; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -verify-machineinstrs -simplifycfg-require-and-preserve-domtree=1 < %s | FileCheck -check-prefixes=GCN,GFX1032 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -verify-machineinstrs -simplifycfg-require-and-preserve-domtree=1 < %s | FileCheck -check-prefixes=GCN,GFX1064 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-early-ifcvt=1 -verify-machineinstrs -simplifycfg-require-and-preserve-domtree=1 < %s | FileCheck -check-prefixes=GCN,GFX1032 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-early-ifcvt=1 -verify-machineinstrs -simplifycfg-require-and-preserve-domtree=1 < %s | FileCheck -check-prefixes=GCN,GFX1064 %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs -simplifycfg-require-and-preserve-domtree=1 < %s | FileCheck -check-prefixes=GCN,GFX1032,GFX10DEFWAVE %s

define amdgpu_kernel void @test_vopc_i32(ptr addrspace(1) %arg) {
; GFX1032-LABEL: test_vopc_i32:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1032-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dword v1, v0, s[0:1]
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_cmp_lt_i32_e32 vcc_lo, 0, v1
; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 2, 1, vcc_lo
; GFX1032-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_vopc_i32:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1064-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dword v1, v0, s[0:1]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_cmp_lt_i32_e32 vcc, 0, v1
; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 2, 1, vcc
; GFX1064-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX1064-NEXT:    s_endpgm
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr addrspace(1) %arg, i32 %lid
  %load = load i32, ptr addrspace(1) %gep, align 4
  %cmp = icmp sgt i32 %load, 0
  %sel = select i1 %cmp, i32 1, i32 2
  store i32 %sel, ptr addrspace(1) %gep, align 4
  ret void
}

define amdgpu_kernel void @test_vopc_f32(ptr addrspace(1) %arg) {
; GFX1032-LABEL: test_vopc_f32:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1032-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dword v1, v0, s[0:1]
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_cmp_nge_f32_e32 vcc_lo, 0, v1
; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 2.0, 1.0, vcc_lo
; GFX1032-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_vopc_f32:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1064-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dword v1, v0, s[0:1]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_cmp_nge_f32_e32 vcc, 0, v1
; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 2.0, 1.0, vcc
; GFX1064-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX1064-NEXT:    s_endpgm
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds float, ptr addrspace(1) %arg, i32 %lid
  %load = load float, ptr addrspace(1) %gep, align 4
  %cmp = fcmp ugt float %load, 0.0
  %sel = select i1 %cmp, float 1.0, float 2.0
  store float %sel, ptr addrspace(1) %gep, align 4
  ret void
}

define amdgpu_ps void @test_vopc_vcmp(float %x) {
; GFX1032-LABEL: test_vopc_vcmp:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    v_cmp_nle_f32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, vcc_lo
; GFX1032-NEXT:    s_cbranch_scc0 .LBB2_1
; GFX1032-NEXT:    s_endpgm
; GFX1032-NEXT:  .LBB2_1:
; GFX1032-NEXT:    s_mov_b32 exec_lo, 0
; GFX1032-NEXT:    exp null off, off, off, off done vm
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_vopc_vcmp:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    v_cmp_nle_f32_e32 vcc, 0, v0
; GFX1064-NEXT:    s_andn2_b64 exec, exec, vcc
; GFX1064-NEXT:    s_cbranch_scc0 .LBB2_1
; GFX1064-NEXT:    s_endpgm
; GFX1064-NEXT:  .LBB2_1:
; GFX1064-NEXT:    s_mov_b64 exec, 0
; GFX1064-NEXT:    exp null off, off, off, off done vm
; GFX1064-NEXT:    s_endpgm
  %cmp = fcmp oge float %x, 0.0
  call void @llvm.amdgcn.kill(i1 %cmp)
  ret void
}

define amdgpu_kernel void @test_vopc_2xf16(ptr addrspace(1) %arg) {
; GFX1032-LABEL: test_vopc_2xf16:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1032-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dword v1, v0, s[0:1]
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_cmp_le_f16_sdwa vcc_lo, v1, v2 src0_sel:WORD_1 src1_sel:DWORD
; GFX1032-NEXT:    v_cndmask_b32_e32 v1, 0x3c003c00, v1, vcc_lo
; GFX1032-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_vopc_2xf16:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1064-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dword v1, v0, s[0:1]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_cmp_le_f16_sdwa vcc, v1, v2 src0_sel:WORD_1 src1_sel:DWORD
; GFX1064-NEXT:    v_cndmask_b32_e32 v1, 0x3c003c00, v1, vcc
; GFX1064-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX1064-NEXT:    s_endpgm
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds <2 x half>, ptr addrspace(1) %arg, i32 %lid
  %load = load <2 x half>, ptr addrspace(1) %gep, align 4
  %elt = extractelement <2 x half> %load, i32 1
  %cmp = fcmp ugt half %elt, 0.0
  %sel = select i1 %cmp, <2 x half> <half 1.0, half 1.0>, <2 x half> %load
  store <2 x half> %sel, ptr addrspace(1) %gep, align 4
  ret void
}

define amdgpu_kernel void @test_vopc_class(ptr addrspace(1) %out, float %x) #0 {
; GFX1032-LABEL: test_vopc_class:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_cmp_class_f32_e64 s0, s4, 0x204
; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 0, 1, s0
; GFX1032-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_vopc_class:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_cmp_class_f32_e64 s[0:1], s4, 0x204
; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 0, 1, s[0:1]
; GFX1064-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX1064-NEXT:    s_endpgm
  %fabs = tail call float @llvm.fabs.f32(float %x)
  %cmp = fcmp oeq float %fabs, 0x7FF0000000000000
  %ext = zext i1 %cmp to i32
  store i32 %ext, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @test_vcmp_vcnd_f16(ptr addrspace(1) %out, half %x) #0 {
; GFX1032-LABEL: test_vcmp_vcnd_f16:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_mov_b32_e32 v0, s4
; GFX1032-NEXT:    v_cmp_neq_f16_e64 vcc_lo, 0x7c00, s4
; GFX1032-NEXT:    v_cndmask_b32_e32 v0, 0x3c00, v0, vcc_lo
; GFX1032-NEXT:    global_store_short v1, v0, s[2:3]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_vcmp_vcnd_f16:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_mov_b32_e32 v0, s4
; GFX1064-NEXT:    v_cmp_neq_f16_e64 vcc, 0x7c00, s4
; GFX1064-NEXT:    v_cndmask_b32_e32 v0, 0x3c00, v0, vcc
; GFX1064-NEXT:    global_store_short v1, v0, s[2:3]
; GFX1064-NEXT:    s_endpgm
  %cmp = fcmp oeq half %x, 0x7FF0000000000000
  %sel = select i1 %cmp, half 1.0, half %x
  store half %sel, ptr addrspace(1) %out, align 2
  ret void
}

define amdgpu_kernel void @test_vop3_cmp_f32_sop_and(ptr addrspace(1) %arg) {
; GFX1032-LABEL: test_vop3_cmp_f32_sop_and:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1032-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_cmp_nge_f32_e32 vcc_lo, 0, v1
; GFX1032-NEXT:    v_cmp_nle_f32_e64 s0, 1.0, v1
; GFX1032-NEXT:    s_and_b32 s0, vcc_lo, s0
; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 2.0, 1.0, s0
; GFX1032-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_vop3_cmp_f32_sop_and:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1064-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_cmp_nge_f32_e32 vcc, 0, v1
; GFX1064-NEXT:    v_cmp_nle_f32_e64 s[0:1], 1.0, v1
; GFX1064-NEXT:    s_and_b64 s[0:1], vcc, s[0:1]
; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 2.0, 1.0, s[0:1]
; GFX1064-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX1064-NEXT:    s_endpgm
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds float, ptr addrspace(1) %arg, i32 %lid
  %load = load float, ptr addrspace(1) %gep, align 4
  %cmp = fcmp ugt float %load, 0.0
  %cmp2 = fcmp ult float %load, 1.0
  %and = and i1 %cmp, %cmp2
  %sel = select i1 %and, float 1.0, float 2.0
  store float %sel, ptr addrspace(1) %gep, align 4
  ret void
}

define amdgpu_kernel void @test_vop3_cmp_i32_sop_xor(ptr addrspace(1) %arg) {
; GFX1032-LABEL: test_vop3_cmp_i32_sop_xor:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1032-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_cmp_lt_i32_e32 vcc_lo, 0, v1
; GFX1032-NEXT:    v_cmp_gt_i32_e64 s0, 1, v1
; GFX1032-NEXT:    s_xor_b32 s0, vcc_lo, s0
; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 2, 1, s0
; GFX1032-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_vop3_cmp_i32_sop_xor:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1064-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_cmp_lt_i32_e32 vcc, 0, v1
; GFX1064-NEXT:    v_cmp_gt_i32_e64 s[0:1], 1, v1
; GFX1064-NEXT:    s_xor_b64 s[0:1], vcc, s[0:1]
; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 2, 1, s[0:1]
; GFX1064-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX1064-NEXT:    s_endpgm
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr addrspace(1) %arg, i32 %lid
  %load = load i32, ptr addrspace(1) %gep, align 4
  %cmp = icmp sgt i32 %load, 0
  %cmp2 = icmp slt i32 %load, 1
  %xor = xor i1 %cmp, %cmp2
  %sel = select i1 %xor, i32 1, i32 2
  store i32 %sel, ptr addrspace(1) %gep, align 4
  ret void
}

define amdgpu_kernel void @test_vop3_cmp_u32_sop_or(ptr addrspace(1) %arg) {
; GFX1032-LABEL: test_vop3_cmp_u32_sop_or:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1032-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_cmp_lt_u32_e32 vcc_lo, 3, v1
; GFX1032-NEXT:    v_cmp_gt_u32_e64 s0, 2, v1
; GFX1032-NEXT:    s_or_b32 s0, vcc_lo, s0
; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 2, 1, s0
; GFX1032-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_vop3_cmp_u32_sop_or:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1064-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dword v1, v0, s[2:3]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_cmp_lt_u32_e32 vcc, 3, v1
; GFX1064-NEXT:    v_cmp_gt_u32_e64 s[0:1], 2, v1
; GFX1064-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 2, 1, s[0:1]
; GFX1064-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX1064-NEXT:    s_endpgm
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %gep = getelementptr inbounds i32, ptr addrspace(1) %arg, i32 %lid
  %load = load i32, ptr addrspace(1) %gep, align 4
  %cmp = icmp ugt i32 %load, 3
  %cmp2 = icmp ult i32 %load, 2
  %or = or i1 %cmp, %cmp2
  %sel = select i1 %or, i32 1, i32 2
  store i32 %sel, ptr addrspace(1) %gep, align 4
  ret void
}

define amdgpu_kernel void @test_mask_if(ptr addrspace(1) %arg) #0 {
; GFX1032-LABEL: test_mask_if:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    v_cmp_lt_u32_e32 vcc_lo, 10, v0
; GFX1032-NEXT:    s_and_saveexec_b32 s2, vcc_lo
; GFX1032-NEXT:    s_cbranch_execz .LBB9_2
; GFX1032-NEXT:  ; %bb.1: ; %if
; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_store_dword v0, v0, s[0:1]
; GFX1032-NEXT:  .LBB9_2: ; %endif
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_mask_if:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    v_cmp_lt_u32_e32 vcc, 10, v0
; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
; GFX1064-NEXT:    s_cbranch_execz .LBB9_2
; GFX1064-NEXT:  ; %bb.1: ; %if
; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_store_dword v0, v0, s[0:1]
; GFX1064-NEXT:  .LBB9_2: ; %endif
; GFX1064-NEXT:    s_endpgm
  %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
  %cmp = icmp ugt i32 %lid, 10
  br i1 %cmp, label %if, label %endif

if:
  store i32 0, ptr addrspace(1) %arg, align 4
  br label %endif

endif:
  ret void
}

define amdgpu_kernel void @test_loop_with_if(ptr addrspace(1) %arg) #0 {
; GFX1032-LABEL: test_loop_with_if:
; GFX1032:       ; %bb.0: ; %bb
; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
; GFX1032-NEXT:    s_mov_b32 s2, 0
; GFX1032-NEXT:    ; implicit-def: $vgpr2_vgpr3
; GFX1032-NEXT:    s_branch .LBB10_2
; GFX1032-NEXT:  .LBB10_1: ; %bb13
; GFX1032-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s4
; GFX1032-NEXT:    v_cmp_lt_i32_e32 vcc_lo, 0xfe, v4
; GFX1032-NEXT:    v_add_nc_u32_e32 v1, 1, v4
; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
; GFX1032-NEXT:    s_cbranch_execz .LBB10_8
; GFX1032-NEXT:  .LBB10_2: ; %bb2
; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
; GFX1032-NEXT:    v_cmp_ge_i32_e64 s4, v1, v0
; GFX1032-NEXT:    v_cmp_lt_i32_e32 vcc_lo, v1, v0
; GFX1032-NEXT:    s_mov_b32 s3, 0
; GFX1032-NEXT:    s_and_saveexec_b32 s5, vcc_lo
; GFX1032-NEXT:    s_cbranch_execz .LBB10_4
; GFX1032-NEXT:  ; %bb.3: ; %bb5
; GFX1032-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1032-NEXT:    v_ashrrev_i32_e32 v2, 31, v1
; GFX1032-NEXT:    s_andn2_b32 s4, s4, exec_lo
; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
; GFX1032-NEXT:    v_lshlrev_b64 v[2:3], 2, v[1:2]
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_add_co_u32 v2, vcc_lo, s0, v2
; GFX1032-NEXT:    v_add_co_ci_u32_e32 v3, vcc_lo, s1, v3, vcc_lo
; GFX1032-NEXT:    global_load_dword v4, v[2:3], off
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_cmp_gt_i32_e32 vcc_lo, 11, v4
; GFX1032-NEXT:    s_and_b32 s6, vcc_lo, exec_lo
; GFX1032-NEXT:    s_or_b32 s4, s4, s6
; GFX1032-NEXT:  .LBB10_4: ; %Flow
; GFX1032-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s5
; GFX1032-NEXT:    ; implicit-def: $vgpr4
; GFX1032-NEXT:    s_and_saveexec_b32 s5, s4
; GFX1032-NEXT:    s_xor_b32 s4, exec_lo, s5
; GFX1032-NEXT:  ; %bb.5: ; %bb11
; GFX1032-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1032-NEXT:    v_lshrrev_b32_e32 v4, 31, v1
; GFX1032-NEXT:    s_andn2_b32 s3, s3, exec_lo
; GFX1032-NEXT:    v_add_nc_u32_e32 v4, v1, v4
; GFX1032-NEXT:    v_ashrrev_i32_e32 v4, 1, v4
; GFX1032-NEXT:  ; %bb.6: ; %Flow1
; GFX1032-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s4
; GFX1032-NEXT:    s_and_saveexec_b32 s4, s3
; GFX1032-NEXT:    s_cbranch_execz .LBB10_1
; GFX1032-NEXT:  ; %bb.7: ; %bb10
; GFX1032-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1032-NEXT:    v_mov_b32_e32 v4, v1
; GFX1032-NEXT:    global_store_dword v[2:3], v0, off
; GFX1032-NEXT:    s_branch .LBB10_1
; GFX1032-NEXT:  .LBB10_8: ; %bb1
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_loop_with_if:
; GFX1064:       ; %bb.0: ; %bb
; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
; GFX1064-NEXT:    ; implicit-def: $vgpr2_vgpr3
; GFX1064-NEXT:    s_branch .LBB10_2
; GFX1064-NEXT:  .LBB10_1: ; %bb13
; GFX1064-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
; GFX1064-NEXT:    s_or_b64 exec, exec, s[6:7]
; GFX1064-NEXT:    v_cmp_lt_i32_e32 vcc, 0xfe, v4
; GFX1064-NEXT:    v_add_nc_u32_e32 v1, 1, v4
; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
; GFX1064-NEXT:    s_cbranch_execz .LBB10_8
; GFX1064-NEXT:  .LBB10_2: ; %bb2
; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT:    v_cmp_ge_i32_e64 s[6:7], v1, v0
; GFX1064-NEXT:    v_cmp_lt_i32_e32 vcc, v1, v0
; GFX1064-NEXT:    s_mov_b64 s[4:5], 0
; GFX1064-NEXT:    s_and_saveexec_b64 s[8:9], vcc
; GFX1064-NEXT:    s_cbranch_execz .LBB10_4
; GFX1064-NEXT:  ; %bb.3: ; %bb5
; GFX1064-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1064-NEXT:    v_ashrrev_i32_e32 v2, 31, v1
; GFX1064-NEXT:    s_andn2_b64 s[6:7], s[6:7], exec
; GFX1064-NEXT:    s_mov_b64 s[4:5], exec
; GFX1064-NEXT:    v_lshlrev_b64 v[2:3], 2, v[1:2]
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_add_co_u32 v2, vcc, s0, v2
; GFX1064-NEXT:    v_add_co_ci_u32_e32 v3, vcc, s1, v3, vcc
; GFX1064-NEXT:    global_load_dword v4, v[2:3], off
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_cmp_gt_i32_e32 vcc, 11, v4
; GFX1064-NEXT:    s_and_b64 s[10:11], vcc, exec
; GFX1064-NEXT:    s_or_b64 s[6:7], s[6:7], s[10:11]
; GFX1064-NEXT:  .LBB10_4: ; %Flow
; GFX1064-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1064-NEXT:    s_or_b64 exec, exec, s[8:9]
; GFX1064-NEXT:    ; implicit-def: $vgpr4
; GFX1064-NEXT:    s_and_saveexec_b64 s[8:9], s[6:7]
; GFX1064-NEXT:    s_xor_b64 s[6:7], exec, s[8:9]
; GFX1064-NEXT:  ; %bb.5: ; %bb11
; GFX1064-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1064-NEXT:    v_lshrrev_b32_e32 v4, 31, v1
; GFX1064-NEXT:    s_andn2_b64 s[4:5], s[4:5], exec
; GFX1064-NEXT:    v_add_nc_u32_e32 v4, v1, v4
; GFX1064-NEXT:    v_ashrrev_i32_e32 v4, 1, v4
; GFX1064-NEXT:  ; %bb.6: ; %Flow1
; GFX1064-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1064-NEXT:    s_or_b64 exec, exec, s[6:7]
; GFX1064-NEXT:    s_and_saveexec_b64 s[6:7], s[4:5]
; GFX1064-NEXT:    s_cbranch_execz .LBB10_1
; GFX1064-NEXT:  ; %bb.7: ; %bb10
; GFX1064-NEXT:    ; in Loop: Header=BB10_2 Depth=1
; GFX1064-NEXT:    v_mov_b32_e32 v4, v1
; GFX1064-NEXT:    global_store_dword v[2:3], v0, off
; GFX1064-NEXT:    s_branch .LBB10_1
; GFX1064-NEXT:  .LBB10_8: ; %bb1
; GFX1064-NEXT:    s_endpgm
bb:
  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
  br label %bb2

