEXT:    v_or_b32_e32 v0, v0, v2
; GFX10-NEXT:    v_or_b32_e32 v0, v0, v1
; GFX10-NEXT:    v_perm_b32 v1, v9, v4, 0x50205
; GFX10-NEXT:    v_and_b32_e32 v0, 15, v0
; GFX10-NEXT:    global_store_byte v[7:8], v0, off
; GFX10-NEXT:    global_store_dword v[5:6], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: trunc_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    v_mov_b32_e32 v0, 1
; GFX9-NEXT:    s_mov_b32 s4, 0x50205
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_and_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_bfe_u32 v2, v9, 8, 1
; GFX9-NEXT:    v_lshlrev_b16_e32 v0, 1, v0
; GFX9-NEXT:    v_perm_b32 v1, v9, v4, s4
; GFX9-NEXT:    v_lshlrev_b16_e32 v3, 3, v4
; GFX9-NEXT:    v_lshlrev_b16_e32 v4, 2, v2
; GFX9-NEXT:    v_or_b32_e32 v0, v2, v0
; GFX9-NEXT:    v_or_b32_e32 v0, v0, v4
; GFX9-NEXT:    v_or_b32_e32 v0, v0, v3
; GFX9-NEXT:    v_and_b32_e32 v0, 15, v0
; GFX9-NEXT:    global_store_byte v[7:8], v0, off
; GFX9-NEXT:    global_store_dword v[5:6], v1, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 5, i32 2, i32 5, i32 0>
  %insvec = trunc <4 x i8> %shuffle0_0 to <4 x i1>
  store <4 x i1> %insvec, ptr addrspace(1) %out1
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0
  ret void
}

define hidden void @udiv(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: udiv:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v9, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_cvt_f32_ubyte3_e32 v0, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_cvt_f32_ubyte0_e32 v1, v9
; GFX10-NEXT:    v_cvt_f32_ubyte2_e32 v2, v9
; GFX10-NEXT:    v_cvt_f32_ubyte3_e32 v3, v9
; GFX10-NEXT:    v_cvt_f32_ubyte1_e32 v10, v9
; GFX10-NEXT:    v_cvt_f32_ubyte0_e32 v11, v4
; GFX10-NEXT:    v_rcp_f32_e32 v12, v1
; GFX10-NEXT:    v_rcp_f32_e32 v13, v2
; GFX10-NEXT:    v_rcp_f32_e32 v3, v3
; GFX10-NEXT:    v_rcp_f32_e32 v10, v10
; GFX10-NEXT:    v_add_nc_u32_e32 v0, 1, v0
; GFX10-NEXT:    v_add_nc_u32_e32 v11, 1, v11
; GFX10-NEXT:    v_add_nc_u32_e32 v1, 1, v1
; GFX10-NEXT:    v_add_nc_u32_e32 v2, 1, v2
; GFX10-NEXT:    v_mul_f32_e32 v0, v0, v12
; GFX10-NEXT:    v_mul_f32_e32 v11, v11, v13
; GFX10-NEXT:    v_mul_f32_e32 v1, v1, v3
; GFX10-NEXT:    v_mul_f32_e32 v2, v2, v10
; GFX10-NEXT:    v_trunc_f32_e32 v0, v0
; GFX10-NEXT:    v_trunc_f32_e32 v3, v11
; GFX10-NEXT:    v_trunc_f32_e32 v1, v1
; GFX10-NEXT:    v_trunc_f32_e32 v2, v2
; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
; GFX10-NEXT:    v_cvt_u32_f32_e32 v3, v3
; GFX10-NEXT:    v_cvt_u32_f32_sdwa v1, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD
; GFX10-NEXT:    v_cvt_u32_f32_sdwa v2, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v1, v3, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    v_perm_b32 v1, v4, v9, 0x40207
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: udiv:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x40207
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_cvt_f32_ubyte3_e32 v0, v4
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v1, v9
; GFX9-NEXT:    v_cvt_f32_ubyte1_e32 v2, v9
; GFX9-NEXT:    v_cvt_f32_ubyte2_e32 v3, v9
; GFX9-NEXT:    v_cvt_f32_ubyte3_e32 v11, v9
; GFX9-NEXT:    v_rcp_f32_e32 v12, v1
; GFX9-NEXT:    v_rcp_f32_e32 v2, v2
; GFX9-NEXT:    v_add_u32_e32 v13, 1, v3
; GFX9-NEXT:    v_rcp_f32_e32 v3, v3
; GFX9-NEXT:    v_rcp_f32_e32 v11, v11
; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v10, v4
; GFX9-NEXT:    v_add_u32_e32 v0, 1, v0
; GFX9-NEXT:    v_add_u32_e32 v10, 1, v10
; GFX9-NEXT:    v_add_u32_e32 v1, 1, v1
; GFX9-NEXT:    v_mul_f32_e32 v0, v0, v12
; GFX9-NEXT:    v_mul_f32_e32 v2, v13, v2
; GFX9-NEXT:    v_mul_f32_e32 v3, v10, v3
; GFX9-NEXT:    v_mul_f32_e32 v1, v1, v11
; GFX9-NEXT:    v_trunc_f32_e32 v0, v0
; GFX9-NEXT:    v_trunc_f32_e32 v2, v2
; GFX9-NEXT:    v_trunc_f32_e32 v3, v3
; GFX9-NEXT:    v_trunc_f32_e32 v1, v1
; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
; GFX9-NEXT:    v_cvt_u32_f32_e32 v3, v3
; GFX9-NEXT:    v_cvt_u32_f32_sdwa v1, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD
; GFX9-NEXT:    v_cvt_u32_f32_sdwa v2, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD
; GFX9-NEXT:    v_perm_b32 v4, v4, v9, s4
; GFX9-NEXT:    v_or_b32_sdwa v1, v3, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    global_store_dword v[7:8], v4, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 3, i32 6, i32 0, i32 4>
  %vecins = udiv <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @uitofp_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: uitofp_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[2:3], off
; GFX10-NEXT:    global_load_dword v9, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_cvt_f32_ubyte1_e32 v3, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_cvt_f32_ubyte2_e32 v2, v9
; GFX10-NEXT:    v_cvt_f32_ubyte1_e32 v1, v9
; GFX10-NEXT:    v_cvt_f32_ubyte0_e32 v0, v4
; GFX10-NEXT:    v_perm_b32 v4, v4, v9, 0x5020104
; GFX10-NEXT:    global_store_dwordx4 v[7:8], v[0:3], off
; GFX10-NEXT:    global_store_dword v[5:6], v4, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: uitofp_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x5020104
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_cvt_f32_ubyte2_e32 v2, v4
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_cvt_f32_ubyte1_e32 v3, v9
; GFX9-NEXT:    v_cvt_f32_ubyte1_e32 v1, v4
; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, v9
; GFX9-NEXT:    v_perm_b32 v10, v9, v4, s4
; GFX9-NEXT:    global_store_dwordx4 v[7:8], v[0:3], off
; GFX9-NEXT:    global_store_dword v[5:6], v10, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 4, i32 1, i32 2, i32 5>
  %insvec = uitofp <4 x i8> %shuffle0_0 to <4 x float>
  store <4 x float> %insvec, ptr addrspace(1) %out1
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0
  ret void
}


