diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp index e657f668cc656..271c8d45fd4a2 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp @@ -289,9 +289,11 @@ static const LLT F64 = LLT::float64(); static const LLT S96 = LLT::scalar(96); static const LLT S128 = LLT::scalar(128); static const LLT S160 = LLT::scalar(160); +static const LLT S192 = LLT::scalar(192); static const LLT S224 = LLT::scalar(224); static const LLT S256 = LLT::scalar(256); static const LLT S512 = LLT::scalar(512); +static const LLT S1024 = LLT::scalar(1024); static const LLT MaxScalar = LLT::scalar(MaxRegisterSize); static const LLT V2S8 = LLT::fixed_vector(2, 8); @@ -332,8 +334,8 @@ static const LLT V16S64 = LLT::fixed_vector(16, 64); static const LLT V2S128 = LLT::fixed_vector(2, 128); static const LLT V4S128 = LLT::fixed_vector(4, 128); -static std::initializer_list AllScalarTypes = {S32, S64, S96, S128, - S160, S224, S256, S512}; +static std::initializer_list AllScalarTypes = { + S32, S64, S96, S128, S160, S192, S224, S256, S512, S1024}; static std::initializer_list AllS16Vectors{ V2S16, V4S16, V6S16, V8S16, V10S16, V12S16, V16S16, V2S128, V4S128}; @@ -889,10 +891,11 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_, .clampScalar(0, S16, S64); getActionDefinitionsBuilder({G_IMPLICIT_DEF, G_FREEZE}) - .legalIf(isRegisterType(0)) + .legalIf(isRegisterClassType(0)) // s1 and s16 are special cases because they have legal operations on // them, but don't really occupy registers in the normal way. .legalFor({S1, S16}) + .clampNumElements(0, V16S32, V32S32) .moreElementsIf(isSmallOddVector(0), oneMoreElement(0)) .clampScalarOrElt(0, S32, MaxScalar) .widenScalarToNextPow2(0, 32) diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-unmerge-values.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-unmerge-values.mir index bec5f646b7839..837f65d4bdec6 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-unmerge-values.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-unmerge-values.mir @@ -171,11 +171,9 @@ body: | ; GCN-LABEL: name: test_unmerge_values_s_s64_s_s64_s64_s_s192 ; GCN: liveins: $sgpr0_sgpr1_sgpr2_sgpr3 ; GCN-NEXT: {{ $}} - ; GCN-NEXT: [[DEF:%[0-9]+]]:sgpr_192 = IMPLICIT_DEF - ; GCN-NEXT: [[COPY:%[0-9]+]]:sreg_64 = COPY [[DEF]].sub0_sub1 - ; GCN-NEXT: [[COPY1:%[0-9]+]]:sreg_64 = COPY [[DEF]].sub2_sub3 - ; GCN-NEXT: [[COPY2:%[0-9]+]]:sreg_64 = COPY [[DEF]].sub4_sub5 - ; GCN-NEXT: S_ENDPGM 0, implicit [[COPY]], implicit [[COPY1]], implicit [[COPY2]] + ; GCN-NEXT: [[DEF:%[0-9]+]]:sgpr(s192) = G_IMPLICIT_DEF + ; GCN-NEXT: [[UV:%[0-9]+]]:sgpr(s64), [[UV1:%[0-9]+]]:sgpr(s64), [[UV2:%[0-9]+]]:sgpr(s64) = G_UNMERGE_VALUES [[DEF]](s192) + ; GCN-NEXT: S_ENDPGM 0, implicit [[UV]](s64), implicit [[UV1]](s64), implicit [[UV2]](s64) %0:sgpr(s192) = G_IMPLICIT_DEF %1:sgpr(s64), %2:sgpr(s64), %3:sgpr(s64) = G_UNMERGE_VALUES %0 S_ENDPGM 0, implicit %1, implicit %2, implicit %3 @@ -294,11 +292,11 @@ body: | ; GCN-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:sgpr_384(<12 x s32>) = G_CONCAT_VECTORS [[COPY]](<3 x s32>), [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>) ; GCN-NEXT: [[COPY4:%[0-9]+]]:sgpr_96(<3 x s32>) = COPY [[CONCAT_VECTORS]].sub0_sub1_sub2(<12 x s32>) ; GCN-NEXT: [[COPY5:%[0-9]+]]:sgpr_96(<3 x s32>) = COPY [[CONCAT_VECTORS]].sub3_sub4_sub5(<12 x s32>) - ; GCN-NEXT: [[UV:%[0-9]+]]:sgpr_96(<3 x s32>), [[UV1:%[0-9]+]]:sgpr_96(<3 x s32>), [[UV2:%[0-9]+]]:sgpr_96(<3 x s32>), [[UV3:%[0-9]+]]:sgpr_96(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) - ; GCN-NEXT: $sgpr0_sgpr1_sgpr2 = COPY [[UV]](<3 x s32>) - ; GCN-NEXT: $sgpr4_sgpr5_sgpr6 = COPY [[UV1]](<3 x s32>) - ; GCN-NEXT: $sgpr8_sgpr9_sgpr10 = COPY [[UV2]](<3 x s32>) - ; GCN-NEXT: $sgpr12_sgpr13_sgpr14 = COPY [[UV3]](<3 x s32>) + ; GCN-NEXT: [[COPY4:%[0-9]+]]:sgpr_96(<3 x s32>), [[COPY5:%[0-9]+]]:sgpr_96(<3 x s32>), [[UV:%[0-9]+]]:sgpr_96(<3 x s32>), [[UV1:%[0-9]+]]:sgpr_96(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) + ; GCN-NEXT: $sgpr0_sgpr1_sgpr2 = COPY [[COPY4]](<3 x s32>) + ; GCN-NEXT: $sgpr4_sgpr5_sgpr6 = COPY [[COPY5]](<3 x s32>) + ; GCN-NEXT: $sgpr8_sgpr9_sgpr10 = COPY [[UV]](<3 x s32>) + ; GCN-NEXT: $sgpr12_sgpr13_sgpr14 = COPY [[UV1]](<3 x s32>) %0:sgpr(<3 x s32>) = COPY $sgpr0_sgpr1_sgpr2 %1:sgpr(<3 x s32>) = COPY $sgpr4_sgpr5_sgpr6 %2:sgpr(<3 x s32>) = COPY $sgpr8_sgpr9_sgpr10 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-freeze.