Google Git
Sign in
llvm/llvm-project/refs/heads/users/cdevadas/enhance-createFrom-function/./llvm/test/CodeGen/AMDGPU/wqm-propagate-for-execz-side-effect.mir
blob: d77704857bba53ea2687401dc240b8b7ab499ba5 [file] [edit]
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 6
# RUN: llc -mtriple=amdgpu12.00 -run-pass=si-wqm -verify-machineinstrs -o - %s | FileCheck %s
# Make sure the sources of execz side-effect instruction are calculated under wqm too.
--- |
define amdgpu_ps float @execz_side_effect_wqm_propagation(float %FragCoordY, i32 %arg, i1 %arg1) {
.entry:
ret float poison
}
...
---
name: execz_side_effect_wqm_propagation
alignment: 4
tracksRegLiveness: true
noPhis: true
isSSA: false
machineFunctionInfo:
isEntryFunction: true
returnsVoid: false
body: |
; CHECK-LABEL: name: execz_side_effect_wqm_propagation
; CHECK: bb.0:
; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.8(0x40000000)
; CHECK-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $exec_lo
; CHECK-NEXT: $exec_lo = S_WQM_B32 $exec_lo, implicit-def $scc
; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK-NEXT: [[V_AND_B32_e64_:%[0-9]+]]:vgpr_32 = V_AND_B32_e64 1, [[COPY1]], implicit $exec
; CHECK-NEXT: [[V_CMP_NE_U32_e64_:%[0-9]+]]:sreg_32 = V_CMP_NE_U32_e64 1, [[V_AND_B32_e64_]], implicit $exec
; CHECK-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
; CHECK-NEXT: [[COPY4:%[0-9]+]]:sreg_32 = COPY $exec_lo, implicit-def $exec_lo
; CHECK-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32 = S_AND_B32 [[COPY4]], [[V_CMP_NE_U32_e64_]], implicit-def dead $scc
; CHECK-NEXT: $exec_lo = S_MOV_B32_term [[S_AND_B32_]]
; CHECK-NEXT: S_CBRANCH_EXECZ %bb.8, implicit $exec
; CHECK-NEXT: S_BRANCH %bb.1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.1:
; CHECK-NEXT: successors: %bb.5(0x40000000), %bb.2(0x40000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: undef [[S_MOV_B32_:%[0-9]+]].sub0:sgpr_128 = S_MOV_B32 0
; CHECK-NEXT: [[S_MOV_B32_:%[0-9]+]].sub1:sgpr_128 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: [[S_MOV_B32_:%[0-9]+]].sub2:sgpr_128 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: [[S_MOV_B32_:%[0-9]+]].sub3:sgpr_128 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: undef [[COPY5:%[0-9]+]].sub0:sgpr_256 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: [[COPY5:%[0-9]+]].sub1:sgpr_256 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: [[COPY5:%[0-9]+]].sub2:sgpr_256 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: [[COPY5:%[0-9]+]].sub3:sgpr_256 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: [[COPY5:%[0-9]+]].sub4:sgpr_256 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
; CHECK-NEXT: [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32 = S_AND_SAVEEXEC_B32 [[COPY]], implicit-def $exec, implicit-def $scc, implicit $exec
; CHECK-NEXT: [[COPY5:%[0-9]+]].sub5:sgpr_256 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: [[COPY5:%[0-9]+]].sub6:sgpr_256 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: [[COPY5:%[0-9]+]].sub7:sgpr_256 = COPY [[S_MOV_B32_]].sub0
; CHECK-NEXT: [[IMAGE_SAMPLE_V1_V3_gfx12_:%[0-9]+]]:vgpr_32 = IMAGE_SAMPLE_V1_V3_gfx12 [[V_MOV_B32_e32_]], [[V_MOV_B32_e32_]], [[COPY3]], [[COPY5]], [[S_MOV_B32_]], 1, 3, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 8)