define hidden void @urem_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: urem_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[2:3], off
; GFX10-NEXT:    global_load_dword v9, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_cvt_f32_ubyte0_e32 v0, v4
; GFX10-NEXT:    v_cvt_f32_ubyte2_e32 v1, v4
; GFX10-NEXT:    v_cvt_f32_ubyte3_e32 v2, v4
; GFX10-NEXT:    v_cvt_f32_ubyte1_e32 v3, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_cvt_f32_ubyte2_e32 v10, v9
; GFX10-NEXT:    v_rcp_f32_e32 v0, v0
; GFX10-NEXT:    v_rcp_f32_e32 v1, v1
; GFX10-NEXT:    v_rcp_f32_e32 v2, v2
; GFX10-NEXT:    v_rcp_f32_e32 v11, v3
; GFX10-NEXT:    v_add_nc_u32_e32 v3, 1, v3
; GFX10-NEXT:    v_add_nc_u32_e32 v10, 1, v10
; GFX10-NEXT:    v_lshrrev_b32_e32 v12, 16, v4
; GFX10-NEXT:    v_mul_f32_e32 v0, v3, v0
; GFX10-NEXT:    v_mul_f32_e32 v1, v3, v1
; GFX10-NEXT:    v_mul_f32_e32 v2, v10, v2
; GFX10-NEXT:    v_mul_f32_e32 v3, v3, v11
; GFX10-NEXT:    v_lshrrev_b32_e32 v10, 8, v4
; GFX10-NEXT:    v_trunc_f32_e32 v0, v0
; GFX10-NEXT:    v_trunc_f32_e32 v1, v1
; GFX10-NEXT:    v_trunc_f32_e32 v2, v2
; GFX10-NEXT:    v_trunc_f32_e32 v3, v3
; GFX10-NEXT:    v_lshrrev_b32_e32 v11, 24, v4
; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
; GFX10-NEXT:    v_cvt_u32_f32_e32 v1, v1
; GFX10-NEXT:    v_cvt_u32_f32_e32 v2, v2
; GFX10-NEXT:    v_cvt_u32_f32_e32 v3, v3
; GFX10-NEXT:    v_mul_lo_u32 v0, v0, v4
; GFX10-NEXT:    v_mul_lo_u32 v1, v1, v12
; GFX10-NEXT:    v_mul_lo_u32 v2, v2, v11
; GFX10-NEXT:    v_mul_lo_u32 v3, v3, v10
; GFX10-NEXT:    v_sub_nc_u32_e32 v0, v10, v0
; GFX10-NEXT:    v_sub_nc_u32_e32 v1, v10, v1
; GFX10-NEXT:    v_sub_nc_u32_sdwa v2, v9, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT:    v_sub_nc_u32_sdwa v3, v10, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    v_perm_b32 v1, v4, v9, 0x2050505
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: urem_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[2:3], off
; GFX9-NEXT:    global_load_dword v9, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x2050505
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v2, v4
; GFX9-NEXT:    v_cvt_f32_ubyte1_e32 v3, v4
; GFX9-NEXT:    v_cvt_f32_ubyte2_e32 v11, v4
; GFX9-NEXT:    v_cvt_f32_ubyte3_e32 v14, v4
; GFX9-NEXT:    v_rcp_f32_e32 v2, v2
; GFX9-NEXT:    v_add_u32_e32 v15, 1, v3
; GFX9-NEXT:    v_rcp_f32_e32 v3, v3
; GFX9-NEXT:    v_rcp_f32_e32 v11, v11
; GFX9-NEXT:    v_rcp_f32_e32 v14, v14
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_cvt_f32_ubyte2_e32 v13, v9
; GFX9-NEXT:    v_add_u32_e32 v13, 1, v13
; GFX9-NEXT:    v_mul_f32_e32 v2, v15, v2
; GFX9-NEXT:    v_mul_f32_e32 v3, v15, v3
; GFX9-NEXT:    v_mul_f32_e32 v11, v15, v11
; GFX9-NEXT:    v_mul_f32_e32 v13, v13, v14
; GFX9-NEXT:    v_trunc_f32_e32 v2, v2
; GFX9-NEXT:    v_trunc_f32_e32 v3, v3
; GFX9-NEXT:    v_trunc_f32_e32 v11, v11
; GFX9-NEXT:    v_trunc_f32_e32 v13, v13
; GFX9-NEXT:    v_cvt_u32_f32_e32 v2, v2
; GFX9-NEXT:    v_cvt_u32_f32_e32 v3, v3
; GFX9-NEXT:    v_cvt_u32_f32_e32 v11, v11
; GFX9-NEXT:    v_cvt_u32_f32_e32 v13, v13
; GFX9-NEXT:    v_lshrrev_b32_e32 v0, 16, v4
; GFX9-NEXT:    v_lshrrev_b32_e32 v10, 8, v4
; GFX9-NEXT:    v_lshrrev_b32_e32 v12, 24, v4
; GFX9-NEXT:    v_perm_b32 v1, v4, v9, s4
; GFX9-NEXT:    v_mul_lo_u32 v2, v2, v4
; GFX9-NEXT:    v_mul_lo_u32 v3, v3, v10
; GFX9-NEXT:    v_mul_lo_u32 v0, v11, v0
; GFX9-NEXT:    v_mul_lo_u32 v4, v13, v12
; GFX9-NEXT:    v_sub_u32_e32 v2, v10, v2
; GFX9-NEXT:    v_sub_u32_sdwa v3, v10, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT:    v_sub_u32_e32 v0, v10, v0
; GFX9-NEXT:    v_sub_u32_sdwa v4, v9, v4 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    global_store_dword v[7:8], v1, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 5, i32 5, i32 5, i32 2>
  %vecins = urem <4 x i8> %shuffle0_0, %vec1
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @xor_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: xor_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[2:3], off
; GFX10-NEXT:    global_load_dword v9, v[0:1], off
; GFX10-NEXT:    v_mov_b32_e32 v0, 0xffffff00
; GFX10-NEXT:    v_mov_b32_e32 v1, 2
; GFX10-NEXT:    v_mov_b32_e32 v2, 1
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_and_b32_e32 v3, 0xffffff00, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_and_b32_sdwa v0, v9, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT:    v_xor_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT:    v_xor_b32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:DWORD
; GFX10-NEXT:    v_xor_b32_e32 v3, 0x100, v3
; GFX10-NEXT:    v_xor_b32_e32 v0, 0x200, v0
; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX10-NEXT:    v_or_b32_e32 v0, v2, v0
; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX10-NEXT:    v_perm_b32 v1, v4, v9, 0x5060307
; GFX10-NEXT:    global_store_dword v[5:6], v0, off
; GFX10-NEXT:    global_store_dword v[7:8], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: xor_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_movk_i32 s4, 0xff00
; GFX9-NEXT:    v_mov_b32_e32 v0, 2
; GFX9-NEXT:    v_mov_b32_e32 v1, 1
; GFX9-NEXT:    s_mov_b32 s5, 0x5060307
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_and_b32_sdwa v2, v4, s4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v3, 0xffffff00, v9
; GFX9-NEXT:    v_xor_b32_sdwa v0, v9, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT:    v_xor_b32_sdwa v1, v9, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:DWORD
; GFX9-NEXT:    v_xor_b32_e32 v3, 0x100, v3
; GFX9-NEXT:    v_xor_b32_e32 v2, 0x200, v2
; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT:    v_or_b32_e32 v1, v1, v2
; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; GFX9-NEXT:    v_perm_b32 v4, v9, v4, s5
; GFX9-NEXT:    global_store_dword v[5:6], v0, off
; GFX9-NEXT:    global_store_dword v[7:8], v4, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 7, i32 3, i32 6, i32 5>
  %vecins = xor <4 x i8> %shuffle0_0, <i8 1, i8 2, i8 2, i8 1>
  store <4 x i8> %vecins, ptr addrspace(1) %out0
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out1
  ret void
}