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-freeze.mir index c06df6312c9c5..b08f850b5b2b1 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-freeze.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-freeze.mir @@ -171,12 +171,8 @@ body: | ; CHECK-LABEL: name: test_freeze_s448 ; CHECK: [[COPY:%[0-9]+]]:_(s512) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 - ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s448) = G_TRUNC [[COPY]](s512) - ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(s448) = G_FREEZE [[TRUNC]] - ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64), [[UV6:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[FREEZE]](s448) - ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF - ; CHECK-NEXT: [[MV:%[0-9]+]]:_(s512) = G_MERGE_VALUES [[UV]](s64), [[UV1]](s64), [[UV2]](s64), [[UV3]](s64), [[UV4]](s64), [[UV5]](s64), [[UV6]](s64), [[DEF]](s64) - ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[MV]](s512) + ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(s512) = G_FREEZE [[COPY]] + ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[FREEZE]](s512) %0:_(s512) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 %1:_(s448) = G_TRUNC %0 %2:_(s448) = G_FREEZE %1 @@ -399,14 +395,12 @@ body: | bb.0: ; CHECK-LABEL: name: test_freeze_v33s32 - ; CHECK: [[DEF:%[0-9]+]]:_(<16 x s32>) = G_IMPLICIT_DEF + ; CHECK: [[DEF:%[0-9]+]]:_(<32 x s32>) = G_IMPLICIT_DEF ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<16 x s32>) = G_FREEZE [[DEF]] - ; CHECK-NEXT: [[FREEZE1:%[0-9]+]]:_(<16 x s32>) = G_FREEZE [[DEF]] - ; CHECK-NEXT: [[FREEZE2:%[0-9]+]]:_(s32) = G_FREEZE [[DEF1]] - ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[FREEZE]](<16 x s32>) - ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[FREEZE1]](<16 x s32>) - ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<33 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[UV14]](s32), [[UV15]](s32), [[UV16]](s32), [[UV17]](s32), [[UV18]](s32), [[UV19]](s32), [[UV20]](s32), [[UV21]](s32), [[UV22]](s32), [[UV23]](s32), [[UV24]](s32), [[UV25]](s32), [[UV26]](s32), [[UV27]](s32), [[UV28]](s32), [[UV29]](s32), [[UV30]](s32), [[UV31]](s32), [[FREEZE2]](s32) + ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<32 x s32>) = G_FREEZE [[DEF]] + ; CHECK-NEXT: [[FREEZE1:%[0-9]+]]:_(s32) = G_FREEZE [[DEF1]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32), [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[FREEZE]](<32 x s32>) + ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<33 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[UV14]](s32), [[UV15]](s32), [[UV16]](s32), [[UV17]](s32), [[UV18]](s32), [[UV19]](s32), [[UV20]](s32), [[UV21]](s32), [[UV22]](s32), [[UV23]](s32), [[UV24]](s32), [[UV25]](s32), [[UV26]](s32), [[UV27]](s32), [[UV28]](s32), [[UV29]](s32), [[UV30]](s32), [[UV31]](s32), [[FREEZE1]](s32) ; CHECK-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<33 x s32>) %0:_(<33 x s32>) = G_IMPLICIT_DEF %1:_(<33 x s32>) = G_FREEZE %0 @@ -419,12 +413,10 @@ body: | bb.0: ; CHECK-LABEL: name: test_freeze_v64s32 - ; CHECK: [[DEF:%[0-9]+]]:_(<16 x s32>) = G_IMPLICIT_DEF - ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<16 x s32>) = G_FREEZE [[DEF]] - ; CHECK-NEXT: [[FREEZE1:%[0-9]+]]:_(<16 x s32>) = G_FREEZE [[DEF]] - ; CHECK-NEXT: [[FREEZE2:%[0-9]+]]:_(<16 x s32>) = G_FREEZE [[DEF]] - ; CHECK-NEXT: [[FREEZE3:%[0-9]+]]:_(<16 x s32>) = G_FREEZE [[DEF]] - ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<64 x s32>) = G_CONCAT_VECTORS [[FREEZE]](<16 x