bb1:
  ret void

bb2:
  %tmp3 = phi i32 [ 0, %bb ], [ %tmp15, %bb13 ]
  %tmp4 = icmp slt i32 %tmp3, %tmp
  br i1 %tmp4, label %bb5, label %bb11

bb5:
  %tmp6 = sext i32 %tmp3 to i64
  %tmp7 = getelementptr inbounds i32, ptr addrspace(1) %arg, i64 %tmp6
  %tmp8 = load i32, ptr addrspace(1) %tmp7, align 4
  %tmp9 = icmp sgt i32 %tmp8, 10
  br i1 %tmp9, label %bb10, label %bb11

bb10:
  store i32 %tmp, ptr addrspace(1) %tmp7, align 4
  br label %bb13

bb11:
  %tmp12 = sdiv i32 %tmp3, 2
  br label %bb13

bb13:
  %tmp14 = phi i32 [ %tmp3, %bb10 ], [ %tmp12, %bb11 ]
  %tmp15 = add nsw i32 %tmp14, 1
  %tmp16 = icmp slt i32 %tmp14, 255
  br i1 %tmp16, label %bb2, label %bb1
}


define amdgpu_kernel void @test_loop_with_if_else_break(ptr addrspace(1) %arg) #0 {
; GFX1032-LABEL: test_loop_with_if_else_break:
; GFX1032:       ; %bb.0: ; %bb
; GFX1032-NEXT:    v_cmp_ne_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    s_mov_b32 s2, 0
; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
; GFX1032-NEXT:    s_cbranch_execz .LBB11_6
; GFX1032-NEXT:  ; %bb.1: ; %.preheader
; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1032-NEXT:    v_min_u32_e32 v1, 0x100, v0
; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
; GFX1032-NEXT:    s_mov_b32 s3, 0
; GFX1032-NEXT:    ; implicit-def: $sgpr4
; GFX1032-NEXT:    s_branch .LBB11_4
; GFX1032-NEXT:  .LBB11_2: ; %bb8
; GFX1032-NEXT:    ; in Loop: Header=BB11_4 Depth=1
; GFX1032-NEXT:    s_add_i32 s3, s3, 1
; GFX1032-NEXT:    global_store_dword v2, v0, s[0:1]
; GFX1032-NEXT:    v_cmp_ge_u32_e32 vcc_lo, s3, v1
; GFX1032-NEXT:    s_add_u32 s0, s0, 4
; GFX1032-NEXT:    s_addc_u32 s1, s1, 0
; GFX1032-NEXT:    s_andn2_b32 s4, s4, exec_lo
; GFX1032-NEXT:    s_and_b32 s5, vcc_lo, exec_lo
; GFX1032-NEXT:    s_or_b32 s4, s4, s5
; GFX1032-NEXT:  .LBB11_3: ; %Flow
; GFX1032-NEXT:    ; in Loop: Header=BB11_4 Depth=1
; GFX1032-NEXT:    s_and_b32 s5, exec_lo, s4
; GFX1032-NEXT:    s_or_b32 s2, s5, s2
; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
; GFX1032-NEXT:    s_cbranch_execz .LBB11_6
; GFX1032-NEXT:  .LBB11_4: ; %bb2
; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dword v3, v2, s[0:1]
; GFX1032-NEXT:    s_or_b32 s4, s4, exec_lo
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_cmp_gt_i32_e32 vcc_lo, 11, v3
; GFX1032-NEXT:    s_cbranch_vccz .LBB11_2
; GFX1032-NEXT:  ; %bb.5: ; in Loop: Header=BB11_4 Depth=1
; GFX1032-NEXT:    ; implicit-def: $sgpr3
; GFX1032-NEXT:    ; implicit-def: $sgpr0_sgpr1
; GFX1032-NEXT:    s_branch .LBB11_3
; GFX1032-NEXT:  .LBB11_6: ; %.loopexit
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_loop_with_if_else_break:
; GFX1064:       ; %bb.0: ; %bb
; GFX1064-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
; GFX1064-NEXT:    s_mov_b32 s6, 0
; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
; GFX1064-NEXT:    s_cbranch_execz .LBB11_6
; GFX1064-NEXT:  ; %bb.1: ; %.preheader
; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1064-NEXT:    v_min_u32_e32 v1, 0x100, v0
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
; GFX1064-NEXT:    ; implicit-def: $sgpr4_sgpr5
; GFX1064-NEXT:    s_branch .LBB11_4
; GFX1064-NEXT:  .LBB11_2: ; %bb8
; GFX1064-NEXT:    ; in Loop: Header=BB11_4 Depth=1
; GFX1064-NEXT:    s_add_i32 s6, s6, 1
; GFX1064-NEXT:    global_store_dword v2, v0, s[0:1]
; GFX1064-NEXT:    v_cmp_ge_u32_e32 vcc, s6, v1
; GFX1064-NEXT:    s_add_u32 s0, s0, 4
; GFX1064-NEXT:    s_addc_u32 s1, s1, 0
; GFX1064-NEXT:    s_andn2_b64 s[4:5], s[4:5], exec
; GFX1064-NEXT:    s_and_b64 s[8:9], vcc, exec
; GFX1064-NEXT:    s_or_b64 s[4:5], s[4:5], s[8:9]
; GFX1064-NEXT:  .LBB11_3: ; %Flow
; GFX1064-NEXT:    ; in Loop: Header=BB11_4 Depth=1
; GFX1064-NEXT:    s_and_b64 s[8:9], exec, s[4:5]
; GFX1064-NEXT:    s_or_b64 s[2:3], s[8:9], s[2:3]
; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
; GFX1064-NEXT:    s_cbranch_execz .LBB11_6
; GFX1064-NEXT:  .LBB11_4: ; %bb2
; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dword v3, v2, s[0:1]
; GFX1064-NEXT:    s_or_b64 s[4:5], s[4:5], exec
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_cmp_gt_i32_e32 vcc, 11, v3
; GFX1064-NEXT:    s_cbranch_vccz .LBB11_2
; GFX1064-NEXT:  ; %bb.5: ; in Loop: Header=BB11_4 Depth=1
; GFX1064-NEXT:    ; implicit-def: $sgpr6
; GFX1064-NEXT:    ; implicit-def: $sgpr0_sgpr1
; GFX1064-NEXT:    s_branch .LBB11_3
; GFX1064-NEXT:  .LBB11_6: ; %.loopexit
; GFX1064-NEXT:    s_endpgm
bb:
  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
  %tmp1 = icmp eq i32 %tmp, 0
  br i1 %tmp1, label %.loopexit, label %.preheader

.preheader:
  br label %bb2

bb2:
  %tmp3 = phi i32 [ %tmp9, %bb8 ], [ 0, %.preheader ]
  %tmp4 = zext i32 %tmp3 to i64
  %tmp5 = getelementptr inbounds i32, ptr addrspace(1) %arg, i64 %tmp4
  %tmp6 = load i32, ptr addrspace(1) %tmp5, align 4
  %tmp7 = icmp sgt i32 %tmp6, 10
  br i1 %tmp7, label %bb8, label %.loopexit

bb8:
  store i32 %tmp, ptr addrspace(1) %tmp5, align 4
  %tmp9 = add nuw nsw i32 %tmp3, 1
  %tmp10 = icmp ult i32 %tmp9, 256
  %tmp11 = icmp ult i32 %tmp9, %tmp
  %tmp12 = and i1 %tmp10, %tmp11
  br i1 %tmp12, label %bb2, label %.loopexit

.loopexit:
  ret void
}

define amdgpu_kernel void @test_addc_vop2b(ptr addrspace(1) %arg, i64 %arg1) #0 {
; GFX1032-LABEL: test_addc_vop2b:
; GFX1032:       ; %bb.0: ; %bb
; GFX1032-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1032-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dwordx2 v[0:1], v2, s[0:1]
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_add_co_u32 v0, vcc_lo, v0, s2
; GFX1032-NEXT:    v_add_co_ci_u32_e32 v1, vcc_lo, s3, v1, vcc_lo
; GFX1032-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_addc_vop2b:
; GFX1064:       ; %bb.0: ; %bb
; GFX1064-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1064-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dwordx2 v[0:1], v2, s[0:1]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_add_co_u32 v0, vcc, v0, s2
; GFX1064-NEXT:    v_add_co_ci_u32_e32 v1, vcc, s3, v1, vcc
; GFX1064-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1064-NEXT:    s_endpgm
bb:
  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
  %tmp3 = getelementptr inbounds i64, ptr addrspace(1) %arg, i32 %tmp
  %tmp4 = load i64, ptr addrspace(1) %tmp3, align 8
  %tmp5 = add nsw i64 %tmp4, %arg1
  store i64 %tmp5, ptr addrspace(1) %tmp3, align 8
  ret void
}

define amdgpu_kernel void @test_subbrev_vop2b(ptr addrspace(1) %arg, i64 %arg1) #0 {
; GFX1032-LABEL: test_subbrev_vop2b:
; GFX1032:       ; %bb.0: ; %bb
; GFX1032-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1032-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dwordx2 v[0:1], v2, s[0:1]
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_sub_co_u32 v0, vcc_lo, v0, s2
; GFX1032-NEXT:    v_subrev_co_ci_u32_e32 v1, vcc_lo, s3, v1, vcc_lo
; GFX1032-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_subbrev_vop2b:
; GFX1064:       ; %bb.0: ; %bb
; GFX1064-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1064-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dwordx2 v[0:1], v2, s[0:1]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_sub_co_u32 v0, vcc, v0, s2
; GFX1064-NEXT:    v_subrev_co_ci_u32_e32 v1, vcc, s3, v1, vcc
; GFX1064-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1064-NEXT:    s_endpgm
bb:
  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
  %tmp3 = getelementptr inbounds i64, ptr addrspace(1) %arg, i32 %tmp
  %tmp4 = load i64, ptr addrspace(1) %tmp3, align 8
  %tmp5 = sub nsw i64 %tmp4, %arg1
  store i64 %tmp5, ptr addrspace(1) %tmp3, align 8
  ret void
}

define amdgpu_kernel void @test_subb_vop2b(ptr addrspace(1) %arg, i64 %arg1) #0 {
; GFX1032-LABEL: test_subb_vop2b:
; GFX1032:       ; %bb.0: ; %bb
; GFX1032-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1032-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dwordx2 v[0:1], v2, s[0:1]
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_sub_co_u32 v0, vcc_lo, s2, v0
; GFX1032-NEXT:    v_sub_co_ci_u32_e32 v1, vcc_lo, s3, v1, vcc_lo
; GFX1032-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_subb_vop2b:
; GFX1064:       ; %bb.0: ; %bb
; GFX1064-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1064-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dwordx2 v[0:1], v2, s[0:1]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_sub_co_u32 v0, vcc, s2, v0
; GFX1064-NEXT:    v_sub_co_ci_u32_e32 v1, vcc, s3, v1, vcc
; GFX1064-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1064-NEXT:    s_endpgm
bb:
  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
  %tmp3 = getelementptr inbounds i64, ptr addrspace(1) %arg, i32 %tmp
  %tmp4 = load i64, ptr addrspace(1) %tmp3, align 8
  %tmp5 = sub nsw i64 %arg1, %tmp4
  store i64 %tmp5, ptr addrspace(1) %tmp3, align 8
  ret void
}