define hidden void @zext_store_div(ptr addrspace(1) %in0, ptr addrspace(1) %in1, i8 %elt, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: zext_store_div:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX10-NEXT:    v_add_co_u32 v0, vcc_lo, v0, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
; GFX10-NEXT:    v_add_co_u32 v2, vcc_lo, v2, v4
; GFX10-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, 0, v3, vcc_lo
; GFX10-NEXT:    global_load_dword v4, v[0:1], off
; GFX10-NEXT:    global_load_dword v9, v[2:3], off
; GFX10-NEXT:    v_mov_b32_e32 v0, 0xff
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshrrev_b16 v1, 8, v4
; GFX10-NEXT:    v_and_b32_e32 v2, 0xff, v4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v3, 0xff, v9
; GFX10-NEXT:    v_and_b32_sdwa v10, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX10-NEXT:    v_perm_b32 v0, v1, v2, 0x5040100
; GFX10-NEXT:    v_perm_b32 v2, v4, v9, 0x60504
; GFX10-NEXT:    v_perm_b32 v1, v3, v10, 0x5040100
; GFX10-NEXT:    global_store_dwordx2 v[7:8], v[0:1], off
; GFX10-NEXT:    global_store_dword v[5:6], v2, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: zext_store_div:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v4, 0x3ff, v31
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 2, v4
; GFX9-NEXT:    v_add_co_u32_e32 v0, vcc, v0, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
; GFX9-NEXT:    v_add_co_u32_e32 v2, vcc, v2, v4
; GFX9-NEXT:    v_addc_co_u32_e32 v3, vcc, 0, v3, vcc
; GFX9-NEXT:    global_load_dword v4, v[0:1], off
; GFX9-NEXT:    global_load_dword v9, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x60504
; GFX9-NEXT:    s_movk_i32 s5, 0xff
; GFX9-NEXT:    s_mov_b32 s6, 0x5040100
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_lshrrev_b16_e32 v0, 8, v4
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v2, v4, v9, s4
; GFX9-NEXT:    v_and_b32_e32 v1, 0xff, v4
; GFX9-NEXT:    v_and_b32_e32 v3, 0xff, v9
; GFX9-NEXT:    v_and_b32_sdwa v4, v4, s5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; GFX9-NEXT:    v_perm_b32 v0, v0, v1, s6
; GFX9-NEXT:    v_perm_b32 v1, v3, v4, s6
; GFX9-NEXT:    global_store_dwordx2 v[7:8], v[0:1], off
; GFX9-NEXT:    global_store_dword v[5:6], v2, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %tid = call i32 @llvm.amdgcn.workitem.id.x()
  %gep0 = getelementptr <4 x i8>, ptr addrspace(1) %in0, i32 %tid
  %gep1 = getelementptr <4 x i8>, ptr addrspace(1) %in1, i32 %tid
  %vec0 = load <4 x i8>, ptr addrspace(1) %gep0, align 4
  %vec1 = load <4 x i8>, ptr addrspace(1) %gep1, align 4
  %shuffle0_0 = shufflevector <4 x i8> %vec0, <4 x i8> %vec1, <4 x i32> <i32 0, i32 1, i32 2, i32 4>
  %insvec = zext <4 x i8> %shuffle0_0 to <4 x i16>
  store <4 x i16> %insvec, ptr addrspace(1) %out1
  store <4 x i8> %shuffle0_0, ptr addrspace(1) %out0
  ret void
}

define void @Source16Bit(i16 %in, <2 x i16> %reg) {
; GFX10-LABEL: Source16Bit:
; GFX10:       ; %bb.0: ; %entry
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v1, 0x3050204
; GFX10-NEXT:    global_store_dword v[0:1], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: Source16Bit:
; GFX9:       ; %bb.0: ; %entry
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    s_mov_b32 s4, 0x3050204
; GFX9-NEXT:    v_perm_b32 v0, v0, v1, s4
; GFX9-NEXT:    global_store_dword v[0:1], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
entry:
  %elt0 = extractelement <2 x i16> %reg, i32 1
  %e0b0 = and i16 %elt0, 255
  %e0b1 = and i16 %elt0, -256
  %e1b0 = and i16 %in, 255
  %e1b1 = and i16 %in, -256
  %tmp0 = shl i16 %e0b0, 8
  %byte0 = or i16 %tmp0, %e1b0
  %tmp2 = lshr i16 %e1b1, 8
  %byte1 = or i16 %e0b1, %tmp2
  %ext0 = zext i16 %byte0 to i32
  %ext1 = zext i16 %byte1 to i32
  %shifted = shl i32 %ext1, 16
  %result = or i32 %shifted, %ext0
  store i32 %result, ptr addrspace(1) poison
  ret void
}

define hidden void @extract3744(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: extract3744:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v6, v7, 0x3070404
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract3744:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x3070404
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v6, v7, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec2 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <4 x i8> %vec1, i64 0
  %zv1e0 = zext i8 %v1e0 to i32
  %byte1 = shl i32 %zv1e0, 8

  %v1e3 = extractelement <4 x i8> %vec1, i64 3
  %zv1e3 = zext i8 %v1e3 to i32
  %byte2 = shl i32 %zv1e3, 16
  %v2e3 = extractelement <4 x i8> %vec2, i64 3
  %zv2e3 = zext i8 %v2e3 to i32
  %byte3 = shl i32 %zv2e3, 24