s32>), [[FREEZE1]](<16 x s32>), [[FREEZE2]](<16 x s32>), [[FREEZE3]](<16 x s32>) + ; CHECK: [[DEF:%[0-9]+]]:_(<32 x s32>) = G_IMPLICIT_DEF + ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<32 x s32>) = G_FREEZE [[DEF]] + ; CHECK-NEXT: [[FREEZE1:%[0-9]+]]:_(<32 x s32>) = G_FREEZE [[DEF]] + ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<64 x s32>) = G_CONCAT_VECTORS [[FREEZE]](<32 x s32>), [[FREEZE1]](<32 x s32>) ; CHECK-NEXT: S_NOP 0, implicit [[CONCAT_VECTORS]](<64 x s32>) %0:_(<64 x s32>) = G_IMPLICIT_DEF %1:_(<64 x s32>) = G_FREEZE %0 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-implicit-def.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-implicit-def.mir index b9edfbfa6d0a9..8113ebfa5362e 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-implicit-def.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-implicit-def.mir @@ -135,8 +135,9 @@ body: | bb.0: ; CHECK-LABEL: name: test_implicit_def_s448 - ; CHECK: [[DEF:%[0-9]+]]:_(s448) = G_IMPLICIT_DEF - ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s32) = G_EXTRACT [[DEF]](s448), 0 + ; CHECK: [[DEF:%[0-9]+]]:_(s512) = G_IMPLICIT_DEF + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s448) = G_TRUNC [[DEF]](s512) + ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s32) = G_EXTRACT [[TRUNC]](s448), 0 ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](s32) %0:_(s448) = G_IMPLICIT_DEF %1:_(s32) = G_EXTRACT %0, 0 @@ -295,18 +296,6 @@ body: | $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY %0 ... ---- -name: test_implicit_def_v17s32 -body: | - bb.0: - - ; CHECK-LABEL: name: test_implicit_def_v17s32 - ; CHECK: [[DEF:%[0-9]+]]:_(<17 x s32>) = G_IMPLICIT_DEF - ; CHECK-NEXT: S_NOP 0, implicit [[DEF]](<17 x s32>) - %0:_(<17 x s32>) = G_IMPLICIT_DEF - S_NOP 0, implicit %0 -... - --- name: test_implicit_def_v32s32 body: | @@ -328,9 +317,9 @@ body: | ; CHECK-LABEL: name: test_implicit_def_v33s32 ; CHECK: liveins: $vgpr0_vgpr1 ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<16 x s32>) = G_IMPLICIT_DEF + ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<32 x s32>) = G_IMPLICIT_DEF ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) + ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32), [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CHECK-NEXT: G_STORE [[UV]](s32), [[COPY]](p1) :: (volatile store (s32), addrspace 1) ; CHECK-NEXT: G_STORE [[DEF1]](s32), [[COPY]](p1) :: (volatile store (s32), addrspace 1) @@ -348,10 +337,9 @@ body: | bb.0: ; CHECK-LABEL: name: test_implicit_def_v64s32 - ; CHECK: [[DEF:%[0-9]+]]:_(<16 x s32>) = G_IMPLICIT_DEF - ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<64 x s32>) = G_CONCAT_VECTORS [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<32 x s32>) = G_CONCAT_VECTORS [[DEF]](<16 x s32>), [[DEF]](<16 x s32>) - ; CHECK-NEXT: S_NOP 0, implicit [[CONCAT_VECTORS]](<64 x s32>), implicit [[CONCAT_VECTORS1]](<32 x s32>) + ; CHECK: [[DEF:%[0-9]+]]:_(<32 x s32>) = G_IMPLICIT_DEF + ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<64 x s32>) = G_CONCAT_VECTORS [[DEF]](<32 x s32>), [[DEF]](<32 x s32>) + ; CHECK-NEXT: S_NOP 0, implicit [[CONCAT_VECTORS]](<64 x s32>), implicit [[DEF]](<32 x s32>) %0:_(<64 x s32>) = G_IMPLICIT_DEF %1:_(<32 x s32>), %2:_(<32 x s32>) = G_UNMERGE_VALUES %0 S_NOP 0, implicit %0, implicit %1 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-insert-vector-elt.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-insert-vector-elt.mir index b57dd396ae355..bebbf2a262256 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-insert-vector-elt.