define amdgpu_kernel void @test_udiv64(ptr addrspace(1) %arg) #0 {
; GFX1032-LABEL: test_udiv64:
; GFX1032:       ; %bb.0: ; %bb
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    s_or_b64 s[8:9], s[6:7], s[4:5]
; GFX1032-NEXT:    s_mov_b32 s8, 0
; GFX1032-NEXT:    s_cmp_lg_u64 s[8:9], 0
; GFX1032-NEXT:    s_cbranch_scc0 .LBB15_4
; GFX1032-NEXT:  ; %bb.1:
; GFX1032-NEXT:    v_cvt_f32_u32_e32 v0, s4
; GFX1032-NEXT:    v_cvt_f32_u32_e32 v1, s5
; GFX1032-NEXT:    s_sub_u32 s9, 0, s4
; GFX1032-NEXT:    s_subb_u32 s10, 0, s5
; GFX1032-NEXT:    v_madmk_f32 v0, v1, 0x4f800000, v0
; GFX1032-NEXT:    v_rcp_f32_e32 v0, v0
; GFX1032-NEXT:    v_mul_f32_e32 v0, 0x5f7ffffc, v0
; GFX1032-NEXT:    v_mul_f32_e32 v1, 0x2f800000, v0
; GFX1032-NEXT:    v_trunc_f32_e32 v1, v1
; GFX1032-NEXT:    v_madmk_f32 v0, v1, 0xcf800000, v0
; GFX1032-NEXT:    v_cvt_u32_f32_e32 v1, v1
; GFX1032-NEXT:    v_cvt_u32_f32_e32 v0, v0
; GFX1032-NEXT:    v_readfirstlane_b32 s0, v1
; GFX1032-NEXT:    v_readfirstlane_b32 s1, v0
; GFX1032-NEXT:    s_mul_i32 s11, s9, s0
; GFX1032-NEXT:    s_mul_hi_u32 s13, s9, s1
; GFX1032-NEXT:    s_mul_i32 s12, s10, s1
; GFX1032-NEXT:    s_add_i32 s11, s13, s11
; GFX1032-NEXT:    s_mul_i32 s14, s9, s1
; GFX1032-NEXT:    s_add_i32 s11, s11, s12
; GFX1032-NEXT:    s_mul_hi_u32 s13, s1, s14
; GFX1032-NEXT:    s_mul_hi_u32 s15, s0, s14
; GFX1032-NEXT:    s_mul_i32 s12, s0, s14
; GFX1032-NEXT:    s_mul_hi_u32 s14, s1, s11
; GFX1032-NEXT:    s_mul_i32 s1, s1, s11
; GFX1032-NEXT:    s_mul_hi_u32 s16, s0, s11
; GFX1032-NEXT:    s_add_u32 s1, s13, s1
; GFX1032-NEXT:    s_addc_u32 s13, 0, s14
; GFX1032-NEXT:    s_add_u32 s1, s1, s12
; GFX1032-NEXT:    s_mul_i32 s11, s0, s11
; GFX1032-NEXT:    s_addc_u32 s1, s13, s15
; GFX1032-NEXT:    s_addc_u32 s12, s16, 0
; GFX1032-NEXT:    s_add_u32 s1, s1, s11
; GFX1032-NEXT:    s_addc_u32 s11, 0, s12
; GFX1032-NEXT:    v_add_co_u32 v0, s1, v0, s1
; GFX1032-NEXT:    s_cmp_lg_u32 s1, 0
; GFX1032-NEXT:    s_addc_u32 s0, s0, s11
; GFX1032-NEXT:    v_readfirstlane_b32 s1, v0
; GFX1032-NEXT:    s_mul_i32 s11, s9, s0
; GFX1032-NEXT:    s_mul_hi_u32 s12, s9, s1
; GFX1032-NEXT:    s_mul_i32 s10, s10, s1
; GFX1032-NEXT:    s_add_i32 s11, s12, s11
; GFX1032-NEXT:    s_mul_i32 s9, s9, s1
; GFX1032-NEXT:    s_add_i32 s11, s11, s10
; GFX1032-NEXT:    s_mul_hi_u32 s12, s0, s9
; GFX1032-NEXT:    s_mul_i32 s13, s0, s9
; GFX1032-NEXT:    s_mul_hi_u32 s9, s1, s9
; GFX1032-NEXT:    s_mul_hi_u32 s14, s1, s11
; GFX1032-NEXT:    s_mul_i32 s1, s1, s11
; GFX1032-NEXT:    s_mul_hi_u32 s10, s0, s11
; GFX1032-NEXT:    s_add_u32 s1, s9, s1
; GFX1032-NEXT:    s_addc_u32 s9, 0, s14
; GFX1032-NEXT:    s_add_u32 s1, s1, s13
; GFX1032-NEXT:    s_mul_i32 s11, s0, s11
; GFX1032-NEXT:    s_addc_u32 s1, s9, s12
; GFX1032-NEXT:    s_addc_u32 s9, s10, 0
; GFX1032-NEXT:    s_add_u32 s1, s1, s11
; GFX1032-NEXT:    s_addc_u32 s9, 0, s9
; GFX1032-NEXT:    v_add_co_u32 v0, s1, v0, s1
; GFX1032-NEXT:    s_cmp_lg_u32 s1, 0
; GFX1032-NEXT:    s_addc_u32 s0, s0, s9
; GFX1032-NEXT:    v_readfirstlane_b32 s1, v0
; GFX1032-NEXT:    s_mul_i32 s10, s6, s0
; GFX1032-NEXT:    s_mul_hi_u32 s9, s6, s0
; GFX1032-NEXT:    s_mul_hi_u32 s11, s7, s0
; GFX1032-NEXT:    s_mul_i32 s0, s7, s0
; GFX1032-NEXT:    s_mul_hi_u32 s12, s6, s1
; GFX1032-NEXT:    s_mul_hi_u32 s13, s7, s1
; GFX1032-NEXT:    s_mul_i32 s1, s7, s1
; GFX1032-NEXT:    s_add_u32 s10, s12, s10
; GFX1032-NEXT:    s_addc_u32 s9, 0, s9
; GFX1032-NEXT:    s_add_u32 s1, s10, s1
; GFX1032-NEXT:    s_addc_u32 s1, s9, s13
; GFX1032-NEXT:    s_addc_u32 s9, s11, 0
; GFX1032-NEXT:    s_add_u32 s1, s1, s0
; GFX1032-NEXT:    s_addc_u32 s9, 0, s9
; GFX1032-NEXT:    s_mul_hi_u32 s0, s4, s1
; GFX1032-NEXT:    s_mul_i32 s11, s4, s9
; GFX1032-NEXT:    s_mul_i32 s12, s4, s1
; GFX1032-NEXT:    s_add_i32 s0, s0, s11
; GFX1032-NEXT:    v_sub_co_u32 v0, s11, s6, s12
; GFX1032-NEXT:    s_mul_i32 s10, s5, s1
; GFX1032-NEXT:    s_add_i32 s0, s0, s10
; GFX1032-NEXT:    v_sub_co_u32 v1, s12, v0, s4
; GFX1032-NEXT:    s_sub_i32 s10, s7, s0
; GFX1032-NEXT:    s_cmp_lg_u32 s11, 0
; GFX1032-NEXT:    s_subb_u32 s10, s10, s5
; GFX1032-NEXT:    s_cmp_lg_u32 s12, 0
; GFX1032-NEXT:    v_cmp_le_u32_e32 vcc_lo, s4, v1
; GFX1032-NEXT:    s_subb_u32 s10, s10, 0
; GFX1032-NEXT:    s_cmp_ge_u32 s10, s5
; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 0, -1, vcc_lo
; GFX1032-NEXT:    s_cselect_b32 s12, -1, 0
; GFX1032-NEXT:    s_cmp_eq_u32 s10, s5
; GFX1032-NEXT:    s_cselect_b32 vcc_lo, -1, 0
; GFX1032-NEXT:    s_add_u32 s10, s1, 1
; GFX1032-NEXT:    v_cndmask_b32_e32 v1, s12, v1, vcc_lo
; GFX1032-NEXT:    s_addc_u32 s12, s9, 0
; GFX1032-NEXT:    s_add_u32 s13, s1, 2
; GFX1032-NEXT:    s_addc_u32 s14, s9, 0
; GFX1032-NEXT:    s_cmp_lg_u32 s11, 0
; GFX1032-NEXT:    v_cmp_le_u32_e32 vcc_lo, s4, v0
; GFX1032-NEXT:    s_subb_u32 s0, s7, s0
; GFX1032-NEXT:    v_mov_b32_e32 v2, s13
; GFX1032-NEXT:    s_cmp_ge_u32 s0, s5
; GFX1032-NEXT:    v_cndmask_b32_e64 v0, 0, -1, vcc_lo
; GFX1032-NEXT:    s_cselect_b32 s7, -1, 0
; GFX1032-NEXT:    s_cmp_eq_u32 s0, s5
; GFX1032-NEXT:    v_cmp_ne_u32_e32 vcc_lo, 0, v1
; GFX1032-NEXT:    s_cselect_b32 s0, -1, 0
; GFX1032-NEXT:    v_mov_b32_e32 v1, s14
; GFX1032-NEXT:    v_cndmask_b32_e64 v0, s7, v0, s0
; GFX1032-NEXT:    v_cndmask_b32_e32 v2, s10, v2, vcc_lo
; GFX1032-NEXT:    v_cndmask_b32_e32 v1, s12, v1, vcc_lo
; GFX1032-NEXT:    v_cmp_ne_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    v_cndmask_b32_e32 v1, s9, v1, vcc_lo
; GFX1032-NEXT:    v_cndmask_b32_e32 v0, s1, v2, vcc_lo
; GFX1032-NEXT:    s_andn2_b32 vcc_lo, exec_lo, s8
; GFX1032-NEXT:    s_cbranch_vccnz .LBB15_3
; GFX1032-NEXT:  .LBB15_2:
; GFX1032-NEXT:    v_cvt_f32_u32_e32 v0, s4
; GFX1032-NEXT:    s_sub_i32 s1, 0, s4
; GFX1032-NEXT:    v_rcp_iflag_f32_e32 v0, v0
; GFX1032-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX1032-NEXT:    v_cvt_u32_f32_e32 v0, v0
; GFX1032-NEXT:    v_readfirstlane_b32 s0, v0
; GFX1032-NEXT:    s_mul_i32 s1, s1, s0
; GFX1032-NEXT:    s_mul_hi_u32 s1, s0, s1
; GFX1032-NEXT:    s_add_i32 s0, s0, s1
; GFX1032-NEXT:    s_mul_hi_u32 s0, s6, s0
; GFX1032-NEXT:    s_mul_i32 s1, s0, s4
; GFX1032-NEXT:    s_add_i32 s5, s0, 1
; GFX1032-NEXT:    s_sub_i32 s1, s6, s1
; GFX1032-NEXT:    s_sub_i32 s6, s1, s4
; GFX1032-NEXT:    s_cmp_ge_u32 s1, s4
; GFX1032-NEXT:    s_cselect_b32 s0, s5, s0
; GFX1032-NEXT:    s_cselect_b32 s1, s6, s1
; GFX1032-NEXT:    s_add_i32 s5, s0, 1
; GFX1032-NEXT:    s_cmp_ge_u32 s1, s4
; GFX1032-NEXT:    s_mov_b32 s1, 0
; GFX1032-NEXT:    s_cselect_b32 s0, s5, s0
; GFX1032-NEXT:    v_mov_b32_e32 v0, s0
; GFX1032-NEXT:    v_mov_b32_e32 v1, s1
; GFX1032-NEXT:  .LBB15_3:
; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
; GFX1032-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3] offset:16
; GFX1032-NEXT:    s_endpgm
; GFX1032-NEXT:  .LBB15_4:
; GFX1032-NEXT:    ; implicit-def: $vgpr0_vgpr1
; GFX1032-NEXT:    s_branch .LBB15_2
;
; GFX1064-LABEL: test_udiv64:
; GFX1064:       ; %bb.0: ; %bb
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    s_or_b64 s[0:1], s[6:7], s[4:5]
; GFX1064-NEXT:    s_mov_b32 s0, 0
; GFX1064-NEXT:    s_cmp_lg_u64 s[0:1], 0
; GFX1064-NEXT:    s_cbranch_scc0 .LBB15_4
; GFX1064-NEXT:  ; %bb.1:
; GFX1064-NEXT:    v_cvt_f32_u32_e32 v0, s4
; GFX1064-NEXT:    v_cvt_f32_u32_e32 v1, s5
; GFX1064-NEXT:    s_sub_u32 s9, 0, s4
; GFX1064-NEXT:    s_subb_u32 s10, 0, s5
; GFX1064-NEXT:    v_madmk_f32 v0, v1, 0x4f800000, v0
; GFX1064-NEXT:    v_rcp_f32_e32 v0, v0
; GFX1064-NEXT:    v_mul_f32_e32 v0, 0x5f7ffffc, v0
; GFX1064-NEXT:    v_mul_f32_e32 v1, 0x2f800000, v0
; GFX1064-NEXT:    v_trunc_f32_e32 v1, v1
; GFX1064-NEXT:    v_madmk_f32 v0, v1, 0xcf800000, v0
; GFX1064-NEXT:    v_cvt_u32_f32_e32 v1, v1
; GFX1064-NEXT:    v_cvt_u32_f32_e32 v0, v0
; GFX1064-NEXT:    v_readfirstlane_b32 s8, v1
; GFX1064-NEXT:    v_readfirstlane_b32 s0, v0
; GFX1064-NEXT:    s_mul_i32 s1, s9, s8
; GFX1064-NEXT:    s_mul_hi_u32 s12, s9, s0
; GFX1064-NEXT:    s_mul_i32 s11, s10, s0
; GFX1064-NEXT:    s_add_i32 s1, s12, s1
; GFX1064-NEXT:    s_mul_i32 s13, s9, s0
; GFX1064-NEXT:    s_add_i32 s1, s1, s11
; GFX1064-NEXT:    s_mul_hi_u32 s12, s0, s13
; GFX1064-NEXT:    s_mul_hi_u32 s14, s8, s13
; GFX1064-NEXT:    s_mul_i32 s11, s8, s13
; GFX1064-NEXT:    s_mul_hi_u32 s13, s0, s1
; GFX1064-NEXT:    s_mul_i32 s0, s0, s1
; GFX1064-NEXT:    s_mul_hi_u32 s15, s8, s1
; GFX1064-NEXT:    s_add_u32 s0, s12, s0
; GFX1064-NEXT:    s_addc_u32 s12, 0, s13
; GFX1064-NEXT:    s_add_u32 s0, s0, s11
; GFX1064-NEXT:    s_mul_i32 s1, s8, s1
; GFX1064-NEXT:    s_addc_u32 s0, s12, s14
; GFX1064-NEXT:    s_addc_u32 s11, s15, 0
; GFX1064-NEXT:    s_add_u32 s0, s0, s1
; GFX1064-NEXT:    s_addc_u32 s11, 0, s11
; GFX1064-NEXT:    v_add_co_u32 v0, s[0:1], v0, s0
; GFX1064-NEXT:    s_cmp_lg_u64 s[0:1], 0
; GFX1064-NEXT:    s_addc_u32 s8, s8, s11
; GFX1064-NEXT:    v_readfirstlane_b32 s0, v0
; GFX1064-NEXT:    s_mul_i32 s1, s9, s8
; GFX1064-NEXT:    s_mul_hi_u32 s11, s9, s0
; GFX1064-NEXT:    s_mul_i32 s10, s10, s0
; GFX1064-NEXT:    s_add_i32 s1, s11, s1
; GFX1064-NEXT:    s_mul_i32 s9, s9, s0
; GFX1064-NEXT:    s_add_i32 s1, s1, s10
; GFX1064-NEXT:    s_mul_hi_u32 s11, s8, s9
; GFX1064-NEXT:    s_mul_i32 s12, s8, s9
; GFX1064-NEXT:    s_mul_hi_u32 s9, s0, s9
; GFX1064-NEXT:    s_mul_hi_u32 s13, s0, s1
; GFX1064-NEXT:    s_mul_i32 s0, s0, s1
; GFX1064-NEXT:    s_mul_hi_u32 s10, s8, s1
; GFX1064-NEXT:    s_add_u32 s0, s9, s0
; GFX1064-NEXT:    s_addc_u32 s9, 0, s13
; GFX1064-NEXT:    s_add_u32 s0, s0, s12
; GFX1064-NEXT:    s_mul_i32 s1, s8, s1
; GFX1064-NEXT:    s_addc_u32 s0, s9, s11
; GFX1064-NEXT:    s_addc_u32 s9, s10, 0
; GFX1064-NEXT:    s_add_u32 s0, s0, s1
; GFX1064-NEXT:    s_addc_u32 s9, 0, s9
; GFX1064-NEXT:    v_add_co_u32 v0, s[0:1], v0, s0
; GFX1064-NEXT:    s_cmp_lg_u64 s[0:1], 0
; GFX1064-NEXT:    s_addc_u32 s0, s8, s9
; GFX1064-NEXT:    v_readfirstlane_b32 s1, v0
; GFX1064-NEXT:    s_mul_i32 s9, s6, s0
; GFX1064-NEXT:    s_mul_hi_u32 s8, s6, s0
; GFX1064-NEXT:    s_mul_hi_u32 s10, s7, s0
; GFX1064-NEXT:    s_mul_i32 s0, s7, s0
; GFX1064-NEXT:    s_mul_hi_u32 s11, s6, s1
; GFX1064-NEXT:    s_mul_hi_u32 s12, s7, s1
; GFX1064-NEXT:    s_mul_i32 s1, s7, s1
; GFX1064-NEXT:    s_add_u32 s9, s11, s9
; GFX1064-NEXT:    s_addc_u32 s8, 0, s8
; GFX1064-NEXT:    s_add_u32 s1, s9, s1
; GFX1064-NEXT:    s_addc_u32 s1, s8, s12
; GFX1064-NEXT:    s_addc_u32 s8, s10, 0
; GFX1064-NEXT:    s_add_u32 s10, s1, s0
; GFX1064-NEXT:    s_addc_u32 s11, 0, s8
; GFX1064-NEXT:    s_mul_hi_u32 s0, s4, s10
; GFX1064-NEXT:    s_mul_i32 s1, s4, s11
; GFX1064-NEXT:    s_mul_i32 s9, s4, s10
; GFX1064-NEXT:    s_add_i32 s12, s0, s1
; GFX1064-NEXT:    v_sub_co_u32 v0, s[0:1], s6, s9
; GFX1064-NEXT:    s_mul_i32 s8, s5, s10
; GFX1064-NEXT:    s_add_i32 s12, s12, s8
; GFX1064-NEXT:    v_sub_co_u32 v1, s[8:9], v0, s4
; GFX1064-NEXT:    s_sub_i32 s13, s7, s12
; GFX1064-NEXT:    s_cmp_lg_u64 s[0:1], 0
; GFX1064-NEXT:    s_subb_u32 s13, s13, s5
; GFX1064-NEXT:    s_cmp_lg_u64 s[8:9], 0
; GFX1064-NEXT:    v_cmp_le_u32_e32 vcc, s4, v1
; GFX1064-NEXT:    s_subb_u32 s8, s13, 0
; GFX1064-NEXT:    s_cmp_ge_u32 s8, s5
; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 0, -1, vcc
; GFX1064-NEXT:    s_cselect_b32 s9, -1, 0
; GFX1064-NEXT:    s_cmp_eq_u32 s8, s5
; GFX1064-NEXT:    s_cselect_b64 vcc, -1, 0
; GFX1064-NEXT:    s_add_u32 s8, s10, 1
; GFX1064-NEXT:    v_cndmask_b32_e32 v1, s9, v1, vcc
; GFX1064-NEXT:    s_addc_u32 s9, s11, 0
; GFX1064-NEXT:    s_add_u32 s13, s10, 2
; GFX1064-NEXT:    s_addc_u32 s14, s11, 0
; GFX1064-NEXT:    s_cmp_lg_u64 s[0:1], 0
; GFX1064-NEXT:    v_cmp_le_u32_e32 vcc, s4, v0
; GFX1064-NEXT:    s_subb_u32 s0, s7, s12
; GFX1064-NEXT:    v_mov_b32_e32 v2, s13
; GFX1064-NEXT:    s_cmp_ge_u32 s0, s5
; GFX1064-NEXT:    v_cndmask_b32_e64 v0, 0, -1, vcc
; GFX1064-NEXT:    s_cselect_b32 s7, -1, 0
; GFX1064-NEXT:    s_cmp_eq_u32 s0, s5
; GFX1064-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v1
; GFX1064-NEXT:    s_cselect_b64 s[0:1], -1, 0
; GFX1064-NEXT:    v_mov_b32_e32 v1, s14
; GFX1064-NEXT:    v_cndmask_b32_e64 v0, s7, v0, s[0:1]
; GFX1064-NEXT:    v_cndmask_b32_e32 v2, s8, v2, vcc
; GFX1064-NEXT:    v_cndmask_b32_e32 v1, s9, v1, vcc
; GFX1064-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
; GFX1064-NEXT:    v_cndmask_b32_e32 v1, s11, v1, vcc
; GFX1064-NEXT:    v_cndmask_b32_e32 v0, s10, v2, vcc
; GFX1064-NEXT:    s_cbranch_execnz .LBB15_3
; GFX1064-NEXT:  .LBB15_2:
; GFX1064-NEXT:    v_cvt_f32_u32_e32 v0, s4
; GFX1064-NEXT:    s_sub_i32 s1, 0, s4
; GFX1064-NEXT:    v_rcp_iflag_f32_e32 v0, v0
; GFX1064-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
; GFX1064-NEXT:    v_cvt_u32_f32_e32 v0, v0
; GFX1064-NEXT:    v_readfirstlane_b32 s0, v0
; GFX1064-NEXT:    s_mul_i32 s1, s1, s0
; GFX1064-NEXT:    s_mul_hi_u32 s1, s0, s1
; GFX1064-NEXT:    s_add_i32 s0, s0, s1
; GFX1064-NEXT:    s_mul_hi_u32 s0, s6, s0
; GFX1064-NEXT:    s_mul_i32 s1, s0, s4
; GFX1064-NEXT:    s_add_i32 s5, s0, 1
; GFX1064-NEXT:    s_sub_i32 s1, s6, s1
; GFX1064-NEXT:    s_sub_i32 s6, s1, s4
; GFX1064-NEXT:    s_cmp_ge_u32 s1, s4
; GFX1064-NEXT:    s_cselect_b32 s0, s5, s0
; GFX1064-NEXT:    s_cselect_b32 s1, s6, s1
; GFX1064-NEXT:    s_add_i32 s5, s0, 1
; GFX1064-NEXT:    s_cmp_ge_u32 s1, s4
; GFX1064-NEXT:    s_mov_b32 s1, 0
; GFX1064-NEXT:    s_cselect_b32 s0, s5, s0
; GFX1064-NEXT:    v_mov_b32_e32 v0, s0
; GFX1064-NEXT:    v_mov_b32_e32 v1, s1
; GFX1064-NEXT:  .LBB15_3:
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3] offset:16
; GFX1064-NEXT:    s_endpgm
; GFX1064-NEXT:  .LBB15_4:
; GFX1064-NEXT:    ; implicit-def: $vgpr0_vgpr1
; GFX1064-NEXT:    s_branch .LBB15_2
bb:
  %tmp = getelementptr inbounds i64, ptr addrspace(1) %arg, i64 1
  %tmp1 = load i64, ptr addrspace(1) %tmp, align 8
  %tmp2 = load i64, ptr addrspace(1) %arg, align 8
  %tmp3 = udiv i64 %tmp1, %tmp2
  %tmp4 = getelementptr inbounds i64, ptr addrspace(1) %arg, i64 2
  store i64 %tmp3, ptr addrspace(1) %tmp4, align 8
  ret void
}

define amdgpu_kernel void @test_div_scale_f32(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
; GFX1032-LABEL: test_div_scale_f32:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1032-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dword v1, v0, s[2:3] glc dlc
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    global_load_dword v2, v0, s[2:3] offset:4 glc dlc
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    v_div_scale_f32 v1, s2, v2, v2, v1
; GFX1032-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_div_scale_f32:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1064-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dword v1, v0, s[2:3] glc dlc
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    global_load_dword v2, v0, s[2:3] offset:4 glc dlc
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    v_div_scale_f32 v1, s[2:3], v2, v2, v1
; GFX1064-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX1064-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
  %gep.0 = getelementptr float, ptr addrspace(1) %in, i32 %tid
  %gep.1 = getelementptr float, ptr addrspace(1) %gep.0, i32 1