  %tmp0 = or i32 %zv1e0, %byte1
  %tmp1 = or i32 %tmp0, %byte2
  %res = or i32 %tmp1, %byte3
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

declare i32 @llvm.amdgcn.perm(i32, i32, i32)

define hidden void @extract_perm_3744(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: extract_perm_3744:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v6, v7, 0x3070404
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_perm_3744:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x3070404
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v6, v7, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec2 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %cast1 = bitcast <4 x i8> %vec1 to i32
  %cast2 = bitcast <4 x i8> %vec2 to i32
  %lo24 = call i32 @llvm.amdgcn.perm(i32 %cast1, i32 %cast1, i32 201523200)
  %hi8 = call i32 @llvm.amdgcn.perm(i32 %cast2, i32 %cast2, i32 51121164)
  %res = or i32 %hi8, %lo24
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @extract1347_v2i16(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: extract1347_v2i16:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x1030407
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract1347_v2i16:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x1030407
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %b0t0 = and i16 -256, %v2e1
  %b0t1 = lshr i16 %b0t0, 8
  %byte0 = zext i16 %b0t1 to i32

  %b1t0 = and i16 255, %v2e0
  %b1t1 = zext i16 %b1t0 to i32
  %byte1 = shl i32 %b1t1, 8

  %b2t0 = and i16 -256, %v1e1
  %b2t1 = lshr i16 %b2t0, 8
  %b2t2 = zext i16 %b2t1 to i32
  %byte2 = shl i32 %b2t2, 16

  %b3t0 = and i16 -256, %v1e0
  %b3t1 = lshr i16 %b3t0, 8
  %b3t2 = zext i16 %b3t1 to i32
  %byte3 = shl i32 %b3t2, 24

  %tmp0 = or i32 %byte0, %byte1
  %tmp1 = or i32 %tmp0, %byte2
  %res = or i32 %tmp1, %byte3
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}


declare i16 @llvm.fshr.i16(i16, i16, i16)

define hidden void @fshri16_8(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: fshri16_8:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x30407
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: fshri16_8:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x30407
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %tmp01.0 = call i16 @llvm.fshr.i16(i16 %v2e0, i16 %v2e1, i16 8)
  %byte01 = zext i16 %tmp01.0 to i32

  %tmp23.0 = call i16 @llvm.fshr.i16(i16 %v1e0, i16 %v1e1, i16 8)
  %tmp23.1 = zext i16 %tmp23.0 to i32
  %byte23 = shl i32 %tmp23.1, 16
  %res = or i32 %byte01, %byte23
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @fshri16_16(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: fshri16_16:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x3020706
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: fshri16_16:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x3020706
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %tmp01.0 = call i16 @llvm.fshr.i16(i16 %v2e0, i16 %v2e1, i16 16)
  %byte01 = zext i16 %tmp01.0 to i32

  %tmp23.0 = call i16 @llvm.fshr.i16(i16 %v1e0, i16 %v1e1, i16 16)
  %tmp23.1 = zext i16 %tmp23.0 to i32
  %byte23 = shl i32 %tmp23.1, 16
  %res = or i32 %byte01, %byte23
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @fshri16_24(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: fshri16_24:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x30407
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: fshri16_24:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x30407
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %tmp01.0 = call i16 @llvm.fshr.i16(i16 %v2e0, i16 %v2e1, i16 24)
  %byte01 = zext i16 %tmp01.0 to i32

  %tmp23.0 = call i16 @llvm.fshr.i16(i16 %v1e0, i16 %v1e1, i16 24)
  %tmp23.1 = zext i16 %tmp23.0 to i32
  %byte23 = shl i32 %tmp23.1, 16
  %res = or i32 %byte01, %byte23
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @fshri16_32(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: fshri16_32:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x3020706
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: fshri16_32:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x3020706
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %tmp01.0 = call i16 @llvm.fshr.i16(i16 %v2e0, i16 %v2e1, i16 32)
  %byte01 = zext i16 %tmp01.0 to i32

  %tmp23.0 = call i16 @llvm.fshr.i16(i16 %v1e0, i16 %v1e1, i16 32)
  %tmp23.1 = zext i16 %tmp23.0 to i32
  %byte23 = shl i32 %tmp23.1, 16
  %res = or i32 %byte01, %byte23
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @fshri16_88(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: fshri16_88:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x30407
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: fshri16_88:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x30407
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %tmp01.0 = call i16 @llvm.fshr.i16(i16 %v2e0, i16 %v2e1, i16 88)
  %byte01 = zext i16 %tmp01.0 to i32

  %tmp23.0 = call i16 @llvm.fshr.i16(i16 %v1e0, i16 %v1e1, i16 88)
  %tmp23.1 = zext i16 %tmp23.0 to i32
  %byte23 = shl i32 %tmp23.1, 16
  %res = or i32 %byte01, %byte23
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

declare i16 @llvm.fshl.i16(i16, i16, i16)

define hidden void @fshli16_1347(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: fshli16_1347:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x30407
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: fshli16_1347:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x30407
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %tmp01.0 = call i16 @llvm.fshl.i16(i16 %v2e0, i16 %v2e1, i16 8)
  %byte01 = zext i16 %tmp01.0 to i32

  %tmp23.0 = call i16 @llvm.fshl.i16(i16 %v1e0, i16 %v1e1, i16 8)
  %tmp23.1 = zext i16 %tmp23.0 to i32
  %byte23 = shl i32 %tmp23.1, 16
  %res = or i32 %byte01, %byte23
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @fshli16_16(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: fshli16_16:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x1000504
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: fshli16_16:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x1000504
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %tmp01.0 = call i16 @llvm.fshl.i16(i16 %v2e0, i16 %v2e1, i16 16)
  %byte01 = zext i16 %tmp01.0 to i32

  %tmp23.0 = call i16 @llvm.fshl.i16(i16 %v1e0, i16 %v1e1, i16 16)
  %tmp23.1 = zext i16 %tmp23.0 to i32
  %byte23 = shl i32 %tmp23.1, 16
  %res = or i32 %byte01, %byte23
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @fshli16_24(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: fshli16_24:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x30407
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: fshli16_24:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x30407
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %tmp01.0 = call i16 @llvm.fshl.i16(i16 %v2e0, i16 %v2e1, i16 24)
  %byte01 = zext i16 %tmp01.0 to i32

  %tmp23.0 = call i16 @llvm.fshl.i16(i16 %v1e0, i16 %v1e1, i16 24)
  %tmp23.1 = zext i16 %tmp23.0 to i32
  %byte23 = shl i32 %tmp23.1, 16
  %res = or i32 %byte01, %byte23
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @fshli16_32(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: fshli16_32:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x1000504
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: fshli16_32:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x1000504
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %tmp01.0 = call i16 @llvm.fshl.i16(i16 %v2e0, i16 %v2e1, i16 32)
  %byte01 = zext i16 %tmp01.0 to i32

  %tmp23.0 = call i16 @llvm.fshl.i16(i16 %v1e0, i16 %v1e1, i16 32)
  %tmp23.1 = zext i16 %tmp23.0 to i32
  %byte23 = shl i32 %tmp23.1, 16
  %res = or i32 %byte01, %byte23
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @fshli16_88(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: fshli16_88:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[0:1], off
; GFX10-NEXT:    global_load_dword v7, v[2:3], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x30407
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: fshli16_88:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[0:1], off
; GFX9-NEXT:    global_load_dword v7, v[2:3], off
; GFX9-NEXT:    s_mov_b32 s4, 0x30407
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <2 x i16>, ptr addrspace(1) %in0, align 4
  %vec2 = load <2 x i16>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <2 x i16> %vec1, i64 0
  %v1e1 = extractelement <2 x i16> %vec1, i64 1
  %v2e0 = extractelement <2 x i16> %vec2, i64 0
  %v2e1 = extractelement <2 x i16> %vec2, i64 1