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-insert-vector-elt.mir @@ -190,13 +190,11 @@ body: | ; CHECK-LABEL: name: insert_vector_elt_64_65_v64s32 ; CHECK: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1, $vgpr2_vgpr3 ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<16 x s32>) = G_IMPLICIT_DEF + ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<32 x s32>) = G_IMPLICIT_DEF ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(p1) = COPY $vgpr2_vgpr3 - ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>), [[UV2:%[0-9]+]]:_(<4 x s32>), [[UV3:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<4 x s32>), [[UV5:%[0-9]+]]:_(<4 x s32>), [[UV6:%[0-9]+]]:_(<4 x s32>), [[UV7:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<4 x s32>), [[UV9:%[0-9]+]]:_(<4 x s32>), [[UV10:%[0-9]+]]:_(<4 x s32>), [[UV11:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(<4 x s32>), [[UV13:%[0-9]+]]:_(<4 x s32>), [[UV14:%[0-9]+]]:_(<4 x s32>), [[UV15:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) + ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<4 x s32>), [[UV1:%[0-9]+]]:_(<4 x s32>), [[UV2:%[0-9]+]]:_(<4 x s32>), [[UV3:%[0-9]+]]:_(<4 x s32>), [[UV4:%[0-9]+]]:_(<4 x s32>), [[UV5:%[0-9]+]]:_(<4 x s32>), [[UV6:%[0-9]+]]:_(<4 x s32>), [[UV7:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) + ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<4 x s32>), [[UV9:%[0-9]+]]:_(<4 x s32>), [[UV10:%[0-9]+]]:_(<4 x s32>), [[UV11:%[0-9]+]]:_(<4 x s32>), [[UV12:%[0-9]+]]:_(<4 x s32>), [[UV13:%[0-9]+]]:_(<4 x s32>), [[UV14:%[0-9]+]]:_(<4 x s32>), [[UV15:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) ; CHECK-NEXT: G_STORE [[UV]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16 ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) @@ -243,10 +241,8 @@ body: | ; CHECK-NEXT: [[C14:%[0-9]+]]:_(s64) = G_CONSTANT i64 240 ; CHECK-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C14]](s64) ; CHECK-NEXT: G_STORE [[UV15]](<4 x s32>), [[PTR_ADD14]](p1) :: (store (<4 x s32>) into unknown-address + 240, align 4, addrspace 1) - ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(<4 x s32>), [[UV17:%[0-9]+]]:_(<4 x s32>), [[UV18:%[0-9]+]]:_(<4 x s32>), [[UV19:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV20:%[0-9]+]]:_(<4 x s32>), [[UV21:%[0-9]+]]:_(<4 x s32>), [[UV22:%[0-9]+]]:_(<4 x s32>), [[UV23:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV24:%[0-9]+]]:_(<4 x s32>), [[UV25:%[0-9]+]]:_(<4 x s32>), [[UV26:%[0-9]+]]:_(<4 x s32>), [[UV27:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV28:%[0-9]+]]:_(<4 x s32>), [[UV29:%[0-9]+]]:_(<4 x s32>), [[UV30:%[0-9]+]]:_(<4 x s32>), [[UV31:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) + ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(<4 x s32>), [[UV17:%[0-9]+]]:_(<4 x s32>), [[UV18:%[0-9]+]]:_(<4 x s32>), [[UV19:%[0-9]+]]:_(<4 x s32>), [[UV20:%[0-9]+]]:_(<4 x s32>), [[UV21:%[0-9]+]]:_(<4 x s32>), [[UV22:%[0-9]+]]:_(<4 x s32>), [[UV23:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) + ; CHECK-NEXT: [[UV24:%[0-9]+]]:_(<4 x s32>), [[UV25:%[0-9]+]]:_(<4 x s32>), [[UV26:%[0-9]+]]:_(<4 x s32>), [[UV27:%[0-9]+]]:_(<4 x s32>), [[UV28:%[0-9]+]]:_(<4 x s32>), [[UV29:%[0-9]+]]:_(<4 x s32>), [[UV30:%[0-9]+]]:_(<4 x s32>), [[UV31:%[0-9]+]]:_(<4 x s32>) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) ; CHECK-NEXT: G_STORE [[UV16]](<4 x s32>), [[COPY1]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; CHECK-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY1]], [[C]](s64) ; CHECK-NEXT: G_STORE [[UV17]](<4 x s32>), [[PTR_ADD15]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1) diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-phi.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-phi.mir index 00612d552a104..d82e8328f26ec 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-phi.