  %a = load volatile float, ptr addrspace(1) %gep.0, align 4
  %b = load volatile float, ptr addrspace(1) %gep.1, align 4

  %result = call { float, i1 } @llvm.amdgcn.div.scale.f32(float %a, float %b, i1 false) nounwind readnone
  %result0 = extractvalue { float, i1 } %result, 0
  store float %result0, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @test_div_scale_f64(ptr addrspace(1) %out, ptr addrspace(1) %aptr, ptr addrspace(1) %in) #0 {
; GFX1032-LABEL: test_div_scale_f64:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
; GFX1032-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dwordx2 v[0:1], v4, s[2:3] glc dlc
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3] offset:8 glc dlc
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_div_scale_f64 v[0:1], s2, v[0:1], v[2:3], v[0:1]
; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
; GFX1032-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_div_scale_f64:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
; GFX1064-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dwordx2 v[0:1], v4, s[2:3] glc dlc
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3] offset:8 glc dlc
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_div_scale_f64 v[0:1], s[2:3], v[0:1], v[2:3], v[0:1]
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1064-NEXT:    s_endpgm
  %tid = call i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
  %gep.0 = getelementptr double, ptr addrspace(1) %in, i32 %tid
  %gep.1 = getelementptr double, ptr addrspace(1) %gep.0, i32 1

  %a = load volatile double, ptr addrspace(1) %gep.0, align 8
  %b = load volatile double, ptr addrspace(1) %gep.1, align 8

  %result = call { double, i1 } @llvm.amdgcn.div.scale.f64(double %a, double %b, i1 true) nounwind readnone
  %result0 = extractvalue { double, i1 } %result, 0
  store double %result0, ptr addrspace(1) %out, align 8
  ret void
}

define i64 @test_mad_i64_i32(i32 %arg0, i32 %arg1, i64 %arg2) #0 {
; GFX1032-LABEL: test_mad_i64_i32:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1032-NEXT:    v_mad_i64_i32 v[0:1], s4, v0, v1, v[2:3]
; GFX1032-NEXT:    s_setpc_b64 s[30:31]
;
; GFX1064-LABEL: test_mad_i64_i32:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1064-NEXT:    v_mad_i64_i32 v[0:1], s[4:5], v0, v1, v[2:3]
; GFX1064-NEXT:    s_setpc_b64 s[30:31]
  %sext0 = sext i32 %arg0 to i64
  %sext1 = sext i32 %arg1 to i64
  %mul = mul i64 %sext0, %sext1
  %mad = add i64 %mul, %arg2
  ret i64 %mad
}

define i64 @test_mad_u64_u32(i32 %arg0, i32 %arg1, i64 %arg2) #0 {
; GFX1032-LABEL: test_mad_u64_u32:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1032-NEXT:    v_mad_u64_u32 v[0:1], s4, v0, v1, v[2:3]
; GFX1032-NEXT:    s_setpc_b64 s[30:31]
;
; GFX1064-LABEL: test_mad_u64_u32:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1064-NEXT:    v_mad_u64_u32 v[0:1], s[4:5], v0, v1, v[2:3]
; GFX1064-NEXT:    s_setpc_b64 s[30:31]
  %sext0 = zext i32 %arg0 to i64
  %sext1 = zext i32 %arg1 to i64
  %mul = mul i64 %sext0, %sext1
  %mad = add i64 %mul, %arg2
  ret i64 %mad
}

define amdgpu_kernel void @test_div_fmas_f32(ptr addrspace(1) %out, float %a, float %b, float %c, i1 %d) nounwind {
; GFX1032-LABEL: test_div_fmas_f32:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x2c
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_mov_b32_e32 v0, s5
; GFX1032-NEXT:    v_mov_b32_e32 v1, s6
; GFX1032-NEXT:    s_bitcmp1_b32 s7, 0
; GFX1032-NEXT:    s_cselect_b32 vcc_lo, -1, 0
; GFX1032-NEXT:    v_div_fmas_f32 v0, s4, v0, v1
; GFX1032-NEXT:    global_store_dword v2, v0, s[2:3]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_div_fmas_f32:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x2c
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_mov_b32_e32 v0, s5
; GFX1064-NEXT:    v_mov_b32_e32 v1, s6
; GFX1064-NEXT:    s_bitcmp1_b32 s7, 0
; GFX1064-NEXT:    s_cselect_b64 vcc, -1, 0
; GFX1064-NEXT:    v_div_fmas_f32 v0, s4, v0, v1
; GFX1064-NEXT:    global_store_dword v2, v0, s[2:3]
; GFX1064-NEXT:    s_endpgm
  %result = call float @llvm.amdgcn.div.fmas.f32(float %a, float %b, float %c, i1 %d) nounwind readnone
  store float %result, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @test_div_fmas_f64(ptr addrspace(1) %out, double %a, double %b, double %c, i1 %d) nounwind {
; GFX1032-LABEL: test_div_fmas_f64:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x24
; GFX1032-NEXT:    s_load_dword s2, s[0:1], 0x44
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_mov_b32_e32 v0, s8
; GFX1032-NEXT:    v_mov_b32_e32 v1, s9
; GFX1032-NEXT:    v_mov_b32_e32 v2, s10
; GFX1032-NEXT:    v_mov_b32_e32 v3, s11
; GFX1032-NEXT:    s_bitcmp1_b32 s2, 0
; GFX1032-NEXT:    s_cselect_b32 vcc_lo, -1, 0
; GFX1032-NEXT:    v_div_fmas_f64 v[0:1], s[6:7], v[0:1], v[2:3]
; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
; GFX1032-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_div_fmas_f64:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x24
; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x44
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_mov_b32_e32 v0, s8
; GFX1064-NEXT:    v_mov_b32_e32 v1, s9
; GFX1064-NEXT:    v_mov_b32_e32 v2, s10
; GFX1064-NEXT:    v_mov_b32_e32 v3, s11
; GFX1064-NEXT:    s_bitcmp1_b32 s2, 0
; GFX1064-NEXT:    s_cselect_b64 vcc, -1, 0
; GFX1064-NEXT:    v_div_fmas_f64 v[0:1], s[6:7], v[0:1], v[2:3]
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
; GFX1064-NEXT:    s_endpgm
  %result = call double @llvm.amdgcn.div.fmas.f64(double %a, double %b, double %c, i1 %d) nounwind readnone
  store double %result, ptr addrspace(1) %out, align 8
  ret void
}


define amdgpu_kernel void @test_div_fmas_f32_i1_phi_vcc(ptr addrspace(1) %out, ptr addrspace(1) %in, ptr addrspace(1) %dummy) #0 {
; GFX1032-LABEL: test_div_fmas_f32_i1_phi_vcc:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
; GFX1032-NEXT:    v_lshlrev_b32_e32 v1, 2, v0
; GFX1032-NEXT:    s_mov_b32 null, 0
; GFX1032-NEXT:    v_cmp_eq_u32_e64 s0, 0, v0
; GFX1032-NEXT:    s_mov_b32 vcc_lo, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_load_dwordx3 v[1:3], v1, s[6:7]
; GFX1032-NEXT:    s_and_saveexec_b32 s1, s0
; GFX1032-NEXT:    s_cbranch_execz .LBB22_2
; GFX1032-NEXT:  ; %bb.1: ; %bb
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    global_load_dword v0, v0, s[2:3] glc dlc
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_cmp_ne_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    s_and_b32 vcc_lo, vcc_lo, exec_lo
; GFX1032-NEXT:  .LBB22_2: ; %exit
; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s1
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_div_fmas_f32 v1, v1, v2, v3
; GFX1032-NEXT:    global_store_dword v0, v1, s[4:5] offset:8
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_div_fmas_f32_i1_phi_vcc:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX1064-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x34
; GFX1064-NEXT:    v_lshlrev_b32_e32 v1, 2, v0
; GFX1064-NEXT:    s_mov_b32 null, 0
; GFX1064-NEXT:    v_cmp_eq_u32_e64 s[0:1], 0, v0
; GFX1064-NEXT:    s_mov_b64 vcc, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_load_dwordx3 v[1:3], v1, s[6:7]
; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], s[0:1]
; GFX1064-NEXT:    s_cbranch_execz .LBB22_2
; GFX1064-NEXT:  ; %bb.1: ; %bb
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    global_load_dword v0, v0, s[8:9] glc dlc
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
; GFX1064-NEXT:    s_and_b64 vcc, vcc, exec
; GFX1064-NEXT:  .LBB22_2: ; %exit
; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_div_fmas_f32 v1, v1, v2, v3
; GFX1064-NEXT:    global_store_dword v0, v1, s[4:5] offset:8
; GFX1064-NEXT:    s_endpgm
entry:
  %tid = call i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
  %gep.out = getelementptr float, ptr addrspace(1) %out, i32 2
  %gep.a = getelementptr float, ptr addrspace(1) %in, i32 %tid
  %gep.b = getelementptr float, ptr addrspace(1) %gep.a, i32 1
  %gep.c = getelementptr float, ptr addrspace(1) %gep.a, i32 2

  %a = load float, ptr addrspace(1) %gep.a
  %b = load float, ptr addrspace(1) %gep.b
  %c = load float, ptr addrspace(1) %gep.c

  %cmp0 = icmp eq i32 %tid, 0
  br i1 %cmp0, label %bb, label %exit

bb:
  %val = load volatile i32, ptr addrspace(1) %dummy
  %cmp1 = icmp ne i32 %val, 0
  br label %exit

exit:
  %cond = phi i1 [false, %entry], [%cmp1, %bb]
  %result = call float @llvm.amdgcn.div.fmas.f32(float %a, float %b, float %c, i1 %cond) nounwind readnone
  store float %result, ptr addrspace(1) %gep.out, align 4
  ret void
}


define amdgpu_kernel void @fdiv_f32(ptr addrspace(1) %out, float %a, float %b) #0 {
; GFX1032-LABEL: fdiv_f32:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_div_scale_f32 v0, s4, s3, s3, s2
; GFX1032-NEXT:    v_rcp_f32_e32 v1, v0
; GFX1032-NEXT:    v_fma_f32 v2, -v0, v1, 1.0
; GFX1032-NEXT:    v_fmac_f32_e32 v1, v2, v1
; GFX1032-NEXT:    v_div_scale_f32 v2, vcc_lo, s2, s3, s2
; GFX1032-NEXT:    v_mul_f32_e32 v3, v2, v1
; GFX1032-NEXT:    v_fma_f32 v4, -v0, v3, v2
; GFX1032-NEXT:    v_fmac_f32_e32 v3, v4, v1
; GFX1032-NEXT:    v_fma_f32 v0, -v0, v3, v2
; GFX1032-NEXT:    v_div_fmas_f32 v0, v0, v1, v3
; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
; GFX1032-NEXT:    v_div_fixup_f32 v0, v0, s3, s2
; GFX1032-NEXT:    global_store_dword v1, v0, s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: fdiv_f32:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_div_scale_f32 v0, s[4:5], s3, s3, s2
; GFX1064-NEXT:    v_rcp_f32_e32 v1, v0
; GFX1064-NEXT:    v_fma_f32 v2, -v0, v1, 1.0
; GFX1064-NEXT:    v_fmac_f32_e32 v1, v2, v1
; GFX1064-NEXT:    v_div_scale_f32 v2, vcc, s2, s3, s2
; GFX1064-NEXT:    v_mul_f32_e32 v3, v2, v1
; GFX1064-NEXT:    v_fma_f32 v4, -v0, v3, v2
; GFX1064-NEXT:    v_fmac_f32_e32 v3, v4, v1
; GFX1064-NEXT:    v_fma_f32 v0, -v0, v3, v2
; GFX1064-NEXT:    v_div_fmas_f32 v0, v0, v1, v3
; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
; GFX1064-NEXT:    v_div_fixup_f32 v0, v0, s3, s2
; GFX1064-NEXT:    global_store_dword v1, v0, s[0:1]
; GFX1064-NEXT:    s_endpgm
entry:
  %fdiv = fdiv float %a, %b
  store float %fdiv, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @test_br_cc_f16(
; GFX1032-LABEL: test_br_cc_f16:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    global_load_ushort v1, v0, s[6:7]
; GFX1032-NEXT:    global_load_ushort v2, v0, s[2:3]
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_cmp_nlt_f16_e32 vcc_lo, v1, v2
; GFX1032-NEXT:    s_cbranch_vccnz .LBB24_2
; GFX1032-NEXT:  ; %bb.1: ; %one
; GFX1032-NEXT:    global_store_short v0, v1, s[4:5]
; GFX1032-NEXT:    s_endpgm
; GFX1032-NEXT:  .LBB24_2: ; %two
; GFX1032-NEXT:    global_store_short v0, v2, s[4:5]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_br_cc_f16:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    global_load_ushort v1, v0, s[6:7]
; GFX1064-NEXT:    global_load_ushort v2, v0, s[2:3]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_cmp_nlt_f16_e32 vcc, v1, v2
; GFX1064-NEXT:    s_cbranch_vccnz .LBB24_2
; GFX1064-NEXT:  ; %bb.1: ; %one
; GFX1064-NEXT:    global_store_short v0, v1, s[4:5]
; GFX1064-NEXT:    s_endpgm
; GFX1064-NEXT:  .LBB24_2: ; %two
; GFX1064-NEXT:    global_store_short v0, v2, s[4:5]
; GFX1064-NEXT:    s_endpgm
    ptr addrspace(1) %r,
    ptr addrspace(1) %a,
    ptr addrspace(1) %b) {
entry:
  %a.val = load half, ptr addrspace(1) %a
  %b.val = load half, ptr addrspace(1) %b
  %fcmp = fcmp olt half %a.val, %b.val
  br i1 %fcmp, label %one, label %two

one:
  store half %a.val, ptr addrspace(1) %r
  ret void

two:
  store half %b.val, ptr addrspace(1) %r
  ret void
}

define amdgpu_kernel void @test_brcc_i1(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in, i1 %val) #0 {
; GCN-LABEL: test_brcc_i1:
; GCN:       ; %bb.0:
; GCN-NEXT:    s_load_dword s2, s[0:1], 0x34
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    s_bitcmp0_b32 s2, 0
; GCN-NEXT:    s_cbranch_scc1 .LBB25_2
; GCN-NEXT:  ; %bb.1: ; %store
; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GCN-NEXT:    v_mov_b32_e32 v0, 0
; GCN-NEXT:    v_mov_b32_e32 v1, 0xde
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    global_store_dword v0, v1, s[0:1]
; GCN-NEXT:  .LBB25_2: ; %end
; GCN-NEXT:    s_endpgm
  %cmp0 = icmp ne i1 %val, 0
  br i1 %cmp0, label %store, label %end

store:
  store i32 222, ptr addrspace(1) %out
  ret void

end:
  ret void
}

define amdgpu_kernel void @test_preserve_condition_undef_flag(float %arg, i32 %arg1, float %arg2) #0 {
; GFX1032-LABEL: test_preserve_condition_undef_flag:
; GFX1032:       ; %bb.0: ; %bb0
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    s_load_dword s2, s[0:1], 0x2c
; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x24
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_cmp_nlt_f32_e64 s0, s2, 1.0
; GFX1032-NEXT:    v_cmp_nlt_f32_e64 s1, s3, 1.0
; GFX1032-NEXT:    v_cmp_ngt_f32_e64 s2, s2, 0
; GFX1032-NEXT:    s_or_b32 s0, s0, s1
; GFX1032-NEXT:    s_or_b32 s0, s0, s2
; GFX1032-NEXT:    s_and_b32 vcc_lo, exec_lo, s0
; GFX1032-NEXT:    s_cbranch_vccnz .LBB26_2
; GFX1032-NEXT:  ; %bb.1: ; %bb1
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    global_store_dword v[0:1], v0, off
; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT:  .LBB26_2: ; %bb2
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_preserve_condition_undef_flag:
; GFX1064:       ; %bb.0: ; %bb0
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x24
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_cmp_nlt_f32_e64 s[0:1], s4, 1.0
; GFX1064-NEXT:    v_cmp_nlt_f32_e64 s[2:3], s2, 1.0
; GFX1064-NEXT:    v_cmp_ngt_f32_e64 s[4:5], s4, 0
; GFX1064-NEXT:    s_or_b64 s[0:1], s[0:1], s[2:3]
; GFX1064-NEXT:    s_or_b64 s[0:1], s[0:1], s[4:5]
; GFX1064-NEXT:    s_and_b64 vcc, exec, s[0:1]
; GFX1064-NEXT:    s_cbranch_vccnz .LBB26_2
; GFX1064-NEXT:  ; %bb.1: ; %bb1
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    global_store_dword v[0:1], v0, off
; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT:  .LBB26_2: ; %bb2
; GFX1064-NEXT:    s_endpgm
bb0:
  %tmp = icmp sgt i32 %arg1, 4
  %undef = call i1 @llvm.amdgcn.class.f32(float undef, i32 undef)
  %tmp4 = select i1 %undef, float %arg, float 1.000000e+00
  %tmp5 = fcmp ogt float %arg2, 0.000000e+00
  %tmp6 = fcmp olt float %arg2, 1.000000e+00
  %tmp7 = fcmp olt float %arg, %tmp4
  %tmp8 = and i1 %tmp5, %tmp6
  %tmp9 = and i1 %tmp8, %tmp7
  br i1 %tmp9, label %bb1, label %bb2