  %tmp01.0 = call i16 @llvm.fshl.i16(i16 %v2e0, i16 %v2e1, i16 88)
  %byte01 = zext i16 %tmp01.0 to i32

  %tmp23.0 = call i16 @llvm.fshl.i16(i16 %v1e0, i16 %v1e1, i16 88)
  %tmp23.1 = zext i16 %tmp23.0 to i32
  %byte23 = shl i32 %tmp23.1, 16
  %res = or i32 %byte01, %byte23
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @shlbase(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0, i32 %base) {
; GFX10-LABEL: shlbase:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v7, v[0:1], off
; GFX10-NEXT:    global_load_dword v8, v[2:3], off
; GFX10-NEXT:    v_add_nc_u32_e32 v0, 16, v6
; GFX10-NEXT:    v_add_nc_u32_e32 v1, 24, v6
; GFX10-NEXT:    v_add_nc_u32_e32 v3, 8, v6
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_and_b32_e32 v2, 0xff, v7
; GFX10-NEXT:    v_lshlrev_b32_sdwa v0, v0, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshlrev_b32_sdwa v1, v1, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
; GFX10-NEXT:    v_lshl_or_b32 v2, v2, v3, v2
; GFX10-NEXT:    v_or3_b32 v0, v2, v0, v1
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: shlbase:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v7, v[0:1], off
; GFX9-NEXT:    global_load_dword v8, v[2:3], off
; GFX9-NEXT:    v_add_u32_e32 v0, 8, v6
; GFX9-NEXT:    v_add_u32_e32 v1, 16, v6
; GFX9-NEXT:    v_add_u32_e32 v2, 24, v6
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_and_b32_e32 v3, 0xff, v7
; GFX9-NEXT:    v_lshlrev_b32_sdwa v1, v1, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_lshlrev_b32_sdwa v2, v2, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
; GFX9-NEXT:    v_lshl_or_b32 v0, v3, v0, v3
; GFX9-NEXT:    v_or3_b32 v0, v0, v1, v2
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec2 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <4 x i8> %vec1, i64 0
  %zv1e0 = zext i8 %v1e0 to i32
  %b8 = add i32 %base, 8
  %byte1 = shl i32 %zv1e0, %b8

  %v1e3 = extractelement <4 x i8> %vec1, i64 3
  %zv1e3 = zext i8 %v1e3 to i32
  %b16 = add i32 %base, 16
  %byte2 = shl i32 %zv1e3, %b16
  %v2e3 = extractelement <4 x i8> %vec2, i64 3
  %zv2e3 = zext i8 %v2e3 to i32
  %b24 = add i32 %base, 24
  %byte3 = shl i32 %zv2e3, %b24

  %tmp0 = or i32 %zv1e0, %byte1
  %tmp1 = or i32 %tmp0, %byte2
  %res = or i32 %tmp1, %byte3
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

; TODO -- lower into v_perm
define hidden void @extractbase(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0, i64 %base) {
; GFX10-LABEL: extractbase:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v7, v[0:1], off
; GFX10-NEXT:    global_load_dword v8, v[2:3], off
; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 3, v6
; GFX10-NEXT:    v_add_nc_u32_e32 v1, 24, v0
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_bfe_u32 v2, v7, v1, 8
; GFX10-NEXT:    v_bfe_u32 v0, v7, v0, 8
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshrrev_b32_sdwa v1, v1, v8 dst_sel:BYTE_3 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT:    v_lshl_or_b32 v0, v0, 8, v0
; GFX10-NEXT:    v_or3_b32 v0, v0, v2, v1
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extractbase:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v7, v[0:1], off
; GFX9-NEXT:    global_load_dword v8, v[2:3], off
; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 3, v6
; GFX9-NEXT:    v_add_u32_e32 v1, 24, v0
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_bfe_u32 v0, v7, v0, 8
; GFX9-NEXT:    v_bfe_u32 v2, v7, v1, 8
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_lshrrev_b32_sdwa v1, v1, v8 dst_sel:BYTE_3 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX9-NEXT:    v_lshl_or_b32 v0, v0, 8, v0
; GFX9-NEXT:    v_or3_b32 v0, v0, v2, v1
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <4 x i8>, ptr addrspace(1) %in0, align 4
  %vec2 = load <4 x i8>, ptr addrspace(1) %in1, align 4
  %v1b = extractelement <4 x i8> %vec1, i64 %base
  %zv1b = zext i8 %v1b to i32
  %byte1 = shl i32 %zv1b, 8

  %b3 = add i64 %base, 3
  %v1b3 = extractelement <4 x i8> %vec1, i64 %b3
  %zv1b3 = zext i8 %v1b3 to i32
  %byte2 = shl i32 %zv1b3, 16
  %v2b3 = extractelement <4 x i8> %vec2, i64 %b3
  %zv2b3 = zext i8 %v2b3 to i32
  %byte3 = shl i32 %zv2b3, 24

  %tmp0 = or i32 %zv1b, %byte1
  %tmp1 = or i32 %tmp0, %byte2
  %res = or i32 %tmp1, %byte3
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @extract_hilo(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: extract_hilo:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[2:3], off
; GFX10-NEXT:    global_load_dword v7, v[0:1], off offset:4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x3060505
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_hilo:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[2:3], off
; GFX9-NEXT:    global_load_dword v7, v[0:1], off offset:4
; GFX9-NEXT:    s_mov_b32 s4, 0x3060505
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <8 x i8>, ptr addrspace(1) %in0, align 4
  %vec2 = load <8 x i8>, ptr addrspace(1) %in1, align 4
  %v1e5 = extractelement <8 x i8> %vec1, i64 5
  %zv1e5 = zext i8 %v1e5 to i32
  %byte1 = shl i32 %zv1e5, 8

  %v1e6 = extractelement <8 x i8> %vec1, i64 6
  %zv1e6 = zext i8 %v1e6 to i32
  %byte2 = shl i32 %zv1e6, 16
  %v2e3 = extractelement <8 x i8> %vec2, i64 3
  %zv2e3 = zext i8 %v2e3 to i32
  %byte3 = shl i32 %zv2e3, 24

  %tmp0 = or i32 %zv1e5, %byte1
  %tmp1 = or i32 %tmp0, %byte2
  %res = or i32 %tmp1, %byte3
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @extract_lohi(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: extract_lohi:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[2:3], off offset:4
; GFX10-NEXT:    global_load_dword v7, v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x70404
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_lohi:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[2:3], off offset:4
; GFX9-NEXT:    global_load_dword v7, v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x70404
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <8 x i8>, ptr addrspace(1) %in0, align 4
  %vec2 = load <8 x i8>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <8 x i8> %vec1, i64 0
  %zv1e0 = zext i8 %v1e0 to i32
  %byte1 = shl i32 %zv1e0, 8