; CHECK-NEXT: IMAGE_ATOMIC_ADD_NORTN_V1_V2_gfx12 [[V_MOV_B32_e32_]], [[V_MOV_B32_e32_]], [[V_MOV_B32_e32_]], [[COPY5]], 1, 1, 0, 0, 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), addrspace 8)
; CHECK-NEXT: $exec_lo = COPY [[S_AND_SAVEEXEC_B32_]]
; CHECK-NEXT: [[V_MED3_F32_e64_:%[0-9]+]]:vgpr_32 = V_MED3_F32_e64 0, 0, 0, 0, 0, 0, 0, 0, implicit $mode, implicit $exec
; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[S_MOV_B32_]], 0, 0 :: (dereferenceable invariant load (s32))
; CHECK-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[V_MED3_F32_e64_]], implicit $exec
; CHECK-NEXT: [[S_MUL_F32_:%[0-9]+]]:sgpr_32 = nofpexcept S_MUL_F32 [[V_READFIRSTLANE_B32_]], 0, implicit $mode
; CHECK-NEXT: [[S_MUL_F32_1:%[0-9]+]]:sgpr_32 = nofpexcept S_MUL_F32 [[S_MUL_F32_]], 0, implicit $mode
; CHECK-NEXT: [[S_CVT_I32_F32_:%[0-9]+]]:sreg_32 = nofpexcept S_CVT_I32_F32 [[S_MUL_F32_1]], implicit $mode
; CHECK-NEXT: [[V_CVT_I32_F32_e64_:%[0-9]+]]:vgpr_32 = nofpexcept V_CVT_I32_F32_e64 0, [[IMAGE_SAMPLE_V1_V3_gfx12_]], 0, 0, implicit $mode, implicit $exec
; CHECK-NEXT: [[V_MUL_LO_U32_e64_:%[0-9]+]]:vgpr_32 = V_MUL_LO_U32_e64 [[S_BUFFER_LOAD_DWORD_IMM]], [[COPY2]], implicit $exec
; CHECK-NEXT: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[S_CVT_I32_F32_]], 8, implicit-def dead $scc
; CHECK-NEXT: undef [[V_OR_B16_t16_e64_:%[0-9]+]].lo16:vreg_128 = V_OR_B16_t16_e64 0, [[V_CVT_I32_F32_e64_]].lo16, 0, [[S_LSHL_B32_]], 0, implicit $exec
; CHECK-NEXT: [[V_OR_B16_t16_e64_:%[0-9]+]].hi16:vreg_128 = V_MOV_B16_t16_e64 0, 0, 0, implicit $exec
; CHECK-NEXT: [[COPY6:%[0-9]+]]:sreg_32 = COPY $exec_lo, implicit-def $exec_lo
; CHECK-NEXT: [[S_AND_B32_1:%[0-9]+]]:sreg_32 = S_AND_B32 [[COPY6]], [[V_CMP_NE_U32_e64_]], implicit-def dead $scc
; CHECK-NEXT: [[S_XOR_B32_:%[0-9]+]]:sreg_32 = S_XOR_B32 [[S_AND_B32_1]], [[COPY6]], implicit-def dead $scc
; CHECK-NEXT: $exec_lo = S_MOV_B32_term [[S_AND_B32_1]]
; CHECK-NEXT: S_CBRANCH_EXECZ %bb.2, implicit $exec
; CHECK-NEXT: S_BRANCH %bb.5
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.2:
; CHECK-NEXT: successors: %bb.3(0x40000000), %bb.7(0x40000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[S_OR_SAVEEXEC_B32_:%[0-9]+]]:sreg_32 = S_OR_SAVEEXEC_B32 [[S_XOR_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
; CHECK-NEXT: [[S_AND_B32_2:%[0-9]+]]:sreg_32 = S_AND_B32 $exec_lo, [[S_OR_SAVEEXEC_B32_]], implicit-def $scc
; CHECK-NEXT: $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_B32_2]], implicit-def $scc
; CHECK-NEXT: S_CBRANCH_EXECZ %bb.7, implicit $exec
; CHECK-NEXT: S_BRANCH %bb.3
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.3:
; CHECK-NEXT: successors: %bb.4(0x40000000), %bb.6(0x40000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: undef [[S_MOV_B32_1:%[0-9]+]].sub0:sgpr_128 = S_MOV_B32 0
; CHECK-NEXT: [[S_MOV_B32_1:%[0-9]+]].sub1:sgpr_128 = COPY [[S_MOV_B32_1]].sub0
; CHECK-NEXT: [[S_MOV_B32_1:%[0-9]+]].sub2:sgpr_128 = COPY [[S_MOV_B32_1]].sub0
; CHECK-NEXT: [[S_MOV_B32_1:%[0-9]+]].sub3:sgpr_128 = COPY [[S_MOV_B32_1]].sub0
; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_XY_VBUFFER_IDXEN:%[0-9]+]]:vreg_64 = BUFFER_LOAD_FORMAT_XY_VBUFFER_IDXEN [[V_MUL_LO_U32_e64_]], [[S_MOV_B32_1]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 8)
; CHECK-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
; CHECK-NEXT: [[V_CMP_LT_F32_e64_:%[0-9]+]]:sreg_32 = nofpexcept V_CMP_LT_F32_e64 0, 0, 0, [[BUFFER_LOAD_FORMAT_XY_VBUFFER_IDXEN]].sub0, 0, implicit $mode, implicit $exec
; CHECK-NEXT: [[COPY7:%[0-9]+]]:sreg_32 = COPY $exec_lo, implicit-def $exec_lo
; CHECK-NEXT: [[S_AND_B32_3:%[0-9]+]]:sreg_32 = S_AND_B32 [[COPY7]], [[V_CMP_LT_F32_e64_]], implicit-def dead $scc
; CHECK-NEXT: $exec_lo = S_MOV_B32_term [[S_AND_B32_3]]
; CHECK-NEXT: S_CBRANCH_EXECZ %bb.6, implicit $exec
; CHECK-NEXT: S_BRANCH %bb.4
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.4:
; CHECK-NEXT: successors: %bb.6(0x80000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = COPY [[V_OR_B16_t16_e64_]].sub0
; CHECK-NEXT: S_BRANCH %bb.6
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.5:
; CHECK-NEXT: successors: %bb.2(0x80000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[S_AND_SAVEEXEC_B32_1:%[0-9]+]]:sreg_32 = S_AND_SAVEEXEC_B32 [[COPY]], implicit-def $exec, implicit-def $scc, implicit $exec
; CHECK-NEXT: [[V_OR_B16_t16_e64_:%[0-9]+]].sub1:vreg_128 = V_MOV_B32_e32 0, implicit $exec
; CHECK-NEXT: [[V_OR_B16_t16_e64_:%[0-9]+]].sub2:vreg_128 = COPY [[V_OR_B16_t16_e64_]].sub1
; CHECK-NEXT: [[V_OR_B16_t16_e64_:%[0-9]+]].sub3:vreg_128 = COPY [[V_OR_B16_t16_e64_]].sub1
; CHECK-NEXT: BUFFER_STORE_FORMAT_XYZW_VBUFFER_IDXEN_exact [[V_OR_B16_t16_e64_]], [[V_MUL_LO_U32_e64_]], [[S_MOV_B32_]], $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 8)
; CHECK-NEXT: [[V_MUL_LO_U32_e64_:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
; CHECK-NEXT: undef [[V_OR_B16_t16_e64_:%[0-9]+]].sub0:vreg_128 = IMPLICIT_DEF
; CHECK-NEXT: $exec_lo = COPY [[S_AND_SAVEEXEC_B32_1]]
; CHECK-NEXT: S_BRANCH %bb.2
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.6:
; CHECK-NEXT: successors: %bb.7(0x80000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: $exec_lo = S_OR_B32 $exec_lo, [[COPY7]], implicit-def $scc
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.7:
; CHECK-NEXT: successors: %bb.8(0x80000000)
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: $exec_lo = S_OR_B32 $exec_lo, [[S_AND_B32_2]], implicit-def $scc
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.8:
; CHECK-NEXT: $exec_lo = S_OR_B32 $exec_lo, [[COPY4]], implicit-def $scc
; CHECK-NEXT: $exec_lo = S_AND_B32 $exec_lo, [[COPY]], implicit-def $scc
; CHECK-NEXT: $vgpr0 = COPY [[V_MOV_B32_e32_]]
; CHECK-NEXT: SI_RETURN_TO_EPILOG $vgpr0
bb.0:
successors: %bb.1, %bb.8
liveins: $vgpr0, $vgpr1, $vgpr2
%2:vgpr_32 = COPY $vgpr2
%1:vgpr_32 = COPY $vgpr1
%0:vgpr_32 = COPY $vgpr0
%3:vgpr_32 = V_AND_B32_e64 1, %2, implicit $exec
%4:sreg_32 = V_CMP_NE_U32_e64 1, %3, implicit $exec
%5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
%6:sreg_32 = COPY $exec_lo, implicit-def $exec_lo
%7:sreg_32 = S_AND_B32 %6, %4, implicit-def dead $scc
$exec_lo = S_MOV_B32_term %7
S_CBRANCH_EXECZ %bb.8, implicit $exec
S_BRANCH %bb.1
bb.1:
successors: %bb.5, %bb.2
undef %8.sub0:sgpr_128 = S_MOV_B32 0
%8.sub1:sgpr_128 = COPY %8.sub0
%8.sub2:sgpr_128 = COPY %8.sub0