bb1:
  store volatile i32 0, ptr addrspace(1) undef
  br label %bb2

bb2:
  ret void
}

define amdgpu_kernel void @test_invert_true_phi_cond_break_loop(i32 %arg) #0 {
; GFX1032-LABEL: test_invert_true_phi_cond_break_loop:
; GFX1032:       ; %bb.0: ; %bb
; GFX1032-NEXT:    s_load_dword s0, s[0:1], 0x24
; GFX1032-NEXT:    ; implicit-def: $sgpr1
; GFX1032-NEXT:    ; implicit-def: $sgpr2
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_subrev_nc_u32_e32 v0, s0, v0
; GFX1032-NEXT:    s_mov_b32 s0, 0
; GFX1032-NEXT:    s_branch .LBB27_2
; GFX1032-NEXT:  .LBB27_1: ; %Flow
; GFX1032-NEXT:    ; in Loop: Header=BB27_2 Depth=1
; GFX1032-NEXT:    s_xor_b32 s3, s1, -1
; GFX1032-NEXT:    s_add_i32 s2, s2, 1
; GFX1032-NEXT:    s_and_b32 s3, exec_lo, s3
; GFX1032-NEXT:    s_or_b32 s0, s3, s0
; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s0
; GFX1032-NEXT:    s_cbranch_execz .LBB27_4
; GFX1032-NEXT:  .LBB27_2: ; %bb1
; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
; GFX1032-NEXT:    s_or_b32 s1, s1, exec_lo
; GFX1032-NEXT:    s_cmp_gt_i32 s2, -1
; GFX1032-NEXT:    s_cbranch_scc1 .LBB27_1
; GFX1032-NEXT:  ; %bb.3: ; %bb4
; GFX1032-NEXT:    ; in Loop: Header=BB27_2 Depth=1
; GFX1032-NEXT:    global_load_dword v1, v[0:1], off glc dlc
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    s_andn2_b32 s1, s1, exec_lo
; GFX1032-NEXT:    v_cmp_ge_i32_e32 vcc_lo, v0, v1
; GFX1032-NEXT:    s_and_b32 s3, vcc_lo, exec_lo
; GFX1032-NEXT:    s_or_b32 s1, s1, s3
; GFX1032-NEXT:    s_branch .LBB27_1
; GFX1032-NEXT:  .LBB27_4: ; %bb9
; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s0
; GFX1032-NEXT:    v_mov_b32_e32 v0, 7
; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
; GFX1032-NEXT:    ds_write_b32 v0, v0
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_invert_true_phi_cond_break_loop:
; GFX1064:       ; %bb.0: ; %bb
; GFX1064-NEXT:    s_load_dword s0, s[0:1], 0x24
; GFX1064-NEXT:    ; implicit-def: $sgpr2_sgpr3
; GFX1064-NEXT:    ; implicit-def: $sgpr4
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_subrev_nc_u32_e32 v0, s0, v0
; GFX1064-NEXT:    s_mov_b64 s[0:1], 0
; GFX1064-NEXT:    s_branch .LBB27_2
; GFX1064-NEXT:  .LBB27_1: ; %Flow
; GFX1064-NEXT:    ; in Loop: Header=BB27_2 Depth=1
; GFX1064-NEXT:    s_xor_b64 s[6:7], s[2:3], -1
; GFX1064-NEXT:    s_add_i32 s4, s4, 1
; GFX1064-NEXT:    s_and_b64 s[6:7], exec, s[6:7]
; GFX1064-NEXT:    s_or_b64 s[0:1], s[6:7], s[0:1]
; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[0:1]
; GFX1064-NEXT:    s_cbranch_execz .LBB27_4
; GFX1064-NEXT:  .LBB27_2: ; %bb1
; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT:    s_or_b64 s[2:3], s[2:3], exec
; GFX1064-NEXT:    s_cmp_gt_i32 s4, -1
; GFX1064-NEXT:    s_cbranch_scc1 .LBB27_1
; GFX1064-NEXT:  ; %bb.3: ; %bb4
; GFX1064-NEXT:    ; in Loop: Header=BB27_2 Depth=1
; GFX1064-NEXT:    global_load_dword v1, v[0:1], off glc dlc
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], exec
; GFX1064-NEXT:    v_cmp_ge_i32_e32 vcc, v0, v1
; GFX1064-NEXT:    s_and_b64 s[6:7], vcc, exec
; GFX1064-NEXT:    s_or_b64 s[2:3], s[2:3], s[6:7]
; GFX1064-NEXT:    s_branch .LBB27_1
; GFX1064-NEXT:  .LBB27_4: ; %bb9
; GFX1064-NEXT:    s_or_b64 exec, exec, s[0:1]
; GFX1064-NEXT:    v_mov_b32_e32 v0, 7
; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
; GFX1064-NEXT:    ds_write_b32 v0, v0
; GFX1064-NEXT:    s_endpgm
bb:
  %id = call i32 @llvm.amdgcn.workitem.id.x()
  %tmp = sub i32 %id, %arg
  br label %bb1

bb1:                                              ; preds = %Flow, %bb
  %lsr.iv = phi i32 [ undef, %bb ], [ %tmp2, %Flow ]
  %lsr.iv.next = add i32 %lsr.iv, 1
  %cmp0 = icmp slt i32 %lsr.iv.next, 0
  br i1 %cmp0, label %bb4, label %Flow

bb4:                                              ; preds = %bb1
  %load = load volatile i32, ptr addrspace(1) undef, align 4
  %cmp1 = icmp sge i32 %tmp, %load
  br label %Flow

Flow:                                             ; preds = %bb4, %bb1
  %tmp2 = phi i32 [ %lsr.iv.next, %bb4 ], [ undef, %bb1 ]
  %tmp3 = phi i1 [ %cmp1, %bb4 ], [ true, %bb1 ]
  br i1 %tmp3, label %bb1, label %bb9

bb9:                                              ; preds = %Flow
  store volatile i32 7, ptr addrspace(3) undef
  ret void
}

define amdgpu_kernel void @test_movrels_extract_neg_offset_vgpr(ptr addrspace(1) %out) #0 {
; GFX1032-LABEL: test_movrels_extract_neg_offset_vgpr:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    v_add_nc_u32_e32 v0, 0xfffffe00, v0
; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 1, v0
; GFX1032-NEXT:    v_cndmask_b32_e64 v1, 0, 1, vcc_lo
; GFX1032-NEXT:    v_cmp_ne_u32_e32 vcc_lo, 2, v0
; GFX1032-NEXT:    v_cndmask_b32_e32 v1, 2, v1, vcc_lo
; GFX1032-NEXT:    v_cmp_ne_u32_e32 vcc_lo, 3, v0
; GFX1032-NEXT:    v_cndmask_b32_e32 v0, 3, v1, vcc_lo
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    global_store_dword v2, v0, s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_movrels_extract_neg_offset_vgpr:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    v_add_nc_u32_e32 v0, 0xfffffe00, v0
; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v0
; GFX1064-NEXT:    v_cndmask_b32_e64 v1, 0, 1, vcc
; GFX1064-NEXT:    v_cmp_ne_u32_e32 vcc, 2, v0
; GFX1064-NEXT:    v_cndmask_b32_e32 v1, 2, v1, vcc
; GFX1064-NEXT:    v_cmp_ne_u32_e32 vcc, 3, v0
; GFX1064-NEXT:    v_cndmask_b32_e32 v0, 3, v1, vcc
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    global_store_dword v2, v0, s[0:1]
; GFX1064-NEXT:    s_endpgm
entry:
  %id = call i32 @llvm.amdgcn.workitem.id.x() #1
  %index = add i32 %id, -512
  %value = extractelement <4 x i32> <i32 0, i32 1, i32 2, i32 3>, i32 %index
  store i32 %value, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @test_set_inactive(ptr addrspace(1) %out, i32 %in) #0 {
; GFX1032-LABEL: test_set_inactive:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_mov_b32_e32 v0, s4
; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
; GFX1032-NEXT:    v_mov_b32_e32 v0, 42
; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
; GFX1032-NEXT:    global_store_dword v1, v0, s[2:3]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_set_inactive:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_mov_b32_e32 v0, s4
; GFX1064-NEXT:    s_not_b64 exec, exec
; GFX1064-NEXT:    v_mov_b32_e32 v0, 42
; GFX1064-NEXT:    s_not_b64 exec, exec
; GFX1064-NEXT:    global_store_dword v1, v0, s[2:3]
; GFX1064-NEXT:    s_endpgm
  %tmp = call i32 @llvm.amdgcn.set.inactive.i32(i32 %in, i32 42)
  store i32 %tmp, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @test_set_inactive_64(ptr addrspace(1) %out, i64 %in) #0 {
; GFX1032-LABEL: test_set_inactive_64:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_mov_b32_e32 v0, s2
; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
; GFX1032-NEXT:    s_not_b32 exec_lo, exec_lo
; GFX1032-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_set_inactive_64:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_mov_b32_e32 v0, s2
; GFX1064-NEXT:    v_mov_b32_e32 v1, s3
; GFX1064-NEXT:    s_not_b64 exec, exec
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    v_mov_b32_e32 v1, 0
; GFX1064-NEXT:    s_not_b64 exec, exec
; GFX1064-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1064-NEXT:    s_endpgm
  %tmp = call i64 @llvm.amdgcn.set.inactive.i64(i64 %in, i64 0)
  store i64 %tmp, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_kill_i1_terminator_float() #0 {
; GFX1032-LABEL: test_kill_i1_terminator_float:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, exec_lo
; GFX1032-NEXT:    s_cbranch_scc0 .LBB31_1
; GFX1032-NEXT:    s_endpgm
; GFX1032-NEXT:  .LBB31_1:
; GFX1032-NEXT:    s_mov_b32 exec_lo, 0
; GFX1032-NEXT:    exp null off, off, off, off done vm
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_kill_i1_terminator_float:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_andn2_b64 exec, exec, exec
; GFX1064-NEXT:    s_cbranch_scc0 .LBB31_1
; GFX1064-NEXT:    s_endpgm
; GFX1064-NEXT:  .LBB31_1:
; GFX1064-NEXT:    s_mov_b64 exec, 0
; GFX1064-NEXT:    exp null off, off, off, off done vm
; GFX1064-NEXT:    s_endpgm
  call void @llvm.amdgcn.kill(i1 false)
  ret void
}

define amdgpu_gs void @test_kill_i1_terminator_i1(i32 %a, i32 %b, i32 %c, i32 %d) #0 {
; GFX1032-LABEL: test_kill_i1_terminator_i1:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    v_cmp_lt_i32_e32 vcc_lo, v0, v1
; GFX1032-NEXT:    v_cmp_lt_i32_e64 s0, v2, v3
; GFX1032-NEXT:    s_mov_b32 s1, exec_lo
; GFX1032-NEXT:    s_or_b32 s0, vcc_lo, s0
; GFX1032-NEXT:    s_xor_b32 s0, s0, exec_lo
; GFX1032-NEXT:    s_andn2_b32 s1, s1, s0
; GFX1032-NEXT:    s_and_b32 exec_lo, exec_lo, s1
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    exp mrt0 off, off, off, off
; GFX1032-NEXT:    s_endpgm
; GFX1032-NEXT:  ; %bb.1:
; GFX1032-NEXT:    s_mov_b32 exec_lo, 0
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_kill_i1_terminator_i1:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    v_cmp_lt_i32_e32 vcc, v0, v1
; GFX1064-NEXT:    v_cmp_lt_i32_e64 s[0:1], v2, v3
; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
; GFX1064-NEXT:    s_or_b64 s[0:1], vcc, s[0:1]
; GFX1064-NEXT:    s_xor_b64 s[0:1], s[0:1], exec
; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[0:1]
; GFX1064-NEXT:    s_and_b64 exec, exec, s[2:3]
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    exp mrt0 off, off, off, off
; GFX1064-NEXT:    s_endpgm
; GFX1064-NEXT:  ; %bb.1:
; GFX1064-NEXT:    s_mov_b64 exec, 0
; GFX1064-NEXT:    s_endpgm
  %c1 = icmp slt i32 %a, %b
  %c2 = icmp slt i32 %c, %d
  %x = or i1 %c1, %c2
  call void @llvm.amdgcn.kill(i1 %x)
  call void @llvm.amdgcn.exp.f32(i32 0, i32 0, float 0.0, float 0.0, float 0.0, float 0.0, i1 false, i1 false)
  ret void
}

define amdgpu_ps <4 x float> @test_loop_vcc(<4 x float> %in) #0 {
; GFX1032-LABEL: test_loop_vcc:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    s_mov_b32 s0, exec_lo
; GFX1032-NEXT:    s_wqm_b32 exec_lo, exec_lo
; GFX1032-NEXT:    v_mov_b32_e32 v7, v3
; GFX1032-NEXT:    v_mov_b32_e32 v6, v2
; GFX1032-NEXT:    v_mov_b32_e32 v5, v1
; GFX1032-NEXT:    v_mov_b32_e32 v4, v0
; GFX1032-NEXT:    v_mov_b32_e32 v8, 0
; GFX1032-NEXT:    s_branch .LBB33_2
; GFX1032-NEXT:  .LBB33_1: ; %body
; GFX1032-NEXT:    ; in Loop: Header=BB33_2 Depth=1
; GFX1032-NEXT:    image_sample v[4:7], v0, s[0:7], s[0:3] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX1032-NEXT:    v_add_f32_e32 v8, 2.0, v8
; GFX1032-NEXT:    s_cbranch_execz .LBB33_4
; GFX1032-NEXT:  .LBB33_2: ; %loop
; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
; GFX1032-NEXT:    v_cmp_lt_f32_e32 vcc_lo, 0x40e00000, v8
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_mov_b32_e32 v0, v4
; GFX1032-NEXT:    v_mov_b32_e32 v1, v5
; GFX1032-NEXT:    v_mov_b32_e32 v2, v6
; GFX1032-NEXT:    v_mov_b32_e32 v3, v7
; GFX1032-NEXT:    s_cbranch_vccz .LBB33_1
; GFX1032-NEXT:  ; %bb.3:
; GFX1032-NEXT:    ; implicit-def: $vgpr4_vgpr5_vgpr6_vgpr7
; GFX1032-NEXT:    ; implicit-def: $vgpr8
; GFX1032-NEXT:  .LBB33_4: ; %break
; GFX1032-NEXT:    s_and_b32 exec_lo, exec_lo, s0
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    ; return to shader part epilog
;
; GFX1064-LABEL: test_loop_vcc:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    s_mov_b64 s[0:1], exec
; GFX1064-NEXT:    s_wqm_b64 exec, exec
; GFX1064-NEXT:    v_mov_b32_e32 v7, v3
; GFX1064-NEXT:    v_mov_b32_e32 v6, v2
; GFX1064-NEXT:    v_mov_b32_e32 v5, v1
; GFX1064-NEXT:    v_mov_b32_e32 v4, v0
; GFX1064-NEXT:    v_mov_b32_e32 v8, 0
; GFX1064-NEXT:    s_branch .LBB33_2
; GFX1064-NEXT:  .LBB33_1: ; %body
; GFX1064-NEXT:    ; in Loop: Header=BB33_2 Depth=1
; GFX1064-NEXT:    image_sample v[4:7], v0, s[0:7], s[0:3] dmask:0xf dim:SQ_RSRC_IMG_1D
; GFX1064-NEXT:    v_add_f32_e32 v8, 2.0, v8
; GFX1064-NEXT:    s_cbranch_execz .LBB33_4
; GFX1064-NEXT:  .LBB33_2: ; %loop
; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT:    v_cmp_lt_f32_e32 vcc, 0x40e00000, v8
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_mov_b32_e32 v0, v4
; GFX1064-NEXT:    v_mov_b32_e32 v1, v5
; GFX1064-NEXT:    v_mov_b32_e32 v2, v6
; GFX1064-NEXT:    v_mov_b32_e32 v3, v7
; GFX1064-NEXT:    s_cbranch_vccz .LBB33_1
; GFX1064-NEXT:  ; %bb.3:
; GFX1064-NEXT:    ; implicit-def: $vgpr4_vgpr5_vgpr6_vgpr7
; GFX1064-NEXT:    ; implicit-def: $vgpr8
; GFX1064-NEXT:  .LBB33_4: ; %break
; GFX1064-NEXT:    s_and_b64 exec, exec, s[0:1]
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    ; return to shader part epilog
entry:
  br label %loop

loop:
  %ctr.iv = phi float [ 0.0, %entry ], [ %ctr.next, %body ]
  %c.iv = phi <4 x float> [ %in, %entry ], [ %c.next, %body ]
  %cc = fcmp ogt float %ctr.iv, 7.0
  br i1 %cc, label %break, label %body

body:
  %c.iv0 = extractelement <4 x float> %c.iv, i32 0
  %c.next = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32 15, float %c.iv0, <8 x i32> undef, <4 x i32> undef, i1 0, i32 0, i32 0)
  %ctr.next = fadd float %ctr.iv, 2.0
  br label %loop

break:
  ret <4 x float> %c.iv
}

define amdgpu_ps float @test_wwm1(i32 inreg %idx0, i32 inreg %idx1, float %src0, float %src1) {
; GFX1032-LABEL: test_wwm1:
; GFX1032:       ; %bb.0: ; %main_body
; GFX1032-NEXT:    s_or_saveexec_b32 s0, -1
; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
; GFX1032-NEXT:    v_mov_b32_e32 v3, v0
; GFX1032-NEXT:    v_add_f32_e32 v2, v3, v2
; GFX1032-NEXT:    s_mov_b32 exec_lo, s0
; GFX1032-NEXT:    v_mov_b32_e32 v0, v2
; GFX1032-NEXT:    ; return to shader part epilog
;
; GFX1064-LABEL: test_wwm1:
; GFX1064:       ; %bb.0: ; %main_body
; GFX1064-NEXT:    s_or_saveexec_b64 s[0:1], -1
; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
; GFX1064-NEXT:    v_mov_b32_e32 v3, v0
; GFX1064-NEXT:    v_add_f32_e32 v2, v3, v2
; GFX1064-NEXT:    s_mov_b64 exec, s[0:1]
; GFX1064-NEXT:    v_mov_b32_e32 v0, v2
; GFX1064-NEXT:    ; return to shader part epilog
main_body:
  %out = fadd float %src0, %src1
  %out.0 = call float @llvm.amdgcn.wwm.f32(float %out)
  ret float %out.0
}