  %v1e3 = extractelement <8 x i8> %vec1, i64 3
  %zv1e3 = zext i8 %v1e3 to i32
  %byte2 = shl i32 %zv1e3, 16
  %v2e4 = extractelement <8 x i8> %vec2, i64 4
  %zv2e4 = zext i8 %v2e4 to i32
  %byte3 = shl i32 %zv2e4, 24

  %tmp0 = or i32 %zv1e0, %byte1
  %tmp1 = or i32 %tmp0, %byte2
  %res = or i32 %tmp1, %byte3
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @extract_hihi(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: extract_hihi:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v6, v[2:3], off offset:4
; GFX10-NEXT:    global_load_dword v7, v[0:1], off offset:4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v7, v6, 0x2070505
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_hihi:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v6, v[2:3], off offset:4
; GFX9-NEXT:    global_load_dword v7, v[0:1], off offset:4
; GFX9-NEXT:    s_mov_b32 s4, 0x2070505
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v7, v6, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <8 x i8>, ptr addrspace(1) %in0, align 4
  %vec2 = load <8 x i8>, ptr addrspace(1) %in1, align 4
  %v1e5 = extractelement <8 x i8> %vec1, i64 5
  %zv1e5 = zext i8 %v1e5 to i32
  %byte1 = shl i32 %zv1e5, 8

  %v1e7 = extractelement <8 x i8> %vec1, i64 7
  %zv1e7 = zext i8 %v1e7 to i32
  %byte2 = shl i32 %zv1e7, 16
  %v2e6 = extractelement <8 x i8> %vec2, i64 6
  %zv2e6 = zext i8 %v2e6 to i32
  %byte3 = shl i32 %zv2e6, 24

  %tmp0 = or i32 %zv1e5, %byte1
  %tmp1 = or i32 %tmp0, %byte2
  %res = or i32 %tmp1, %byte3
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @extract_v8i8(ptr addrspace(1) %in0, ptr addrspace(1) %out0) {
; GFX10-LABEL: extract_v8i8:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v1, v0, 0x1070404
; GFX10-NEXT:    global_store_dword v[2:3], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_v8i8:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
; GFX9-NEXT:    s_mov_b32 s4, 0x1070404
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v1, v0, s4
; GFX9-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <8 x i8>, ptr addrspace(1) %in0, align 4
  %v1e4 = extractelement <8 x i8> %vec1, i64 4
  %zv1e4 = zext i8 %v1e4 to i32
  %byte1 = shl i32 %zv1e4, 8

  %v1e7 = extractelement <8 x i8> %vec1, i64 7
  %zv1e7 = zext i8 %v1e7 to i32
  %byte2 = shl i32 %zv1e7, 16
  %v2e1 = extractelement <8 x i8> %vec1, i64 1
  %zv2e1 = zext i8 %v2e1 to i32
  %byte3 = shl i32 %zv2e1, 24

  %tmp0 = or i32 %zv1e4, %byte1
  %tmp1 = or i32 %tmp0, %byte2
  %res = or i32 %tmp1, %byte3
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

define hidden void @extract_v256i8(ptr addrspace(1) %in0, ptr addrspace(1) %out0) {
; GFX10-LABEL: extract_v256i8:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dword v0, v[0:1], off offset:252
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, v0, v0, 0x6050707
; GFX10-NEXT:    global_store_dword v[2:3], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_v256i8:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dword v0, v[0:1], off offset:252
; GFX9-NEXT:    s_mov_b32 s4, 0x6050707
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v0, v0, s4
; GFX9-NEXT:    global_store_dword v[2:3], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <256 x i8>, ptr addrspace(1) %in0, align 4
  %v1e4 = extractelement <256 x i8> %vec1, i64 255
  %zv1e4 = zext i8 %v1e4 to i32
  %byte1 = shl i32 %zv1e4, 8

  %v1e7 = extractelement <256 x i8> %vec1, i64 253
  %zv1e7 = zext i8 %v1e7 to i32
  %byte2 = shl i32 %zv1e7, 16
  %v2e1 = extractelement <256 x i8> %vec1, i64 254
  %zv2e1 = zext i8 %v2e1 to i32
  %byte3 = shl i32 %zv2e1, 24

  %tmp0 = or i32 %zv1e4, %byte1
  %tmp1 = or i32 %tmp0, %byte2
  %res = or i32 %tmp1, %byte3
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

; TODO : support this pattern
define hidden void @extract_3src(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: extract_3src:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dwordx2 v[6:7], v[0:1], off
; GFX10-NEXT:    global_load_dword v8, v[2:3], off offset:4
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 8, v7
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshlrev_b32_e32 v1, 8, v8
; GFX10-NEXT:    v_and_b32_e32 v2, 0xff, v6
; GFX10-NEXT:    v_and_b32_e32 v0, 0xff0000, v0
; GFX10-NEXT:    v_and_b32_e32 v1, 0xff000000, v1
; GFX10-NEXT:    v_lshl_or_b32 v2, v2, 8, v2
; GFX10-NEXT:    v_or3_b32 v0, v2, v0, v1
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_3src:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dwordx2 v[6:7], v[0:1], off
; GFX9-NEXT:    global_load_dword v8, v[2:3], off offset:4
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_and_b32_e32 v0, 0xff, v6
; GFX9-NEXT:    v_lshlrev_b32_e32 v1, 8, v7
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 8, v8
; GFX9-NEXT:    v_and_b32_e32 v1, 0xff0000, v1
; GFX9-NEXT:    v_and_b32_e32 v2, 0xff000000, v2
; GFX9-NEXT:    v_lshl_or_b32 v0, v0, 8, v0
; GFX9-NEXT:    v_or3_b32 v0, v0, v1, v2
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec1 = load <8 x i8>, ptr addrspace(1) %in0, align 4
  %vec2 = load <8 x i8>, ptr addrspace(1) %in1, align 4
  %v1e0 = extractelement <8 x i8> %vec1, i64 0
  %zv1e0 = zext i8 %v1e0 to i32
  %byte1 = shl i32 %zv1e0, 8

  %v1e5 = extractelement <8 x i8> %vec1, i64 5
  %zv1e5 = zext i8 %v1e5 to i32
  %byte2 = shl i32 %zv1e5, 16
  %v2e6 = extractelement <8 x i8> %vec2, i64 6
  %zv2e6 = zext i8 %v2e6 to i32
  %byte3 = shl i32 %zv2e6, 24

  %tmp0 = or i32 %zv1e0, %byte1
  %tmp1 = or i32 %tmp0, %byte2
  %res = or i32 %tmp1, %byte3
  store i32 %res, ptr addrspace(1) %out0, align 4
  ret void
}