%8.sub3:sgpr_128 = COPY %8.sub0
undef %9.sub0:sgpr_256 = COPY %8.sub0
%9.sub1:sgpr_256 = COPY %8.sub0
%9.sub2:sgpr_256 = COPY %8.sub0
%9.sub3:sgpr_256 = COPY %8.sub0
%9.sub4:sgpr_256 = COPY %8.sub0
%5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
%9.sub5:sgpr_256 = COPY %8.sub0
%9.sub6:sgpr_256 = COPY %8.sub0
%9.sub7:sgpr_256 = COPY %8.sub0
%10:vgpr_32 = IMAGE_SAMPLE_V1_V3_gfx12 %5, %5, %0, %9, %8, 1, 3, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), addrspace 8)
IMAGE_ATOMIC_ADD_NORTN_V1_V2_gfx12 %5, %5, %5, %9, 1, 1, 0, 0, 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), addrspace 8)
%11:vgpr_32 = V_MED3_F32_e64 0, 0, 0, 0, 0, 0, 0, 0, implicit $mode, implicit $exec
%12:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM %8, 0, 0 :: (dereferenceable invariant load (s32))
%13:sreg_32_xm0 = V_READFIRSTLANE_B32 %11, implicit $exec
%14:sgpr_32 = nofpexcept S_MUL_F32 %13, 0, implicit $mode
%15:sgpr_32 = nofpexcept S_MUL_F32 %14, 0, implicit $mode
%16:sreg_32 = nofpexcept S_CVT_I32_F32 %15, implicit $mode
%17:vgpr_32 = nofpexcept V_CVT_I32_F32_e64 0, %10, 0, 0, implicit $mode, implicit $exec
%18:vgpr_32 = V_MUL_LO_U32_e64 %12, %1, implicit $exec
%19:sreg_32 = S_LSHL_B32 %16, 8, implicit-def dead $scc
undef %20.lo16:vreg_128 = V_OR_B16_t16_e64 0, %17.lo16, 0, %19, 0, implicit $exec
%20.hi16:vreg_128 = V_MOV_B16_t16_e64 0, 0, 0, implicit $exec
%21:sreg_32 = COPY $exec_lo, implicit-def $exec_lo
%22:sreg_32 = S_AND_B32 %21, %4, implicit-def dead $scc
%23:sreg_32 = S_XOR_B32 %22, %21, implicit-def dead $scc
$exec_lo = S_MOV_B32_term %22
S_CBRANCH_EXECZ %bb.2, implicit $exec
S_BRANCH %bb.5
bb.2:
successors: %bb.3, %bb.7
%24:sreg_32 = S_OR_SAVEEXEC_B32 %23, implicit-def $exec, implicit-def $scc, implicit $exec
%25:sreg_32 = S_AND_B32 $exec_lo, %24, implicit-def $scc
$exec_lo = S_XOR_B32_term $exec_lo, %25, implicit-def $scc
S_CBRANCH_EXECZ %bb.7, implicit $exec
S_BRANCH %bb.3
bb.3:
successors: %bb.4, %bb.6
undef %26.sub0:sgpr_128 = S_MOV_B32 0
%26.sub1:sgpr_128 = COPY %26.sub0
%26.sub2:sgpr_128 = COPY %26.sub0
%26.sub3:sgpr_128 = COPY %26.sub0
%27:vreg_64 = BUFFER_LOAD_FORMAT_XY_VBUFFER_IDXEN %18, %26, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable load (s64), align 1, addrspace 8)
%5:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
%28:sreg_32 = nofpexcept V_CMP_LT_F32_e64 0, 0, 0, %27.sub0, 0, implicit $mode, implicit $exec
%29:sreg_32 = COPY $exec_lo, implicit-def $exec_lo
%30:sreg_32 = S_AND_B32 %29, %28, implicit-def dead $scc
$exec_lo = S_MOV_B32_term %30
S_CBRANCH_EXECZ %bb.6, implicit $exec
S_BRANCH %bb.4
bb.4:
%5:vgpr_32 = COPY %20.sub0
S_BRANCH %bb.6
bb.5:
%20.sub1:vreg_128 = V_MOV_B32_e32 0, implicit $exec
%20.sub2:vreg_128 = COPY %20.sub1
%20.sub3:vreg_128 = COPY %20.sub1
BUFFER_STORE_FORMAT_XYZW_VBUFFER_IDXEN_exact %20, %18, %8, $sgpr_null, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), align 1, addrspace 8)
%18:vgpr_32 = IMPLICIT_DEF
undef %20.sub0:vreg_128 = IMPLICIT_DEF
S_BRANCH %bb.2
bb.6:
$exec_lo = S_OR_B32 $exec_lo, %29, implicit-def $scc
bb.7:
$exec_lo = S_OR_B32 $exec_lo, %25, implicit-def $scc
bb.8:
$exec_lo = S_OR_B32 $exec_lo, %6, implicit-def $scc
$vgpr0 = COPY %5
SI_RETURN_TO_EPILOG $vgpr0
...