define amdgpu_ps float @test_wwm2(i32 inreg %idx) {
; GFX1032-LABEL: test_wwm2:
; GFX1032:       ; %bb.0: ; %main_body
; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, -1, 0
; GFX1032-NEXT:    v_mbcnt_hi_u32_b32 v0, -1, v0
; GFX1032-NEXT:    v_cmp_gt_u32_e32 vcc_lo, 16, v0
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    s_and_saveexec_b32 s1, vcc_lo
; GFX1032-NEXT:    s_cbranch_execz .LBB35_2
; GFX1032-NEXT:  ; %bb.1: ; %if
; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
; GFX1032-NEXT:    v_mov_b32_e32 v1, s0
; GFX1032-NEXT:    buffer_load_dword v1, v1, s[0:3], 0 idxen
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_add_f32_e32 v2, v1, v1
; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
; GFX1032-NEXT:    v_mov_b32_e32 v0, v2
; GFX1032-NEXT:    v_add_f32_e32 v0, v1, v0
; GFX1032-NEXT:  .LBB35_2: ; %endif
; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s1
; GFX1032-NEXT:    ; return to shader part epilog
;
; GFX1064-LABEL: test_wwm2:
; GFX1064:       ; %bb.0: ; %main_body
; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, -1, 0
; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, -1, v0
; GFX1064-NEXT:    v_cmp_gt_u32_e32 vcc, 16, v0
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
; GFX1064-NEXT:    s_cbranch_execz .LBB35_2
; GFX1064-NEXT:  ; %bb.1: ; %if
; GFX1064-NEXT:    s_or_saveexec_b64 s[4:5], -1
; GFX1064-NEXT:    v_mov_b32_e32 v1, s0
; GFX1064-NEXT:    buffer_load_dword v1, v1, s[0:3], 0 idxen
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_add_f32_e32 v2, v1, v1
; GFX1064-NEXT:    s_mov_b64 exec, s[4:5]
; GFX1064-NEXT:    v_mov_b32_e32 v0, v2
; GFX1064-NEXT:    v_add_f32_e32 v0, v1, v0
; GFX1064-NEXT:  .LBB35_2: ; %endif
; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
; GFX1064-NEXT:    ; return to shader part epilog
main_body:
  ; use mbcnt to make sure the branch is divergent
  %lo = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %hi = call i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %lo)
  %cc = icmp uge i32 %hi, 16
  br i1 %cc, label %endif, label %if

if:
  %src = call float @llvm.amdgcn.struct.ptr.buffer.load.f32(ptr addrspace(8) undef, i32 %idx, i32 0, i32 0, i32 0)
  %out = fadd float %src, %src
  %out.0 = call float @llvm.amdgcn.wwm.f32(float %out)
  %out.1 = fadd float %src, %out.0
  br label %endif

endif:
  %out.2 = phi float [ %out.1, %if ], [ 0.0, %main_body ]
  ret float %out.2
}

define amdgpu_ps float @test_strict_wwm1(i32 inreg %idx0, i32 inreg %idx1, float %src0, float %src1) {
; GFX1032-LABEL: test_strict_wwm1:
; GFX1032:       ; %bb.0: ; %main_body
; GFX1032-NEXT:    s_or_saveexec_b32 s0, -1
; GFX1032-NEXT:    v_mov_b32_e32 v2, v1
; GFX1032-NEXT:    v_mov_b32_e32 v3, v0
; GFX1032-NEXT:    v_add_f32_e32 v2, v3, v2
; GFX1032-NEXT:    s_mov_b32 exec_lo, s0
; GFX1032-NEXT:    v_mov_b32_e32 v0, v2
; GFX1032-NEXT:    ; return to shader part epilog
;
; GFX1064-LABEL: test_strict_wwm1:
; GFX1064:       ; %bb.0: ; %main_body
; GFX1064-NEXT:    s_or_saveexec_b64 s[0:1], -1
; GFX1064-NEXT:    v_mov_b32_e32 v2, v1
; GFX1064-NEXT:    v_mov_b32_e32 v3, v0
; GFX1064-NEXT:    v_add_f32_e32 v2, v3, v2
; GFX1064-NEXT:    s_mov_b64 exec, s[0:1]
; GFX1064-NEXT:    v_mov_b32_e32 v0, v2
; GFX1064-NEXT:    ; return to shader part epilog
main_body:
  %out = fadd float %src0, %src1
  %out.0 = call float @llvm.amdgcn.strict.wwm.f32(float %out)
  ret float %out.0
}

define amdgpu_ps float @test_strict_wwm2(i32 inreg %idx) {
; GFX1032-LABEL: test_strict_wwm2:
; GFX1032:       ; %bb.0: ; %main_body
; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, -1, 0
; GFX1032-NEXT:    v_mbcnt_hi_u32_b32 v0, -1, v0
; GFX1032-NEXT:    v_cmp_gt_u32_e32 vcc_lo, 16, v0
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    s_and_saveexec_b32 s1, vcc_lo
; GFX1032-NEXT:    s_cbranch_execz .LBB37_2
; GFX1032-NEXT:  ; %bb.1: ; %if
; GFX1032-NEXT:    s_or_saveexec_b32 s2, -1
; GFX1032-NEXT:    v_mov_b32_e32 v1, s0
; GFX1032-NEXT:    buffer_load_dword v1, v1, s[0:3], 0 idxen
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_add_f32_e32 v2, v1, v1
; GFX1032-NEXT:    s_mov_b32 exec_lo, s2
; GFX1032-NEXT:    v_mov_b32_e32 v0, v2
; GFX1032-NEXT:    v_add_f32_e32 v0, v1, v0
; GFX1032-NEXT:  .LBB37_2: ; %endif
; GFX1032-NEXT:    s_or_b32 exec_lo, exec_lo, s1
; GFX1032-NEXT:    ; return to shader part epilog
;
; GFX1064-LABEL: test_strict_wwm2:
; GFX1064:       ; %bb.0: ; %main_body
; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, -1, 0
; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, -1, v0
; GFX1064-NEXT:    v_cmp_gt_u32_e32 vcc, 16, v0
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
; GFX1064-NEXT:    s_cbranch_execz .LBB37_2
; GFX1064-NEXT:  ; %bb.1: ; %if
; GFX1064-NEXT:    s_or_saveexec_b64 s[4:5], -1
; GFX1064-NEXT:    v_mov_b32_e32 v1, s0
; GFX1064-NEXT:    buffer_load_dword v1, v1, s[0:3], 0 idxen
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_add_f32_e32 v2, v1, v1
; GFX1064-NEXT:    s_mov_b64 exec, s[4:5]
; GFX1064-NEXT:    v_mov_b32_e32 v0, v2
; GFX1064-NEXT:    v_add_f32_e32 v0, v1, v0
; GFX1064-NEXT:  .LBB37_2: ; %endif
; GFX1064-NEXT:    s_or_b64 exec, exec, s[2:3]
; GFX1064-NEXT:    ; return to shader part epilog
main_body:
  ; use mbcnt to make sure the branch is divergent
  %lo = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
  %hi = call i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %lo)
  %cc = icmp uge i32 %hi, 16
  br i1 %cc, label %endif, label %if

if:
  %src = call float @llvm.amdgcn.struct.ptr.buffer.load.f32(ptr addrspace(8) undef, i32 %idx, i32 0, i32 0, i32 0)
  %out = fadd float %src, %src
  %out.0 = call float @llvm.amdgcn.strict.wwm.f32(float %out)
  %out.1 = fadd float %src, %out.0
  br label %endif

endif:
  %out.2 = phi float [ %out.1, %if ], [ 0.0, %main_body ]
  ret float %out.2
}


define amdgpu_ps <4 x float> @test_wqm1(i32 inreg, i32 inreg, i32 inreg, i32 inreg %m0, <8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, <2 x float> %pos) #0 {
; GFX1032-LABEL: test_wqm1:
; GFX1032:       ; %bb.0: ; %main_body
; GFX1032-NEXT:    s_mov_b32 s0, exec_lo
; GFX1032-NEXT:    s_wqm_b32 exec_lo, exec_lo
; GFX1032-NEXT:    s_mov_b32 m0, s3
; GFX1032-NEXT:    v_interp_p1_f32_e32 v2, v0, attr0.x
; GFX1032-NEXT:    v_interp_p1_f32_e32 v3, v0, attr0.y
; GFX1032-NEXT:    v_interp_p2_f32_e32 v2, v1, attr0.x
; GFX1032-NEXT:    v_interp_p2_f32_e32 v3, v1, attr0.y
; GFX1032-NEXT:    s_and_b32 exec_lo, exec_lo, s0
; GFX1032-NEXT:    image_sample v[0:3], v[2:3], s[4:11], s[12:15] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    ; return to shader part epilog
;
; GFX1064-LABEL: test_wqm1:
; GFX1064:       ; %bb.0: ; %main_body
; GFX1064-NEXT:    s_mov_b64 s[0:1], exec
; GFX1064-NEXT:    s_wqm_b64 exec, exec
; GFX1064-NEXT:    s_mov_b32 m0, s3
; GFX1064-NEXT:    v_interp_p1_f32_e32 v2, v0, attr0.x
; GFX1064-NEXT:    v_interp_p1_f32_e32 v3, v0, attr0.y
; GFX1064-NEXT:    v_interp_p2_f32_e32 v2, v1, attr0.x
; GFX1064-NEXT:    v_interp_p2_f32_e32 v3, v1, attr0.y
; GFX1064-NEXT:    s_and_b64 exec, exec, s[0:1]
; GFX1064-NEXT:    image_sample v[0:3], v[2:3], s[4:11], s[12:15] dmask:0xf dim:SQ_RSRC_IMG_2D
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    ; return to shader part epilog
main_body:
  %inst23 = extractelement <2 x float> %pos, i32 0
  %inst24 = extractelement <2 x float> %pos, i32 1
  %inst25 = tail call float @llvm.amdgcn.interp.p1(float %inst23, i32 0, i32 0, i32 %m0)
  %inst26 = tail call float @llvm.amdgcn.interp.p2(float %inst25, float %inst24, i32 0, i32 0, i32 %m0)
  %inst28 = tail call float @llvm.amdgcn.interp.p1(float %inst23, i32 1, i32 0, i32 %m0)
  %inst29 = tail call float @llvm.amdgcn.interp.p2(float %inst28, float %inst24, i32 1, i32 0, i32 %m0)
  %tex = call <4 x float> @llvm.amdgcn.image.sample.2d.v4f32.f32(i32 15, float %inst26, float %inst29, <8 x i32> %rsrc, <4 x i32> %sampler, i1 0, i32 0, i32 0)
  ret <4 x float> %tex
}

define amdgpu_ps float @test_wqm2(i32 inreg %idx0, i32 inreg %idx1) #0 {
; GFX1032-LABEL: test_wqm2:
; GFX1032:       ; %bb.0: ; %main_body
; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
; GFX1032-NEXT:    s_wqm_b32 exec_lo, exec_lo
; GFX1032-NEXT:    v_mov_b32_e32 v0, s0
; GFX1032-NEXT:    v_mov_b32_e32 v1, s1
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    buffer_load_dword v2, v0, s[0:3], 0 idxen
; GFX1032-NEXT:    buffer_load_dword v3, v1, s[0:3], 0 idxen
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    v_add_f32_e32 v0, v2, v3
; GFX1032-NEXT:    ; kill: def $vgpr0 killed $vgpr0 killed $exec
; GFX1032-NEXT:    s_and_b32 exec_lo, exec_lo, s2
; GFX1032-NEXT:    ; return to shader part epilog
;
; GFX1064-LABEL: test_wqm2:
; GFX1064:       ; %bb.0: ; %main_body
; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
; GFX1064-NEXT:    s_wqm_b64 exec, exec
; GFX1064-NEXT:    v_mov_b32_e32 v0, s0
; GFX1064-NEXT:    v_mov_b32_e32 v1, s1
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    buffer_load_dword v2, v0, s[0:3], 0 idxen
; GFX1064-NEXT:    buffer_load_dword v3, v1, s[0:3], 0 idxen
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    v_add_f32_e32 v0, v2, v3
; GFX1064-NEXT:    ; kill: def $vgpr0 killed $vgpr0 killed $exec
; GFX1064-NEXT:    s_and_b64 exec, exec, s[2:3]
; GFX1064-NEXT:    ; return to shader part epilog
main_body:
  %src0 = call float @llvm.amdgcn.struct.ptr.buffer.load.f32(ptr addrspace(8) undef, i32 %idx0, i32 0, i32 0, i32 0)
  %src1 = call float @llvm.amdgcn.struct.ptr.buffer.load.f32(ptr addrspace(8) undef, i32 %idx1, i32 0, i32 0, i32 0)
  %out = fadd float %src0, %src1
  %out.0 = bitcast float %out to i32
  %out.1 = call i32 @llvm.amdgcn.wqm.i32(i32 %out.0)
  %out.2 = bitcast i32 %out.1 to float
  ret float %out.2
}

define amdgpu_kernel void @test_intr_fcmp_i64(ptr addrspace(1) %out, float %src, float %a) {
; GFX1032-LABEL: test_intr_fcmp_i64:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_cmp_eq_f32_e64 s2, s2, |s3|
; GFX1032-NEXT:    v_mov_b32_e32 v0, s2
; GFX1032-NEXT:    global_store_dwordx2 v1, v[0:1], s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_intr_fcmp_i64:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_cmp_eq_f32_e64 s[2:3], s2, |s3|
; GFX1064-NEXT:    v_mov_b32_e32 v0, s2
; GFX1064-NEXT:    v_mov_b32_e32 v1, s3
; GFX1064-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1064-NEXT:    s_endpgm
  %temp = call float @llvm.fabs.f32(float %a)
  %result = call i64 @llvm.amdgcn.fcmp.i64.f32(float %src, float %temp, i32 1)
  store i64 %result, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @test_intr_icmp_i64(ptr addrspace(1) %out, i32 %src) {
; GFX1032-LABEL: test_intr_icmp_i64:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v1, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_cmp_eq_u32_e64 s0, 0x64, s4
; GFX1032-NEXT:    v_mov_b32_e32 v0, s0
; GFX1032-NEXT:    global_store_dwordx2 v1, v[0:1], s[2:3]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_intr_icmp_i64:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_cmp_eq_u32_e64 s[0:1], 0x64, s4
; GFX1064-NEXT:    v_mov_b32_e32 v0, s0
; GFX1064-NEXT:    v_mov_b32_e32 v1, s1
; GFX1064-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
; GFX1064-NEXT:    s_endpgm
  %result = call i64 @llvm.amdgcn.icmp.i64.i32(i32 %src, i32 100, i32 32)
  store i64 %result, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @test_intr_fcmp_i32(ptr addrspace(1) %out, float %src, float %a) {
; GFX1032-LABEL: test_intr_fcmp_i32:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_cmp_eq_f32_e64 s2, s2, |s3|
; GFX1032-NEXT:    v_mov_b32_e32 v1, s2
; GFX1032-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_intr_fcmp_i32:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_cmp_eq_f32_e64 s[2:3], s2, |s3|
; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
; GFX1064-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX1064-NEXT:    s_endpgm
  %temp = call float @llvm.fabs.f32(float %a)
  %result = call i32 @llvm.amdgcn.fcmp.i32.f32(float %src, float %temp, i32 1)
  store i32 %result, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @test_intr_icmp_i32(ptr addrspace(1) %out, i32 %src) {
; GFX1032-LABEL: test_intr_icmp_i32:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_clause 0x1
; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_cmp_eq_u32_e64 s0, 0x64, s4
; GFX1032-NEXT:    v_mov_b32_e32 v1, s0
; GFX1032-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_intr_icmp_i32:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_clause 0x1
; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x2c
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_cmp_eq_u32_e64 s[0:1], 0x64, s4
; GFX1064-NEXT:    v_mov_b32_e32 v1, s0
; GFX1064-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX1064-NEXT:    s_endpgm
  %result = call i32 @llvm.amdgcn.icmp.i32.i32(i32 %src, i32 100, i32 32)
  store i32 %result, ptr addrspace(1) %out
  ret void
}

define amdgpu_ps void @test_wqm_vote(float %a) {
; GFX1032-LABEL: test_wqm_vote:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    v_cmp_neq_f32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    s_mov_b32 s0, exec_lo
; GFX1032-NEXT:    v_mov_b32_e32 v0, 0
; GFX1032-NEXT:    s_wqm_b32 s1, vcc_lo
; GFX1032-NEXT:    s_xor_b32 s1, s1, exec_lo
; GFX1032-NEXT:    s_andn2_b32 s0, s0, s1
; GFX1032-NEXT:    s_cbranch_scc0 .LBB44_2
; GFX1032-NEXT:  ; %bb.1:
; GFX1032-NEXT:    s_and_b32 exec_lo, exec_lo, s0
; GFX1032-NEXT:    exp mrt0 off, off, off, off
; GFX1032-NEXT:    s_endpgm
; GFX1032-NEXT:  .LBB44_2:
; GFX1032-NEXT:    s_mov_b32 exec_lo, 0
; GFX1032-NEXT:    exp null off, off, off, off done vm
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_wqm_vote:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    v_cmp_neq_f32_e32 vcc, 0, v0
; GFX1064-NEXT:    s_mov_b64 s[0:1], exec
; GFX1064-NEXT:    v_mov_b32_e32 v0, 0
; GFX1064-NEXT:    s_wqm_b64 s[2:3], vcc
; GFX1064-NEXT:    s_xor_b64 s[2:3], s[2:3], exec
; GFX1064-NEXT:    s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX1064-NEXT:    s_cbranch_scc0 .LBB44_2
; GFX1064-NEXT:  ; %bb.1:
; GFX1064-NEXT:    s_and_b64 exec, exec, s[0:1]
; GFX1064-NEXT:    exp mrt0 off, off, off, off
; GFX1064-NEXT:    s_endpgm
; GFX1064-NEXT:  .LBB44_2:
; GFX1064-NEXT:    s_mov_b64 exec, 0
; GFX1064-NEXT:    exp null off, off, off, off done vm
; GFX1064-NEXT:    s_endpgm
  %c1 = fcmp une float %a, 0.0
  %c2 = call i1 @llvm.amdgcn.wqm.vote(i1 %c1)
  call void @llvm.amdgcn.kill(i1 %c2)
  call void @llvm.amdgcn.exp.f32(i32 0, i32 0, float 0.0, float 0.0, float 0.0, float 0.0, i1 false, i1 false)
  ret void
}