; Should not result in crash
define hidden void @extract_v6i16(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: extract_v6i16:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    s_clause 0x3
; GFX10-NEXT:    global_load_ushort v2, v[0:1], off offset:6
; GFX10-NEXT:    global_load_ushort v3, v[0:1], off
; GFX10-NEXT:    global_load_ushort v8, v[0:1], off offset:2
; GFX10-NEXT:    global_load_ushort v9, v[0:1], off offset:4
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_lshl_or_b32 v0, v8, 16, v3
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_lshl_or_b32 v1, v2, 16, v9
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    global_store_dword v[6:7], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_v6i16:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_ushort v2, v[0:1], off offset:4
; GFX9-NEXT:    global_load_ushort v3, v[0:1], off offset:6
; GFX9-NEXT:    global_load_ushort v8, v[0:1], off
; GFX9-NEXT:    global_load_ushort v9, v[0:1], off offset:2
; GFX9-NEXT:    s_waitcnt vmcnt(2)
; GFX9-NEXT:    v_lshl_or_b32 v0, v3, 16, v2
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_lshl_or_b32 v1, v9, 16, v8
; GFX9-NEXT:    global_store_dword v[4:5], v1, off
; GFX9-NEXT:    global_store_dword v[6:7], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec = load <6 x i16>, ptr addrspace(1) %in0, align 2
  %el0 = extractelement <6 x i16> %vec, i32 0
  %el1 = extractelement <6 x i16> %vec, i32 1
  %el2 = extractelement <6 x i16> %vec, i32 2
  %el3 = extractelement <6 x i16> %vec, i32 3
  %z0 = zext i16 %el0 to i32
  %z1 = zext i16 %el1 to i32
  %s1 = shl nuw i32 %z1, 16
  %o0 = or i32 %s1, %z0
  %z2 = zext i16 %el2 to i32
  %z3 = zext i16 %el3 to i32
  %s3 = shl nuw i32 %z3, 16
  %o1 = or i32 %z2, %s3

  store i32 %o0, ptr addrspace(1) %out0, align 4
  store i32 %o1, ptr addrspace(1) %out1, align 4
  ret void
}


; Should not result in crash
define hidden void @extract_v7i16(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: extract_v7i16:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    global_store_dword v[6:7], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_v7i16:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    global_store_dword v[6:7], v1, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec = load <7 x i16>, ptr addrspace(1) %in0, align 2
  %el0 = extractelement <7 x i16> %vec, i32 0
  %el1 = extractelement <7 x i16> %vec, i32 1
  %el2 = extractelement <7 x i16> %vec, i32 2
  %el3 = extractelement <7 x i16> %vec, i32 3
  %z0 = zext i16 %el0 to i32
  %z1 = zext i16 %el1 to i32
  %s1 = shl nuw i32 %z1, 16
  %o0 = or i32 %s1, %z0
  %z2 = zext i16 %el2 to i32
  %z3 = zext i16 %el3 to i32
  %s3 = shl nuw i32 %z3, 16
  %o1 = or i32 %z2, %s3

  store i32 %o0, ptr addrspace(1) %out0, align 4
  store i32 %o1, ptr addrspace(1) %out1, align 4
  ret void
}

; Should not result in crash
define hidden void @extract_v13i8(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: extract_v13i8:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    s_clause 0x1
; GFX10-NEXT:    global_load_dwordx2 v[2:3], v[0:1], off
; GFX10-NEXT:    global_load_ushort v8, v[0:1], off offset:8
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_bfe_u32 v0, v2, 8, 8
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_and_b32_e32 v1, 0xff, v8
; GFX10-NEXT:    v_perm_b32 v0, v0, v2, 0x5040c00
; GFX10-NEXT:    v_perm_b32 v1, v1, v3, 0x5040c03
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    global_store_dword v[6:7], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_v13i8:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dwordx2 v[2:3], v[0:1], off
; GFX9-NEXT:    global_load_ushort v8, v[0:1], off offset:8
; GFX9-NEXT:    s_mov_b32 s4, 0x5040c00
; GFX9-NEXT:    s_mov_b32 s5, 0x5040c03
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_bfe_u32 v0, v2, 8, 8
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_and_b32_e32 v1, 0xff, v8
; GFX9-NEXT:    v_perm_b32 v0, v0, v2, s4
; GFX9-NEXT:    v_perm_b32 v1, v1, v3, s5
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    global_store_dword v[6:7], v1, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec = load <13 x i8>, ptr addrspace(1) %in0, align 2
  %el0 = extractelement <13 x i8> %vec, i32 0
  %el1 = extractelement <13 x i8> %vec, i32 1
  %el2 = extractelement <13 x i8> %vec, i32 7
  %el3 = extractelement <13 x i8> %vec, i32 8
  %z0 = zext i8 %el0 to i32
  %z1 = zext i8 %el1 to i32
  %s1 = shl nuw i32 %z1, 16
  %o0 = or i32 %s1, %z0
  %z2 = zext i8 %el2 to i32
  %z3 = zext i8 %el3 to i32
  %s3 = shl nuw i32 %z3, 16
  %o1 = or i32 %z2, %s3

  store i32 %o0, ptr addrspace(1) %out0, align 4
  store i32 %o1, ptr addrspace(1) %out1, align 4
  ret void
}

; Should not result in crash
define hidden void @extract_v13i64(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0, ptr addrspace(1) %out1) {
; GFX10-LABEL: extract_v13i64:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    s_clause 0x2
; GFX10-NEXT:    global_load_dwordx4 v[8:11], v[0:1], off offset:48
; GFX10-NEXT:    global_load_dwordx4 v[11:14], v[0:1], off
; GFX10-NEXT:    global_load_dwordx4 v[14:17], v[0:1], off offset:64
; GFX10-NEXT:    ; kill: killed $vgpr0 killed $vgpr1
; GFX10-NEXT:    s_waitcnt vmcnt(1)
; GFX10-NEXT:    v_perm_b32 v0, v12, v13, 0x1000504
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    v_perm_b32 v1, v10, v14, 0x1000504
; GFX10-NEXT:    global_store_dword v[4:5], v0, off
; GFX10-NEXT:    global_store_dword v[6:7], v1, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: extract_v13i64:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_dwordx4 v[8:11], v[0:1], off offset:48
; GFX9-NEXT:    global_load_dwordx4 v[11:14], v[0:1], off
; GFX9-NEXT:    global_load_dwordx4 v[14:17], v[0:1], off offset:64
; GFX9-NEXT:    s_mov_b32 s4, 0x1000504
; GFX9-NEXT:    ; kill: killed $vgpr0 killed $vgpr1
; GFX9-NEXT:    s_waitcnt vmcnt(1)
; GFX9-NEXT:    v_perm_b32 v0, v12, v13, s4
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v1, v10, v14, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    global_store_dword v[6:7], v1, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec = load <13 x i64>, ptr addrspace(1) %in0, align 2
  %el0 = extractelement <13 x i64> %vec, i32 0
  %el1 = extractelement <13 x i64> %vec, i32 1
  %el2 = extractelement <13 x i64> %vec, i32 7
  %el3 = extractelement <13 x i64> %vec, i32 8
  %el00 = lshr i64 %el0, 32
  %t0 = trunc i64 %el00 to i16
  %z0 = zext i16 %t0 to i32
  %z1 = trunc i64 %el1 to i32
  %s1 = shl nuw i32 %z1, 16
  %o0 = or i32 %s1, %z0
  %t2 = trunc i64 %el2 to i16
  %z2 = zext i16 %t2 to i32
  %z3 = trunc i64 %el3 to i32
  %s3 = shl nuw i32 %z3, 16
  %o1 = or i32 %z2, %s3

  store i32 %o0, ptr addrspace(1) %out0, align 4
  store i32 %o1, ptr addrspace(1) %out1, align 4
  ret void
}