define amdgpu_kernel void @test_branch_true() #2 {
; GFX1032-LABEL: test_branch_true:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    s_mov_b32 vcc_lo, exec_lo
; GFX1032-NEXT:    s_cbranch_execnz .LBB45_2
; GFX1032-NEXT:  ; %bb.1: ; %for.body.lr.ph
; GFX1032-NEXT:    s_branch .LBB45_3
; GFX1032-NEXT:  .LBB45_2: ; %Flow
; GFX1032-NEXT:    s_branch .LBB45_5
; GFX1032-NEXT:  .LBB45_3: ; %for.body
; GFX1032-NEXT:    s_mov_b32 vcc_lo, 0
; GFX1032-NEXT:  ; %bb.4: ; %for.end.loopexit
; GFX1032-NEXT:    s_branch .LBB45_2
; GFX1032-NEXT:  .LBB45_5: ; %for.end
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_branch_true:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    s_mov_b64 vcc, exec
; GFX1064-NEXT:    s_cbranch_execnz .LBB45_2
; GFX1064-NEXT:  ; %bb.1: ; %for.body.lr.ph
; GFX1064-NEXT:    s_branch .LBB45_3
; GFX1064-NEXT:  .LBB45_2: ; %Flow
; GFX1064-NEXT:    s_branch .LBB45_5
; GFX1064-NEXT:  .LBB45_3: ; %for.body
; GFX1064-NEXT:    s_mov_b64 vcc, 0
; GFX1064-NEXT:  ; %bb.4: ; %for.end.loopexit
; GFX1064-NEXT:    s_branch .LBB45_2
; GFX1064-NEXT:  .LBB45_5: ; %for.end
; GFX1064-NEXT:    s_endpgm
entry:
  br i1 true, label %for.end, label %for.body.lr.ph

for.body.lr.ph:                                   ; preds = %entry
  br label %for.body

for.body:                                         ; preds = %for.body, %for.body.lr.ph
  br i1 undef, label %for.end, label %for.body

for.end:                                          ; preds = %for.body, %entry
  ret void
}

define amdgpu_ps float @test_ps_live() #0 {
; GFX1032-LABEL: test_ps_live:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_mov_b32 s0, exec_lo
; GFX1032-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s0
; GFX1032-NEXT:    ; return to shader part epilog
;
; GFX1064-LABEL: test_ps_live:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_mov_b64 s[0:1], exec
; GFX1064-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[0:1]
; GFX1064-NEXT:    ; return to shader part epilog
  %live = call i1 @llvm.amdgcn.ps.live()
  %live.32 = zext i1 %live to i32
  %r = bitcast i32 %live.32 to float
  ret float %r
}

define amdgpu_kernel void @test_vccnz_ifcvt_triangle64(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
; GFX1032-LABEL: test_vccnz_ifcvt_triangle64:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_cmp_neq_f64_e64 s4, s[2:3], 1.0
; GFX1032-NEXT:    s_and_b32 vcc_lo, exec_lo, s4
; GFX1032-NEXT:    s_cbranch_vccnz .LBB47_2
; GFX1032-NEXT:  ; %bb.1: ; %if
; GFX1032-NEXT:    v_add_f64 v[0:1], s[2:3], s[2:3]
; GFX1032-NEXT:    s_branch .LBB47_3
; GFX1032-NEXT:  .LBB47_2:
; GFX1032-NEXT:    v_mov_b32_e32 v0, s2
; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
; GFX1032-NEXT:  .LBB47_3: ; %endif
; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
; GFX1032-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: test_vccnz_ifcvt_triangle64:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_cmp_neq_f64_e64 s[4:5], s[2:3], 1.0
; GFX1064-NEXT:    s_and_b64 vcc, exec, s[4:5]
; GFX1064-NEXT:    s_cbranch_vccnz .LBB47_2
; GFX1064-NEXT:  ; %bb.1: ; %if
; GFX1064-NEXT:    v_add_f64 v[0:1], s[2:3], s[2:3]
; GFX1064-NEXT:    s_branch .LBB47_3
; GFX1064-NEXT:  .LBB47_2:
; GFX1064-NEXT:    v_mov_b32_e32 v0, s2
; GFX1064-NEXT:    v_mov_b32_e32 v1, s3
; GFX1064-NEXT:  .LBB47_3: ; %endif
; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
; GFX1064-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
; GFX1064-NEXT:    s_endpgm
entry:
  %v = load double, ptr addrspace(1) %in
  %cc = fcmp oeq double %v, 1.000000e+00
  br i1 %cc, label %if, label %endif

if:
  %u = fadd double %v, %v
  br label %endif

endif:
  %r = phi double [ %v, %entry ], [ %u, %if ]
  store double %r, ptr addrspace(1) %out
  ret void
}

define amdgpu_gs float @test_vgprblocks_w32_attr(float %a, float %b, float %c, float %d, float %e,
; GCN-LABEL: test_vgprblocks_w32_attr:
; GCN:       ; %bb.0: ; %main_body
; GCN-NEXT:    v_add_f32_e32 v0, v0, v1
; GCN-NEXT:    v_add_f32_e32 v0, v0, v2
; GCN-NEXT:    v_add_f32_e32 v0, v0, v3
; GCN-NEXT:    v_add_f32_e32 v0, v0, v4
; GCN-NEXT:    v_add_f32_e32 v0, v0, v5
; GCN-NEXT:    v_add_f32_e32 v0, v0, v6
; GCN-NEXT:    v_add_f32_e32 v0, v0, v7
; GCN-NEXT:    v_add_f32_e32 v0, v0, v8
; GCN-NEXT:    v_add_f32_e32 v0, v0, v9
; GCN-NEXT:    v_add_f32_e32 v0, v0, v10
; GCN-NEXT:    v_add_f32_e32 v0, v0, v11
; GCN-NEXT:    ; return to shader part epilog
                                        float %f, float %g, float %h, float %i, float %j, float %k, float %l) #3 {
main_body:
  %s = fadd float %a, %b
  %s.1 = fadd float %s, %c
  %s.2 = fadd float %s.1, %d
  %s.3 = fadd float %s.2, %e
  %s.4 = fadd float %s.3, %f
  %s.5 = fadd float %s.4, %g
  %s.6 = fadd float %s.5, %h
  %s.7 = fadd float %s.6, %i
  %s.8 = fadd float %s.7, %j
  %s.9 = fadd float %s.8, %k
  %s.10 = fadd float %s.9, %l
  ret float %s.10
}

define amdgpu_gs float @test_vgprblocks_w64_attr(float %a, float %b, float %c, float %d, float %e,
; GCN-LABEL: test_vgprblocks_w64_attr:
; GCN:       ; %bb.0: ; %main_body
; GCN-NEXT:    v_add_f32_e32 v0, v0, v1
; GCN-NEXT:    v_add_f32_e32 v0, v0, v2
; GCN-NEXT:    v_add_f32_e32 v0, v0, v3
; GCN-NEXT:    v_add_f32_e32 v0, v0, v4
; GCN-NEXT:    v_add_f32_e32 v0, v0, v5
; GCN-NEXT:    v_add_f32_e32 v0, v0, v6
; GCN-NEXT:    v_add_f32_e32 v0, v0, v7
; GCN-NEXT:    v_add_f32_e32 v0, v0, v8
; GCN-NEXT:    v_add_f32_e32 v0, v0, v9
; GCN-NEXT:    v_add_f32_e32 v0, v0, v10
; GCN-NEXT:    v_add_f32_e32 v0, v0, v11
; GCN-NEXT:    ; return to shader part epilog
                                        float %f, float %g, float %h, float %i, float %j, float %k, float %l) #4 {
main_body:
  %s = fadd float %a, %b
  %s.1 = fadd float %s, %c
  %s.2 = fadd float %s.1, %d
  %s.3 = fadd float %s.2, %e
  %s.4 = fadd float %s.3, %f
  %s.5 = fadd float %s.4, %g
  %s.6 = fadd float %s.5, %h
  %s.7 = fadd float %s.6, %i
  %s.8 = fadd float %s.7, %j
  %s.9 = fadd float %s.8, %k
  %s.10 = fadd float %s.9, %l
  ret float %s.10
}

define amdgpu_kernel void @icmp64(i32 %n, i32 %s) {
; GFX1032-LABEL: icmp64:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    s_load_dword s0, s[0:1], 0x28
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_cvt_f32_u32_e32 v1, s0
; GFX1032-NEXT:    s_sub_i32 s1, 0, s0
; GFX1032-NEXT:    v_rcp_iflag_f32_e32 v1, v1
; GFX1032-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
; GFX1032-NEXT:    v_cvt_u32_f32_e32 v1, v1
; GFX1032-NEXT:    v_mul_lo_u32 v2, s1, v1
; GFX1032-NEXT:    s_brev_b32 s1, 1
; GFX1032-NEXT:    v_mul_hi_u32 v2, v1, v2
; GFX1032-NEXT:    v_add_nc_u32_e32 v1, v1, v2
; GFX1032-NEXT:    v_mul_hi_u32 v1, v0, v1
; GFX1032-NEXT:    v_mul_lo_u32 v1, v1, s0
; GFX1032-NEXT:    v_sub_nc_u32_e32 v0, v0, v1
; GFX1032-NEXT:    v_subrev_nc_u32_e32 v1, s0, v0
; GFX1032-NEXT:    v_cmp_le_u32_e32 vcc_lo, s0, v0
; GFX1032-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX1032-NEXT:    v_subrev_nc_u32_e32 v1, s0, v0
; GFX1032-NEXT:    v_cmp_le_u32_e32 vcc_lo, s0, v0
; GFX1032-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    s_lshr_b32 s0, vcc_lo, 1
; GFX1032-NEXT:    s_ff1_i32_b64 s0, s[0:1]
; GFX1032-NEXT:    s_cmp_gt_u32 s0, 9
; GFX1032-NEXT:    s_cselect_b32 s0, -1, 0
; GFX1032-NEXT:    s_and_b32 s0, vcc_lo, s0
; GFX1032-NEXT:    s_and_saveexec_b32 s1, s0
; GFX1032-NEXT:  ; %bb.1: ; %if.then
; GFX1032-NEXT:    ; divergent unreachable
; GFX1032-NEXT:  ; %bb.2: ; %UnifiedReturnBlock
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: icmp64:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    s_load_dword s0, s[0:1], 0x28
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_cvt_f32_u32_e32 v1, s0
; GFX1064-NEXT:    s_sub_i32 s1, 0, s0
; GFX1064-NEXT:    v_rcp_iflag_f32_e32 v1, v1
; GFX1064-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
; GFX1064-NEXT:    v_cvt_u32_f32_e32 v1, v1
; GFX1064-NEXT:    v_mul_lo_u32 v2, s1, v1
; GFX1064-NEXT:    v_mul_hi_u32 v2, v1, v2
; GFX1064-NEXT:    v_add_nc_u32_e32 v1, v1, v2
; GFX1064-NEXT:    v_mul_hi_u32 v1, v0, v1
; GFX1064-NEXT:    v_mul_lo_u32 v1, v1, s0
; GFX1064-NEXT:    v_sub_nc_u32_e32 v0, v0, v1
; GFX1064-NEXT:    v_subrev_nc_u32_e32 v1, s0, v0
; GFX1064-NEXT:    v_cmp_le_u32_e32 vcc, s0, v0
; GFX1064-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX1064-NEXT:    v_subrev_nc_u32_e32 v1, s0, v0
; GFX1064-NEXT:    v_cmp_le_u32_e32 vcc, s0, v0
; GFX1064-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT:    s_lshr_b64 s[0:1], vcc, 1
; GFX1064-NEXT:    s_bitset1_b32 s1, 31
; GFX1064-NEXT:    s_ff1_i32_b64 s0, s[0:1]
; GFX1064-NEXT:    s_cmp_gt_u32 s0, 9
; GFX1064-NEXT:    s_cselect_b64 s[0:1], -1, 0
; GFX1064-NEXT:    s_and_b64 s[0:1], vcc, s[0:1]
; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], s[0:1]
; GFX1064-NEXT:  ; %bb.1: ; %if.then
; GFX1064-NEXT:    ; divergent unreachable
; GFX1064-NEXT:  ; %bb.2: ; %UnifiedReturnBlock
; GFX1064-NEXT:    s_endpgm
entry:
  %id = tail call i32 @llvm.amdgcn.workitem.id.x()
  %mul4 = mul nsw i32 %s, %n
  %cmp = icmp slt i32 0, %mul4
  br label %if.end

if.end:                                           ; preds = %entry
  %rem = urem i32 %id, %s
  %icmp = tail call i64 @llvm.amdgcn.icmp.i64.i32(i32 %rem, i32 0, i32 32)
  %shr = lshr i64 %icmp, 1
  %notmask = shl nsw i64 -1, 0
  %and = and i64 %notmask, %shr
  %or = or i64 %and, -9223372036854775808
  %cttz = tail call i64 @llvm.cttz.i64(i64 %or, i1 true)
  %cast = trunc i64 %cttz to i32
  %cmp3 = icmp ugt i32 10, %cast
  %cmp6 = icmp ne i32 %rem, 0
  %brmerge = or i1 %cmp6, %cmp3
  br i1 %brmerge, label %if.end2, label %if.then

if.then:                                          ; preds = %if.end
  unreachable

if.end2:                                          ; preds = %if.end
  ret void
}

define amdgpu_kernel void @fcmp64(float %n, float %s) {
; GFX1032-LABEL: fcmp64:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    s_load_dword s0, s[0:1], 0x28
; GFX1032-NEXT:    v_cvt_f32_u32_e32 v0, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_div_scale_f32 v1, s1, s0, s0, v0
; GFX1032-NEXT:    v_div_scale_f32 v4, vcc_lo, v0, s0, v0
; GFX1032-NEXT:    s_brev_b32 s1, 1
; GFX1032-NEXT:    v_rcp_f32_e32 v2, v1
; GFX1032-NEXT:    v_fma_f32 v3, -v1, v2, 1.0
; GFX1032-NEXT:    v_fmac_f32_e32 v2, v3, v2
; GFX1032-NEXT:    v_mul_f32_e32 v3, v4, v2
; GFX1032-NEXT:    v_fma_f32 v5, -v1, v3, v4
; GFX1032-NEXT:    v_fmac_f32_e32 v3, v5, v2
; GFX1032-NEXT:    v_fma_f32 v1, -v1, v3, v4
; GFX1032-NEXT:    v_div_fmas_f32 v1, v1, v2, v3
; GFX1032-NEXT:    v_div_fixup_f32 v1, v1, s0, v0
; GFX1032-NEXT:    v_trunc_f32_e32 v1, v1
; GFX1032-NEXT:    v_fma_f32 v0, -v1, s0, v0
; GFX1032-NEXT:    v_cmp_eq_f32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    s_lshr_b32 s0, vcc_lo, 1
; GFX1032-NEXT:    v_cmp_nlg_f32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    s_ff1_i32_b64 s0, s[0:1]
; GFX1032-NEXT:    s_cmp_gt_u32 s0, 9
; GFX1032-NEXT:    s_cselect_b32 s0, -1, 0
; GFX1032-NEXT:    s_and_b32 s0, vcc_lo, s0
; GFX1032-NEXT:    s_and_saveexec_b32 s1, s0
; GFX1032-NEXT:  ; %bb.1: ; %if.then
; GFX1032-NEXT:    ; divergent unreachable
; GFX1032-NEXT:  ; %bb.2: ; %UnifiedReturnBlock
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: fcmp64:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x28
; GFX1064-NEXT:    v_cvt_f32_u32_e32 v0, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_div_scale_f32 v1, s[0:1], s2, s2, v0
; GFX1064-NEXT:    v_rcp_f32_e32 v2, v1
; GFX1064-NEXT:    v_fma_f32 v3, -v1, v2, 1.0
; GFX1064-NEXT:    v_fmac_f32_e32 v2, v3, v2
; GFX1064-NEXT:    v_div_scale_f32 v3, vcc, v0, s2, v0
; GFX1064-NEXT:    v_mul_f32_e32 v4, v3, v2
; GFX1064-NEXT:    v_fma_f32 v5, -v1, v4, v3
; GFX1064-NEXT:    v_fmac_f32_e32 v4, v5, v2
; GFX1064-NEXT:    v_fma_f32 v1, -v1, v4, v3
; GFX1064-NEXT:    v_div_fmas_f32 v1, v1, v2, v4
; GFX1064-NEXT:    v_div_fixup_f32 v1, v1, s2, v0
; GFX1064-NEXT:    v_trunc_f32_e32 v1, v1
; GFX1064-NEXT:    v_fma_f32 v0, -v1, s2, v0
; GFX1064-NEXT:    v_cmp_eq_f32_e32 vcc, 0, v0
; GFX1064-NEXT:    s_lshr_b64 s[0:1], vcc, 1
; GFX1064-NEXT:    v_cmp_nlg_f32_e32 vcc, 0, v0
; GFX1064-NEXT:    s_bitset1_b32 s1, 31
; GFX1064-NEXT:    s_ff1_i32_b64 s0, s[0:1]
; GFX1064-NEXT:    s_cmp_gt_u32 s0, 9
; GFX1064-NEXT:    s_cselect_b64 s[0:1], -1, 0
; GFX1064-NEXT:    s_and_b64 s[0:1], vcc, s[0:1]
; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], s[0:1]
; GFX1064-NEXT:  ; %bb.1: ; %if.then
; GFX1064-NEXT:    ; divergent unreachable
; GFX1064-NEXT:  ; %bb.2: ; %UnifiedReturnBlock
; GFX1064-NEXT:    s_endpgm
entry:
  %id = tail call i32 @llvm.amdgcn.workitem.id.x()
  %id.f = uitofp i32 %id to float
  %mul4 = fmul float %s, %n
  %cmp = fcmp ult float 0.0, %mul4
  br label %if.end

if.end:                                           ; preds = %entry
  %rem.f = frem float %id.f, %s
  %fcmp = tail call i64 @llvm.amdgcn.fcmp.i64.f32(float %rem.f, float 0.0, i32 1)
  %shr = lshr i64 %fcmp, 1
  %notmask = shl nsw i64 -1, 0
  %and = and i64 %notmask, %shr
  %or = or i64 %and, -9223372036854775808
  %cttz = tail call i64 @llvm.cttz.i64(i64 %or, i1 true)
  %cast = trunc i64 %cttz to i32
  %cmp3 = icmp ugt i32 10, %cast
  %cmp6 = fcmp one float %rem.f, 0.0
  %brmerge = or i1 %cmp6, %cmp3
  br i1 %brmerge, label %if.end2, label %if.then

if.then:                                          ; preds = %if.end
  unreachable

if.end2:                                          ; preds = %if.end
  ret void
}

define amdgpu_kernel void @icmp32(i32 %n, i32 %s) {
; GFX1032-LABEL: icmp32:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    s_load_dword s0, s[0:1], 0x28
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_cvt_f32_u32_e32 v1, s0
; GFX1032-NEXT:    s_sub_i32 s1, 0, s0
; GFX1032-NEXT:    v_rcp_iflag_f32_e32 v1, v1
; GFX1032-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
; GFX1032-NEXT:    v_cvt_u32_f32_e32 v1, v1
; GFX1032-NEXT:    v_mul_lo_u32 v2, s1, v1
; GFX1032-NEXT:    v_mul_hi_u32 v2, v1, v2
; GFX1032-NEXT:    v_add_nc_u32_e32 v1, v1, v2
; GFX1032-NEXT:    v_mul_hi_u32 v1, v0, v1
; GFX1032-NEXT:    v_mul_lo_u32 v1, v1, s0
; GFX1032-NEXT:    v_sub_nc_u32_e32 v0, v0, v1
; GFX1032-NEXT:    v_subrev_nc_u32_e32 v1, s0, v0
; GFX1032-NEXT:    v_cmp_le_u32_e32 vcc_lo, s0, v0
; GFX1032-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX1032-NEXT:    v_subrev_nc_u32_e32 v1, s0, v0
; GFX1032-NEXT:    v_cmp_le_u32_e32 vcc_lo, s0, v0
; GFX1032-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc_lo
; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    s_lshr_b32 s0, vcc_lo, 1
; GFX1032-NEXT:    s_bitset1_b32 s0, 31
; GFX1032-NEXT:    s_ff1_i32_b32 s0, s0
; GFX1032-NEXT:    s_cmp_gt_u32 s0, 9
; GFX1032-NEXT:    s_cselect_b32 s0, -1, 0
; GFX1032-NEXT:    s_and_b32 s0, vcc_lo, s0
; GFX1032-NEXT:    s_and_saveexec_b32 s1, s0
; GFX1032-NEXT:  ; %bb.1: ; %if.then
; GFX1032-NEXT:    ; divergent unreachable
; GFX1032-NEXT:  ; %bb.2: ; %UnifiedReturnBlock
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: icmp32:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    s_load_dword s0, s[0:1], 0x28
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_cvt_f32_u32_e32 v1, s0
; GFX1064-NEXT:    s_sub_i32 s1, 0, s0
; GFX1064-NEXT:    v_rcp_iflag_f32_e32 v1, v1
; GFX1064-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
; GFX1064-NEXT:    v_cvt_u32_f32_e32 v1, v1
; GFX1064-NEXT:    v_mul_lo_u32 v2, s1, v1
; GFX1064-NEXT:    v_mul_hi_u32 v2, v1, v2
; GFX1064-NEXT:    v_add_nc_u32_e32 v1, v1, v2
; GFX1064-NEXT:    v_mul_hi_u32 v1, v0, v1
; GFX1064-NEXT:    v_mul_lo_u32 v1, v1, s0
; GFX1064-NEXT:    v_sub_nc_u32_e32 v0, v0, v1
; GFX1064-NEXT:    v_subrev_nc_u32_e32 v1, s0, v0
; GFX1064-NEXT:    v_cmp_le_u32_e32 vcc, s0, v0
; GFX1064-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX1064-NEXT:    v_subrev_nc_u32_e32 v1, s0, v0
; GFX1064-NEXT:    v_cmp_le_u32_e32 vcc, s0, v0
; GFX1064-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc
; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
; GFX1064-NEXT:    s_lshr_b32 s0, vcc_lo, 1
; GFX1064-NEXT:    s_bitset1_b32 s0, 31
; GFX1064-NEXT:    s_ff1_i32_b32 s0, s0
; GFX1064-NEXT:    s_cmp_gt_u32 s0, 9
; GFX1064-NEXT:    s_cselect_b64 s[0:1], -1, 0
; GFX1064-NEXT:    s_and_b64 s[0:1], vcc, s[0:1]
; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], s[0:1]
; GFX1064-NEXT:  ; %bb.1: ; %if.then
; GFX1064-NEXT:    ; divergent unreachable
; GFX1064-NEXT:  ; %bb.2: ; %UnifiedReturnBlock
; GFX1064-NEXT:    s_endpgm
entry:
  %id = tail call i32 @llvm.amdgcn.workitem.id.x()
  %mul4 = mul nsw i32 %s, %n
  %cmp = icmp slt i32 0, %mul4
  br label %if.end

if.end:                                           ; preds = %entry
  %rem = urem i32 %id, %s
  %icmp = tail call i32 @llvm.amdgcn.icmp.i32.i32(i32 %rem, i32 0, i32 32)
  %shr = lshr i32 %icmp, 1
  %notmask = shl nsw i32 -1, 0
  %and = and i32 %notmask, %shr
  %or = or i32 %and, 2147483648
  %cttz = tail call i32 @llvm.cttz.i32(i32 %or, i1 true)
  %cmp3 = icmp ugt i32 10, %cttz
  %cmp6 = icmp ne i32 %rem, 0
  %brmerge = or i1 %cmp6, %cmp3
  br i1 %brmerge, label %if.end2, label %if.then

if.then:                                          ; preds = %if.end
  unreachable

if.end2:                                          ; preds = %if.end
  ret void
}

define amdgpu_kernel void @fcmp32(float %n, float %s) {
; GFX1032-LABEL: fcmp32:
; GFX1032:       ; %bb.0: ; %entry
; GFX1032-NEXT:    s_load_dword s0, s[0:1], 0x28
; GFX1032-NEXT:    v_cvt_f32_u32_e32 v0, v0
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    v_div_scale_f32 v1, s1, s0, s0, v0
; GFX1032-NEXT:    v_rcp_f32_e32 v2, v1
; GFX1032-NEXT:    v_fma_f32 v3, -v1, v2, 1.0
; GFX1032-NEXT:    v_fmac_f32_e32 v2, v3, v2
; GFX1032-NEXT:    v_div_scale_f32 v3, vcc_lo, v0, s0, v0
; GFX1032-NEXT:    v_mul_f32_e32 v4, v3, v2
; GFX1032-NEXT:    v_fma_f32 v5, -v1, v4, v3
; GFX1032-NEXT:    v_fmac_f32_e32 v4, v5, v2
; GFX1032-NEXT:    v_fma_f32 v1, -v1, v4, v3
; GFX1032-NEXT:    v_div_fmas_f32 v1, v1, v2, v4
; GFX1032-NEXT:    v_div_fixup_f32 v1, v1, s0, v0
; GFX1032-NEXT:    v_trunc_f32_e32 v1, v1
; GFX1032-NEXT:    v_fma_f32 v0, -v1, s0, v0
; GFX1032-NEXT:    v_cmp_eq_f32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    s_lshr_b32 s0, vcc_lo, 1
; GFX1032-NEXT:    v_cmp_nlg_f32_e32 vcc_lo, 0, v0
; GFX1032-NEXT:    s_bitset1_b32 s0, 31
; GFX1032-NEXT:    s_ff1_i32_b32 s0, s0
; GFX1032-NEXT:    s_cmp_gt_u32 s0, 9
; GFX1032-NEXT:    s_cselect_b32 s0, -1, 0
; GFX1032-NEXT:    s_and_b32 s0, vcc_lo, s0
; GFX1032-NEXT:    s_and_saveexec_b32 s1, s0
; GFX1032-NEXT:  ; %bb.1: ; %if.then
; GFX1032-NEXT:    ; divergent unreachable
; GFX1032-NEXT:  ; %bb.2: ; %UnifiedReturnBlock
; GFX1032-NEXT:    s_endpgm
;
; GFX1064-LABEL: fcmp32:
; GFX1064:       ; %bb.0: ; %entry
; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x28
; GFX1064-NEXT:    v_cvt_f32_u32_e32 v0, v0
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    v_div_scale_f32 v1, s[0:1], s2, s2, v0
; GFX1064-NEXT:    v_rcp_f32_e32 v2, v1
; GFX1064-NEXT:    v_fma_f32 v3, -v1, v2, 1.0
; GFX1064-NEXT:    v_fmac_f32_e32 v2, v3, v2
; GFX1064-NEXT:    v_div_scale_f32 v3, vcc, v0, s2, v0
; GFX1064-NEXT:    v_mul_f32_e32 v4, v3, v2
; GFX1064-NEXT:    v_fma_f32 v5, -v1, v4, v3
; GFX1064-NEXT:    v_fmac_f32_e32 v4, v5, v2
; GFX1064-NEXT:    v_fma_f32 v1, -v1, v4, v3
; GFX1064-NEXT:    v_div_fmas_f32 v1, v1, v2, v4
; GFX1064-NEXT:    v_div_fixup_f32 v1, v1, s2, v0
; GFX1064-NEXT:    v_trunc_f32_e32 v1, v1
; GFX1064-NEXT:    v_fma_f32 v0, -v1, s2, v0
; GFX1064-NEXT:    v_cmp_eq_f32_e32 vcc, 0, v0
; GFX1064-NEXT:    s_lshr_b32 s0, vcc_lo, 1
; GFX1064-NEXT:    v_cmp_nlg_f32_e32 vcc, 0, v0
; GFX1064-NEXT:    s_bitset1_b32 s0, 31
; GFX1064-NEXT:    s_ff1_i32_b32 s0, s0
; GFX1064-NEXT:    s_cmp_gt_u32 s0, 9
; GFX1064-NEXT:    s_cselect_b64 s[0:1], -1, 0
; GFX1064-NEXT:    s_and_b64 s[0:1], vcc, s[0:1]
; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], s[0:1]
; GFX1064-NEXT:  ; %bb.1: ; %if.then
; GFX1064-NEXT:    ; divergent unreachable
; GFX1064-NEXT:  ; %bb.2: ; %UnifiedReturnBlock
; GFX1064-NEXT:    s_endpgm
entry:
  %id = tail call i32 @llvm.amdgcn.workitem.id.x()
  %id.f = uitofp i32 %id to float
  %mul4 = fmul float %s, %n
  %cmp = fcmp ult float 0.0, %mul4
  br label %if.end

if.end:                                           ; preds = %entry
  %rem.f = frem float %id.f, %s
  %fcmp = tail call i32 @llvm.amdgcn.fcmp.i32.f32(float %rem.f, float 0.0, i32 1)
  %shr = lshr i32 %fcmp, 1
  %notmask = shl nsw i32 -1, 0
  %and = and i32 %notmask, %shr
  %or = or i32 %and, 2147483648
  %cttz = tail call i32 @llvm.cttz.i32(i32 %or, i1 true)
  %cmp3 = icmp ugt i32 10, %cttz
  %cmp6 = fcmp one float %rem.f, 0.0
  %brmerge = or i1 %cmp6, %cmp3
  br i1 %brmerge, label %if.end2, label %if.then

if.then:                                          ; preds = %if.end
  unreachable

if.end2:                                          ; preds = %if.end
  ret void
}

declare void @external_void_func_void() #1

define void @callee_no_stack_with_call() #1 {
; GFX1032-LABEL: callee_no_stack_with_call:
; GFX1032:       ; %bb.0:
; GFX1032-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1032-NEXT:    s_mov_b32 s16, s33
; GFX1032-NEXT:    s_mov_b32 s33, s32
; GFX1032-NEXT:    s_or_saveexec_b32 s17, -1
; GFX1032-NEXT:    buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
; GFX1032-NEXT:    s_mov_b32 exec_lo, s17
; GFX1032-NEXT:    s_addk_i32 s32, 0x200
; GFX1032-NEXT:    v_writelane_b32 v40, s16, 2
; GFX1032-NEXT:    s_getpc_b64 s[16:17]
; GFX1032-NEXT:    s_add_u32 s16, s16, external_void_func_void@gotpcrel32@lo+4
; GFX1032-NEXT:    s_addc_u32 s17, s17, external_void_func_void@gotpcrel32@hi+12
; GFX1032-NEXT:    s_load_dwordx2 s[16:17], s[16:17], 0x0
; GFX1032-NEXT:    v_writelane_b32 v40, s30, 0
; GFX1032-NEXT:    v_writelane_b32 v40, s31, 1
; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1032-NEXT:    s_swappc_b64 s[30:31], s[16:17]
; GFX1032-NEXT:    v_readlane_b32 s31, v40, 1
; GFX1032-NEXT:    v_readlane_b32 s30, v40, 0
; GFX1032-NEXT:    v_readlane_b32 s4, v40, 2
; GFX1032-NEXT:    s_or_saveexec_b32 s5, -1
; GFX1032-NEXT:    buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GFX1032-NEXT:    s_waitcnt_depctr 0xffe3
; GFX1032-NEXT:    s_mov_b32 exec_lo, s5
; GFX1032-NEXT:    s_addk_i32 s32, 0xfe00
; GFX1032-NEXT:    s_mov_b32 s33, s4
; GFX1032-NEXT:    s_waitcnt vmcnt(0)
; GFX1032-NEXT:    s_setpc_b64 s[30:31]
;
; GFX1064-LABEL: callee_no_stack_with_call:
; GFX1064:       ; %bb.0:
; GFX1064-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX1064-NEXT:    s_mov_b32 s16, s33
; GFX1064-NEXT:    s_mov_b32 s33, s32
; GFX1064-NEXT:    s_or_saveexec_b64 s[18:19], -1
; GFX1064-NEXT:    buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill
; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
; GFX1064-NEXT:    s_mov_b64 exec, s[18:19]
; GFX1064-NEXT:    s_addk_i32 s32, 0x400
; GFX1064-NEXT:    v_writelane_b32 v40, s16, 2
; GFX1064-NEXT:    s_getpc_b64 s[16:17]
; GFX1064-NEXT:    s_add_u32 s16, s16, external_void_func_void@gotpcrel32@lo+4
; GFX1064-NEXT:    s_addc_u32 s17, s17, external_void_func_void@gotpcrel32@hi+12
; GFX1064-NEXT:    s_load_dwordx2 s[16:17], s[16:17], 0x0
; GFX1064-NEXT:    v_writelane_b32 v40, s30, 0
; GFX1064-NEXT:    v_writelane_b32 v40, s31, 1
; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
; GFX1064-NEXT:    s_swappc_b64 s[30:31], s[16:17]
; GFX1064-NEXT:    v_readlane_b32 s31, v40, 1
; GFX1064-NEXT:    v_readlane_b32 s30, v40, 0
; GFX1064-NEXT:    v_readlane_b32 s4, v40, 2
; GFX1064-NEXT:    s_or_saveexec_b64 s[6:7], -1
; GFX1064-NEXT:    buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GFX1064-NEXT:    s_waitcnt_depctr 0xffe3
; GFX1064-NEXT:    s_mov_b64 exec, s[6:7]
; GFX1064-NEXT:    s_addk_i32 s32, 0xfc00
; GFX1064-NEXT:    s_mov_b32 s33, s4
; GFX1064-NEXT:    s_waitcnt vmcnt(0)
; GFX1064-NEXT:    s_setpc_b64 s[30:31]
  call void @external_void_func_void()
  ret void
}


declare i32 @llvm.amdgcn.workitem.id.x()
declare float @llvm.fabs.f32(float)
declare { float, i1 } @llvm.amdgcn.div.scale.f32(float, float, i1)
declare { double, i1 } @llvm.amdgcn.div.scale.f64(double, double, i1)
declare float @llvm.amdgcn.div.fmas.f32(float, float, float, i1)
declare double @llvm.amdgcn.div.fmas.f64(double, double, double, i1)
declare i1 @llvm.amdgcn.class.f32(float, i32)
declare i32 @llvm.amdgcn.set.inactive.i32(i32, i32)
declare i64 @llvm.amdgcn.set.inactive.i64(i64, i64)
declare <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32, float, <8 x i32>, <4 x i32>, i1, i32, i32)
declare <4 x float> @llvm.amdgcn.image.sample.2d.v4f32.f32(i32, float, float, <8 x i32>, <4 x i32>, i1, i32, i32)
declare float @llvm.amdgcn.strict.wwm.f32(float)
declare float @llvm.amdgcn.wwm.f32(float)
declare i32 @llvm.amdgcn.wqm.i32(i32)
declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32)
declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32)
declare float @llvm.amdgcn.struct.ptr.buffer.load.f32(ptr addrspace(8), i32, i32, i32, i32 immarg)
declare i32 @llvm.amdgcn.mbcnt.lo(i32, i32)
declare i32 @llvm.amdgcn.mbcnt.hi(i32, i32)
declare i64 @llvm.amdgcn.fcmp.i64.f32(float, float, i32)
declare i64 @llvm.amdgcn.icmp.i64.i32(i32, i32, i32)
declare i32 @llvm.amdgcn.fcmp.i32.f32(float, float, i32)
declare i32 @llvm.amdgcn.icmp.i32.i32(i32, i32, i32)
declare void @llvm.amdgcn.kill(i1)
declare i1 @llvm.amdgcn.wqm.vote(i1)
declare i1 @llvm.amdgcn.ps.live()
declare i64 @llvm.cttz.i64(i64, i1)
declare i32 @llvm.cttz.i32(i32, i1)
declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #5

attributes #0 = { nounwind readnone speculatable }
attributes #1 = { nounwind }
attributes #2 = { nounwind readnone optnone noinline }
attributes #3 = { "target-features"="+wavefrontsize32" }
attributes #4 = { "target-features"="+wavefrontsize64" }
attributes #5 = { inaccessiblememonly nounwind }
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; GFX10DEFWAVE: {{.*}}