; Should combine the lower 16 bits from each i32 in load
define hidden void @trunc_vector(ptr addrspace(1) %in0, ptr addrspace(1) %in1, ptr addrspace(1) %out0) {
; GFX10-LABEL: trunc_vector:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    s_clause 0x1
; GFX10-NEXT:    global_load_ushort v2, v[0:1], off
; GFX10-NEXT:    global_load_short_d16_hi v2, v[0:1], off offset:4
; GFX10-NEXT:    s_waitcnt vmcnt(0)
; GFX10-NEXT:    global_store_dword v[4:5], v2, off
; GFX10-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-LABEL: trunc_vector:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    global_load_ushort v2, v[0:1], off
; GFX9-NEXT:    global_load_ushort v3, v[0:1], off offset:4
; GFX9-NEXT:    s_mov_b32 s4, 0x5040100
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    v_perm_b32 v0, v3, v2, s4
; GFX9-NEXT:    global_store_dword v[4:5], v0, off
; GFX9-NEXT:    s_waitcnt vmcnt(0)
; GFX9-NEXT:    s_setpc_b64 s[30:31]
  %vec = load <2 x i32>, ptr addrspace(1) %in0, align 2
  %tvec = trunc <2 x i32> %vec to <2 x i16>
  %el0 = extractelement <2 x i16> %tvec, i32 0
  %el1 = extractelement <2 x i16> %tvec, i32 1
  %z0 = zext i16 %el0 to i32
  %z1 = zext i16 %el1 to i32
  %s1 = shl nuw i32 %z1, 16
  %o0 = or i32 %s1, %z0

  store i32 %o0, ptr addrspace(1) %out0, align 4
  ret void
}

; Should produce a v_perm instead of shift/and/or
define amdgpu_kernel void @any_extend_to_perm(i8 %arg, <4 x i8> %arg1) {
; GFX10-LABEL: any_extend_to_perm:
; GFX10:       ; %bb.0: ; %bb
; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; GFX10-NEXT:    v_mov_b32_e32 v0, 0x60504
; GFX10-NEXT:    v_mov_b32_e32 v1, 0
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    v_perm_b32 v0, s1, s0, v0
; GFX10-NEXT:    ds_write_b32 v1, v0
; GFX10-NEXT:    s_endpgm
;
; GFX9-LABEL: any_extend_to_perm:
; GFX9:       ; %bb.0: ; %bb
; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
; GFX9-NEXT:    v_mov_b32_e32 v0, 0x60504
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    v_mov_b32_e32 v1, s0
; GFX9-NEXT:    v_perm_b32 v0, s1, v1, v0
; GFX9-NEXT:    v_mov_b32_e32 v1, 0
; GFX9-NEXT:    ds_write_b32 v1, v0
; GFX9-NEXT:    s_endpgm
bb:
  %insertelement = insertelement <4 x i8> %arg1, i8 %arg, i64 3
  store <4 x i8> %insertelement, ptr addrspace(3) zeroinitializer, align 4
  ret void
}

; Similar to above to but should produce a series of v_perms with mask reuse. This was reduced from CK-provided IR
define amdgpu_kernel void @more_any_extend_to_perm(i8 %arg, <4 x i8> %arg1, i8 %arg2, ptr addrspace(3) %arg3, ptr addrspace(3) %arg4) {
; GFX10-LABEL: more_any_extend_to_perm:
; GFX10:       ; %bb.0: ; %bb
; GFX10-NEXT:    s_clause 0x1
; GFX10-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
; GFX10-NEXT:    v_mov_b32_e32 v0, 0xc0c0304
; GFX10-NEXT:    v_mov_b32_e32 v1, 0xc0c0104
; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
; GFX10-NEXT:    v_perm_b32 v2, s6, s7, v0
; GFX10-NEXT:    v_perm_b32 v0, s0, s7, v0
; GFX10-NEXT:    v_perm_b32 v3, s0, s7, v1
; GFX10-NEXT:    v_perm_b32 v1, s6, s7, v1
; GFX10-NEXT:    v_mov_b32_e32 v4, s1
; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
; GFX10-NEXT:    v_mov_b32_e32 v5, s2
; GFX10-NEXT:    v_or_b32_e32 v2, v3, v2
; GFX10-NEXT:    v_or_b32_e32 v0, v1, v0
; GFX10-NEXT:    ds_write_b32 v4, v2
; GFX10-NEXT:    ds_write_b32 v5, v0
; GFX10-NEXT:    s_endpgm
;
; GFX9-LABEL: more_any_extend_to_perm:
; GFX9:       ; %bb.0: ; %bb
; GFX9-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
; GFX9-NEXT:    v_mov_b32_e32 v1, 0xc0c0304
; GFX9-NEXT:    v_mov_b32_e32 v0, 0xc0c0104
; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
; GFX9-NEXT:    v_mov_b32_e32 v2, s7
; GFX9-NEXT:    v_perm_b32 v4, s6, v2, v1
; GFX9-NEXT:    v_perm_b32 v1, s0, v2, v1
; GFX9-NEXT:    v_perm_b32 v3, s0, v2, v0
; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 16, v4
; GFX9-NEXT:    v_perm_b32 v0, s6, v2, v0
; GFX9-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
; GFX9-NEXT:    v_or_b32_e32 v3, v3, v4
; GFX9-NEXT:    v_mov_b32_e32 v4, s1
; GFX9-NEXT:    v_or_b32_e32 v0, v0, v1
; GFX9-NEXT:    v_mov_b32_e32 v1, s2
; GFX9-NEXT:    ds_write_b32 v4, v3
; GFX9-NEXT:    ds_write_b32 v1, v0
; GFX9-NEXT:    s_endpgm
bb:
  %insertelement = insertelement <4 x i8> %arg1, i8 %arg2, i64 0
  %insertelement5 = insertelement <4 x i8> %insertelement, i8 %arg, i64 2
  store <4 x i8> %insertelement5, ptr addrspace(3) %arg3, align 4
  %insertelement6 = insertelement <4 x i8> %arg1, i8 %arg, i64 0
  %insertelement7 = insertelement <4 x i8> %insertelement6, i8 %arg2, i64 2
  store <4 x i8> %insertelement7, ptr addrspace(3) %arg4, align 4
  ret void
}