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-phi.mir @@ -673,88 +673,86 @@ body: | ; CHECK-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000) ; CHECK-NEXT: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4 ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<16 x s32>) = G_IMPLICIT_DEF + ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<32 x s32>) = G_IMPLICIT_DEF ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr4 ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[COPY]](s32), [[C]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<16 x s32>), [[UV1:%[0-9]+]]:_(<16 x s32>) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<16 x s32>), [[UV3:%[0-9]+]]:_(<16 x s32>) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) ; CHECK-NEXT: G_BRCOND [[ICMP]](s1), %bb.1 ; CHECK-NEXT: G_BR %bb.2 ; CHECK-NEXT: {{ $}} ; CHECK-NEXT: bb.1: ; CHECK-NEXT: successors: %bb.2(0x80000000) ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV32:%[0-9]+]]:_(s32), [[UV33:%[0-9]+]]:_(s32), [[UV34:%[0-9]+]]:_(s32), [[UV35:%[0-9]+]]:_(s32), [[UV36:%[0-9]+]]:_(s32), [[UV37:%[0-9]+]]:_(s32), [[UV38:%[0-9]+]]:_(s32), [[UV39:%[0-9]+]]:_(s32), [[UV40:%[0-9]+]]:_(s32), [[UV41:%[0-9]+]]:_(s32), [[UV42:%[0-9]+]]:_(s32), [[UV43:%[0-9]+]]:_(s32), [[UV44:%[0-9]+]]:_(s32), [[UV45:%[0-9]+]]:_(s32), [[UV46:%[0-9]+]]:_(s32), [[UV47:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV48:%[0-9]+]]:_(s32), [[UV49:%[0-9]+]]:_(s32), [[UV50:%[0-9]+]]:_(s32), [[UV51:%[0-9]+]]:_(s32), [[UV52:%[0-9]+]]:_(s32), [[UV53:%[0-9]+]]:_(s32), [[UV54:%[0-9]+]]:_(s32), [[UV55:%[0-9]+]]:_(s32), [[UV56:%[0-9]+]]:_(s32), [[UV57:%[0-9]+]]:_(s32), [[UV58:%[0-9]+]]:_(s32), [[UV59:%[0-9]+]]:_(s32), [[UV60:%[0-9]+]]:_(s32), [[UV61:%[0-9]+]]:_(s32), [[UV62:%[0-9]+]]:_(s32), [[UV63:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV64:%[0-9]+]]:_(s32), [[UV65:%[0-9]+]]:_(s32), [[UV66:%[0-9]+]]:_(s32), [[UV67:%[0-9]+]]:_(s32), [[UV68:%[0-9]+]]:_(s32), [[UV69:%[0-9]+]]:_(s32), [[UV70:%[0-9]+]]:_(s32), [[UV71:%[0-9]+]]:_(s32), [[UV72:%[0-9]+]]:_(s32), [[UV73:%[0-9]+]]:_(s32), [[UV74:%[0-9]+]]:_(s32), [[UV75:%[0-9]+]]:_(s32), [[UV76:%[0-9]+]]:_(s32), [[UV77:%[0-9]+]]:_(s32), [[UV78:%[0-9]+]]:_(s32), [[UV79:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV80:%[0-9]+]]:_(s32), [[UV81:%[0-9]+]]:_(s32), [[UV82:%[0-9]+]]:_(s32), [[UV83:%[0-9]+]]:_(s32), [[UV84:%[0-9]+]]:_(s32), [[UV85:%[0-9]+]]:_(s32), [[UV86:%[0-9]+]]:_(s32), [[UV87:%[0-9]+]]:_(s32), [[UV88:%[0-9]+]]:_(s32), [[UV89:%[0-9]+]]:_(s32), [[UV90:%[0-9]+]]:_(s32), [[UV91:%[0-9]+]]:_(s32), [[UV92:%[0-9]+]]:_(s32), [[UV93:%[0-9]+]]:_(s32), [[UV94:%[0-9]+]]:_(s32), [[UV95:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV96:%[0-9]+]]:_(s32), [[UV97:%[0-9]+]]:_(s32), [[UV98:%[0-9]+]]:_(s32), [[UV99:%[0-9]+]]:_(s32), [[UV100:%[0-9]+]]:_(s32), [[UV101:%[0-9]+]]:_(s32), [[UV102:%[0-9]+]]:_(s32), [[UV103:%[0-9]+]]:_(s32), [[UV104:%[0-9]+]]:_(s32), [[UV105:%[0-9]+]]:_(s32), [[UV106:%[0-9]+]]:_(s32), [[UV107:%[0-9]+]]:_(s32), [[UV108:%[0-9]+]]:_(s32), [[UV109:%[0-9]+]]:_(s32), [[UV110:%[0-9]+]]:_(s32), [[UV111:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[UV112:%[0-9]+]]:_(s32), [[UV113:%[0-9]+]]:_(s32), [[UV114:%[0-9]+]]:_(s32), [[UV115:%[0-9]+]]:_(s32), [[UV116:%[0-9]+]]:_(s32), [[UV117:%[0-9]+]]:_(s32), [[UV118:%[0-9]+]]:_(s32), [[UV119:%[0-9]+]]:_(s32), [[UV120:%[0-9]+]]:_(s32), [[UV121:%[0-9]+]]:_(s32), [[UV122:%[0-9]+]]:_(s32), [[UV123:%[0-9]+]]:_(s32), [[UV124:%[0-9]+]]:_(s32), [[UV125:%[0-9]+]]:_(s32), [[UV126:%[0-9]+]]:_(s32), [[UV127:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>) - ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[UV]], [[UV64]] - ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[UV1]], [[UV65]] - ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s32) = G_ADD [[UV2]], [[UV66]] - ; CHECK-NEXT: [[ADD3:%[0-9]+]]:_(s32) = G_ADD [[UV3]], [[UV67]] - ; CHECK-NEXT: [[ADD4:%[0-9]+]]:_(s32) = G_ADD [[UV4]], [[UV68]] - ; CHECK-NEXT: [[ADD5:%[0-9]+]]:_(s32) = G_ADD [[UV5]], [[UV69]] - ; CHECK-NEXT: [[ADD6:%[0-9]+]]:_(s32) = G_ADD [[UV6]], [[UV70]] - ; CHECK-NEXT: [[ADD7:%[0-9]+]]:_(s32) = G_ADD [[UV7]], [[UV71]] - ; CHECK-NEXT: [[ADD8:%[0-9]+]]:_(s32) = G_ADD [[UV8]], [[UV72]] - ; CHECK-NEXT: [[ADD9:%[0-9]+]]:_(s32) = G_ADD [[UV9]], [[UV73]] - ; CHECK-NEXT: [[ADD10:%[0-9]+]]:_(s32) = G_ADD [[UV10]], [[UV74]] - ; CHECK-NEXT: [[ADD11:%[0-9]+]]:_(s32) = G_ADD [[UV11]], [[UV75]] - ; CHECK-NEXT: [[ADD12:%[0-9]+]]:_(s32) = G_ADD [[UV12]], [[UV76]] - ; CHECK-NEXT: [[ADD13:%[0-9]+]]:_(s32) = G_ADD [[UV13]], [[UV77]] - ; CHECK-NEXT: [[ADD14:%[0-9]+]]:_(s32) = G_ADD [[UV14]], [[UV78]] - ; CHECK-NEXT: [[ADD15:%[0-9]+]]:_(s32) = G_ADD [[UV15]], [[UV79]] - ; CHECK-NEXT: [[ADD16:%[0-9]+]]:_(s32) = G_ADD [[UV16]], [[UV80]] - ; CHECK-NEXT: [[ADD17:%[0-9]+]]:_(s32) = G_ADD [[UV17]], [[UV81]] - ; CHECK-NEXT: [[ADD18:%[0-9]+]]:_(s32) = G_ADD [[UV18]], [[UV82]] - ; CHECK-NEXT: [[ADD19:%[0-9]+]]:_(s32) = G_ADD [[UV19]], [[UV83]] - ; CHECK-NEXT: [[ADD20:%[0-9]+]]:_(s32) = G_ADD [[UV20]], [[UV84]] - ; CHECK-NEXT: [[ADD21:%[0-9]+]]:_(s32) = G_ADD [[UV21]], [[UV85]] - ; CHECK-NEXT: [[ADD22:%[0-9]+]]:_(s32) = G_ADD [[UV22]], [[UV86]] - ; CHECK-NEXT: [[ADD23:%[0-9]+]]:_(s32) = G_ADD [[UV23]], [[UV87]] - ; CHECK-NEXT: [[ADD24:%[0-9]+]]:_(s32) = G_ADD [[UV24]], [[UV88]] - ; CHECK-NEXT: [[ADD25:%[0-9]+]]:_(s32) = G_ADD [[UV25]], [[UV89]] - ; CHECK-NEXT: [[ADD26:%[0-9]+]]:_(s32) = G_ADD [[UV26]], [[UV90]] - ; CHECK-NEXT: [[ADD27:%[0-9]+]]:_(s32) = G_ADD [[UV27]], [[UV91]] - ; CHECK-NEXT: [[ADD28:%[0-9]+]]:_(s32) = G_ADD [[UV28]], [[UV92]] - ; CHECK-NEXT: [[ADD29:%[0-9]+]]:_(s32) = G_ADD [[UV29]], [[UV93]] - ; CHECK-NEXT: [[ADD30:%[0-9]+]]:_(s32) = G_ADD [[UV30]], [[UV94]] - ; CHECK-NEXT: [[ADD31:%[0-9]+]]:_(s32) = G_ADD [[UV31]], [[UV95]] - ; CHECK-NEXT: [[ADD32:%[0-9]+]]:_(s32) = G_ADD [[UV32]], [[UV96]] - ; CHECK-NEXT: [[ADD33:%[0-9]+]]:_(s32) = G_ADD [[UV33]], [[UV97]] - ; CHECK-NEXT: [[ADD34:%[0-9]+]]:_(s32) = G_ADD [[UV34]], [[UV98]] - ; CHECK-NEXT: [[ADD35:%[0-9]+]]:_(s32) = G_ADD [[UV35]], [[UV99]] - ; CHECK-NEXT: [[ADD36:%[0-9]+]]:_(s32) = G_ADD [[UV36]], [[UV100]] - ; CHECK-NEXT: [[ADD37:%[0-9]+]]:_(s32) = G_ADD [[UV37]], [[UV101]] - ; CHECK-NEXT: [[ADD38:%[0-9]+]]:_(s32) = G_ADD [[UV38]], [[UV102]] - ; CHECK-NEXT: [[ADD39:%[0-9]+]]:_(s32) = G_ADD [[UV39]], [[UV103]] - ; CHECK-NEXT: [[ADD40:%[0-9]+]]:_(s32) = G_ADD [[UV40]], [[UV104]] - ; CHECK-NEXT: [[ADD41:%[0-9]+]]:_(s32) = G_ADD [[UV41]], [[UV105]] - ; CHECK-NEXT: [[ADD42:%[0-9]+]]:_(s32) = G_ADD [[UV42]], [[UV106]] - ; CHECK-NEXT: [[ADD43:%[0-9]+]]:_(s32) = G_ADD [[UV43]], [[UV107]] - ; CHECK-NEXT: [[ADD44:%[0-9]+]]:_(s32) = G_ADD [[UV44]], [[UV108]] - ; CHECK-NEXT: [[ADD45:%[0-9]+]]:_(s32) = G_ADD [[UV45]], [[UV109]] - ; CHECK-NEXT: [[ADD46:%[0-9]+]]:_(s32) = G_ADD [[UV46]], [[UV110]] - ; CHECK-NEXT: [[ADD47:%[0-9]+]]:_(s32) = G_ADD [[UV47]], [[UV111]] - ; CHECK-NEXT: [[ADD48:%[0-9]+]]:_(s32) = G_ADD [[UV48]], [[UV112]] - ; CHECK-NEXT: [[ADD49:%[0-9]+]]:_(s32) = G_ADD [[UV49]], [[UV113]] - ; CHECK-NEXT: [[ADD50:%[0-9]+]]:_(s32) = G_ADD [[UV50]], [[UV114]] - ; CHECK-NEXT: [[ADD51:%[0-9]+]]:_(s32) = G_ADD [[UV51]], [[UV115]] - ; CHECK-NEXT: [[ADD52:%[0-9]+]]:_(s32) = G_ADD [[UV52]], [[UV116]] - ; CHECK-NEXT: [[ADD53:%[0-9]+]]:_(s32) = G_ADD [[UV53]], [[UV117]] - ; CHECK-NEXT: [[ADD54:%[0-9]+]]:_(s32) = G_ADD [[UV54]], [[UV118]] - ; CHECK-NEXT: [[ADD55:%[0-9]+]]:_(s32) = G_ADD [[UV55]], [[UV119]] - ; CHECK-NEXT: [[ADD56:%[0-9]+]]:_(s32) = G_ADD [[UV56]], [[UV120]] - ; CHECK-NEXT: [[ADD57:%[0-9]+]]:_(s32) = G_ADD [[UV57]], [[UV121]] - ; CHECK-NEXT: [[ADD58:%[0-9]+]]:_(s32) = G_ADD [[UV58]], [[UV122]] - ; CHECK-NEXT: [[ADD59:%[0-9]+]]:_(s32) = G_ADD [[UV59]], [[UV123]] - ; CHECK-NEXT: [[ADD60:%[0-9]+]]:_(s32) = G_ADD [[UV60]], [[UV124]] - ; CHECK-NEXT: [[ADD61:%[0-9]+]]:_(s32) = G_ADD [[UV61]], [[UV125]] - ; CHECK-NEXT: [[ADD62:%[0-9]+]]:_(s32) = G_ADD [[UV62]], [[UV126]] - ; CHECK-NEXT: [[ADD63:%[0-9]+]]:_(s32) = G_ADD [[UV63]], [[UV127]] + ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32), [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32), [[UV32:%[0-9]+]]:_(s32), [[UV33:%[0-9]+]]:_(s32), [[UV34:%[0-9]+]]:_(s32), [[UV35:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) + ; CHECK-NEXT: [[UV36:%[0-9]+]]:_(s32), [[UV37:%[0-9]+]]:_(s32), [[UV38:%[0-9]+]]:_(s32), [[UV39:%[0-9]+]]:_(s32), [[UV40:%[0-9]+]]:_(s32), [[UV41:%[0-9]+]]:_(s32), [[UV42:%[0-9]+]]:_(s32), [[UV43:%[0-9]+]]:_(s32), [[UV44:%[0-9]+]]:_(s32), [[UV45:%[0-9]+]]:_(s32), [[UV46:%[0-9]+]]:_(s32), [[UV47:%[0-9]+]]:_(s32), [[UV48:%[0-9]+]]:_(s32), [[UV49:%[0-9]+]]:_(s32), [[UV50:%[0-9]+]]:_(s32), [[UV51:%[0-9]+]]:_(s32), [[UV52:%[0-9]+]]:_(s32), [[UV53:%[0-9]+]]:_(s32), [[UV54:%[0-9]+]]:_(s32), [[UV55:%[0-9]+]]:_(s32), [[UV56:%[0-9]+]]:_(s32), [[UV57:%[0-9]+]]:_(s32), [[UV58:%[0-9]+]]:_(s32), [[UV59:%[0-9]+]]:_(s32), [[UV60:%[0-9]+]]:_(s32), [[UV61:%[0-9]+]]:_(s32), [[UV62:%[0-9]+]]:_(s32), [[UV63:%[0-9]+]]:_(s32), [[UV64:%[0-9]+]]:_(s32), [[UV65:%[0-9]+]]:_(s32), [[UV66:%[0-9]+]]:_(s32), [[UV67:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) + ; CHECK-NEXT: [[UV68:%[0-9]+]]:_(s32), [[UV69:%[0-9]+]]:_(s32), [[UV70:%[0-9]+]]:_(s32), [[UV71:%[0-9]+]]:_(s32), [[UV72:%[0-9]+]]:_(s32), [[UV73:%[0-9]+]]:_(s32), [[UV74:%[0-9]+]]:_(s32), [[UV75:%[0-9]+]]:_(s32), [[UV76:%[0-9]+]]:_(s32), [[UV77:%[0-9]+]]:_(s32), [[UV78:%[0-9]+]]:_(s32), [[UV79:%[0-9]+]]:_(s32), [[UV80:%[0-9]+]]:_(s32), [[UV81:%[0-9]+]]:_(s32), [[UV82:%[0-9]+]]:_(s32), [[UV83:%[0-9]+]]:_(s32), [[UV84:%[0-9]+]]:_(s32), [[UV85:%[0-9]+]]:_(s32), [[UV86:%[0-9]+]]:_(s32), [[UV87:%[0-9]+]]:_(s32), [[UV88:%[0-9]+]]:_(s32), [[UV89:%[0-9]+]]:_(s32), [[UV90:%[0-9]+]]:_(s32), [[UV91:%[0-9]+]]:_(s32), [[UV92:%[0-9]+]]:_(s32), [[UV93:%[0-9]+]]:_(s32), [[UV94:%[0-9]+]]:_(s32), [[UV95:%[0-9]+]]:_(s32), [[UV96:%[0-9]+]]:_(s32), [[UV97:%[0-9]+]]:_(s32), [[UV98:%[0-9]+]]:_(s32), [[UV99:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) + ; CHECK-NEXT: [[UV100:%[0-9]+]]:_(s32), [[UV101:%[0-9]+]]:_(s32), [[UV102:%[0-9]+]]:_(s32), [[UV103:%[0-9]+]]:_(s32), [[UV104:%[0-9]+]]:_(s32), [[UV105:%[0-9]+]]:_(s32), [[UV106:%[0-9]+]]:_(s32), [[UV107:%[0-9]+]]:_(s32), [[UV108:%[0-9]+]]:_(s32), [[UV109:%[0-9]+]]:_(s32), [[UV110:%[0-9]+]]:_(s32), [[UV111:%[0-9]+]]:_(s32), [[UV112:%[0-9]+]]:_(s32), [[UV113:%[0-9]+]]:_(s32), [[UV114:%[0-9]+]]:_(s32), [[UV115:%[0-9]+]]:_(s32), [[UV116:%[0-9]+]]:_(s32), [[UV117:%[0-9]+]]:_(s32), [[UV118:%[0-9]+]]:_(s32), [[UV119:%[0-9]+]]:_(s32), [[UV120:%[0-9]+]]:_(s32), [[UV121:%[0-9]+]]:_(s32), [[UV122:%[0-9]+]]:_(s32), [[UV123:%[0-9]+]]:_(s32), [[UV124:%[0-9]+]]:_(s32), [[UV125:%[0-9]+]]:_(s32), [[UV126:%[0-9]+]]:_(s32), [[UV127:%[0-9]+]]:_(s32), [[UV128:%[0-9]+]]:_(s32), [[UV129:%[0-9]+]]:_(s32), [[UV130:%[0-9]+]]:_(s32), [[UV131:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<32 x s32>) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[UV4]], [[UV68]] + ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[UV5]], [[UV69]] + ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s32) = G_ADD [[UV6]], [[UV70]] + ; CHECK-NEXT: [[ADD3:%[0-9]+]]:_(s32) = G_ADD [[UV7]], [[UV71]] + ; CHECK-NEXT: [[ADD4:%[0-9]+]]:_(s32) = G_ADD [[UV8]], [[UV72]] + ; CHECK-NEXT: [[ADD5:%[0-9]+]]:_(s32) = G_ADD [[UV9]], [[UV73]] + ; CHECK-NEXT: [[ADD6:%[0-9]+]]:_(s32) = G_ADD [[UV10]], [[UV74]] + ; CHECK-NEXT: [[ADD7:%[0-9]+]]:_(s32) = G_ADD [[UV11]], [[UV75]] + ; CHECK-NEXT: [[ADD8:%[0-9]+]]:_(s32) = G_ADD [[UV12]], [[UV76]] + ; CHECK-NEXT: [[ADD9:%[0-9]+]]:_(s32) = G_ADD [[UV13]], [[UV77]] + ; CHECK-NEXT: [[ADD10:%[0-9]+]]:_(s32) = G_ADD [[UV14]], [[UV78]] + ; CHECK-NEXT: [[ADD11:%[0-9]+]]:_(s32) = G_ADD [[UV15]], [[UV79]] + ; CHECK-NEXT: [[ADD12:%[0-9]+]]:_(s32) = G_ADD [[UV16]], [[UV80]] + ; CHECK-NEXT: [[ADD13:%[0-9]+]]:_(s32) = G_ADD [[UV17]], [[UV81]] + ; CHECK-NEXT: [[ADD14:%[0-9]+]]:_(s32) = G_ADD [[UV18]], [[UV82]] + ; CHECK-NEXT: [[ADD15:%[0-9]+]]:_(s32) = G_ADD [[UV19]], [[UV83]] + ; CHECK-NEXT: [[ADD16:%[0-9]+]]:_(s32) = G_ADD [[UV20]], [[UV84]] + ; CHECK-NEXT: [[ADD17:%[0-9]+]]:_(s32) = G_ADD [[UV21]], [[UV85]] + ; CHECK-NEXT: [[ADD18:%[0-9]+]]:_(s32) = G_ADD [[UV22]], [[UV86]] + ; CHECK-NEXT: [[ADD19:%[0-9]+]]:_(s32) = G_ADD [[UV23]], [[UV87]] + ; CHECK-NEXT: [[ADD20:%[0-9]+]]:_(s32) = G_ADD [[UV24]], [[UV88]] + ; CHECK-NEXT: [[ADD21:%[0-9]+]]:_(s32) = G_ADD [[UV25]], [[UV89]] + ; CHECK-NEXT: [[ADD22:%[0-9]+]]:_(s32) = G_ADD [[UV26]], [[UV90]] + ; CHECK-NEXT: [[ADD23:%[0-9]+]]:_(s32) = G_ADD [[UV27]], [[UV91]] + ; CHECK-NEXT: [[ADD24:%[0-9]+]]:_(s32) = G_ADD [[UV28]], [[UV92]] + ; CHECK-NEXT: [[ADD25:%[0-9]+]]:_(s32) = G_ADD [[UV29]], [[UV93]] + ; CHECK-NEXT: [[ADD26:%[0-9]+]]:_(s32) = G_ADD [[UV30]], [[UV94]] + ; CHECK-NEXT: [[ADD27:%[0-9]+]]:_(s32) = G_ADD [[UV31]], [[UV95]] + ; CHECK-NEXT: [[ADD28:%[0-9]+]]:_(s32) = G_ADD [[UV32]], [[UV96]] + ; CHECK-NEXT: [[ADD29:%[0-9]+]]:_(s32) = G_ADD [[UV33]], [[UV97]] + ; CHECK-NEXT: [[ADD30:%[0-9]+]]:_(s32) = G_ADD [[UV34]], [[UV98]] + ; CHECK-NEXT: [[ADD31:%[0-9]+]]:_(s32) = G_ADD [[UV35]], [[UV99]] + ; CHECK-NEXT: [[ADD32:%[0-9]+]]:_(s32) = G_ADD [[UV36]], [[UV100]] + ; CHECK-NEXT: [[ADD33:%[0-9]+]]:_(s32) = G_ADD [[UV37]], [[UV101]] + ; CHECK-NEXT: [[ADD34:%[0-9]+]]:_(s32) = G_ADD [[UV38]], [[UV102]] + ; CHECK-NEXT: [[ADD35:%[0-9]+]]:_(s32) = G_ADD [[UV39]], [[UV103]] + ; CHECK-NEXT: [[ADD36:%[0-9]+]]:_(s32) = G_ADD [[UV40]], [[UV104]] + ; CHECK-NEXT: [[ADD37:%[0-9]+]]:_(s32) = G_ADD [[UV41]], [[UV105]] + ; CHECK-NEXT: [[ADD38:%[0-9]+]]:_(s32) = G_ADD [[UV42]], [[UV106]] + ; CHECK-NEXT: [[ADD39:%[0-9]+]]:_(s32) = G_ADD [[UV43]], [[UV107]] + ; CHECK-NEXT: [[ADD40:%[0-9]+]]:_(s32) = G_ADD [[UV44]], [[UV108]] + ; CHECK-NEXT: [[ADD41:%[0-9]+]]:_(s32) = G_ADD [[UV45]], [[UV109]] + ; CHECK-NEXT: [[ADD42:%[0-9]+]]:_(s32) = G_ADD [[UV46]], [[UV110]] + ; CHECK-NEXT: [[ADD43:%[0-9]+]]:_(s32) = G_ADD [[UV47]], [[UV111]] + ; CHECK-NEXT: [[ADD44:%[0-9]+]]:_(s32) = G_ADD [[UV48]], [[UV112]] + ; CHECK-NEXT: [[ADD45:%[0-9]+]]:_(s32) = G_ADD [[UV49]], [[UV113]] + ; CHECK-NEXT: [[ADD46:%[0-9]+]]:_(s32) = G_ADD [[UV50]], [[UV114]] + ; CHECK-NEXT: [[ADD47:%[0-9]+]]:_(s32) = G_ADD [[UV51]], [[UV115]] + ; CHECK-NEXT: [[ADD48:%[0-9]+]]:_(s32) = G_ADD [[UV52]], [[UV116]] + ; CHECK-NEXT: [[ADD49:%[0-9]+]]:_(s32) = G_ADD [[UV53]], [[UV117]] + ; CHECK-NEXT: [[ADD50:%[0-9]+]]:_(s32) = G_ADD [[UV54]], [[UV118]] + ; CHECK-NEXT: [[ADD51:%[0-9]+]]:_(s32) = G_ADD [[UV55]], [[UV119]] + ; CHECK-NEXT: [[ADD52:%[0-9]+]]:_(s32) = G_ADD [[UV56]], [[UV120]] + ; CHECK-NEXT: [[ADD53:%[0-9]+]]:_(s32) = G_ADD [[UV57]], [[UV121]] + ; CHECK-NEXT: [[ADD54:%[0-9]+]]:_(s32) = G_ADD [[UV58]], [[UV122]] + ; CHECK-NEXT: [[ADD55:%[0-9]+]]:_(s32) = G_ADD [[UV59]], [[UV123]] + ; CHECK-NEXT: [[ADD56:%[0-9]+]]:_(s32) = G_ADD [[UV60]], [[UV124]] + ; CHECK-NEXT: [[ADD57:%[0-9]+]]:_(s32) = G_ADD [[UV61]], [[UV125]] + ; CHECK-NEXT: [[ADD58:%[0-9]+]]:_(s32) = G_ADD [[UV62]], [[UV126]] + ; CHECK-NEXT: [[ADD59:%[0-9]+]]:_(s32) = G_ADD [[UV63]], [[UV127]] + ; CHECK-NEXT: [[ADD60:%[0-9]+]]:_(s32) = G_ADD [[UV64]], [[UV128]] + ; CHECK-NEXT: [[ADD61:%[0-9]+]]:_(s32) = G_ADD [[UV65]], [[UV129]] + ; CHECK-NEXT: [[ADD62:%[0-9]+]]:_(s32) = G_ADD [[UV66]], [[UV130]] + ; CHECK-NEXT: [[ADD63:%[0-9]+]]:_(s32) = G_ADD [[UV67]], [[UV131]] ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<16 x s32>) = G_BUILD_VECTOR [[ADD]](s32), [[ADD1]](s32), [[ADD2]](s32), [[ADD3]](s32), [[ADD4]](s32), [[ADD5]](s32), [[ADD6]](s32), [[ADD7]](s32), [[ADD8]](s32), [[ADD9]](s32), [[ADD10]](s32), [[ADD11]](s32), [[ADD12]](s32), [[ADD13]](s32), [[ADD14]](s32), [[ADD15]](s32) ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<16 x s32>) = G_BUILD_VECTOR [[ADD16]](s32), [[ADD17]](s32), [[ADD18]](s32), [[ADD19]](s32), [[ADD20]](s32), [[ADD21]](s32), [[ADD22]](s32), [[ADD23]](s32), [[ADD24]](s32), [[ADD25]](s32), [[ADD26]](s32), [[ADD27]](s32), [[ADD28]](s32), [[ADD29]](s32), [[ADD30]](s32), [[ADD31]](s32) ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<16 x s32>) = G_BUILD_VECTOR [[ADD32]](s32), [[ADD33]](s32), [[ADD34]](s32), [[ADD35]](s32), [[ADD36]](s32), [[ADD37]](s32), [[ADD38]](s32), [[ADD39]](s32), [[ADD40]](s32), [[ADD41]](s32), [[ADD42]](s32), [[ADD43]](s32), [[ADD44]](s32), [[ADD45]](s32), [[ADD46]](s32), [[ADD47]](s32) @@ -762,10 +760,10 @@ body: | ; CHECK-NEXT: G_BR %bb.2 ; CHECK-NEXT: {{ $}} ; CHECK-NEXT: bb.2: - ; CHECK-NEXT: [[PHI:%[0-9]+]]:_(<16 x s32>) = G_PHI [[DEF]](<16 x s32>), %bb.0, [[BUILD_VECTOR]](<16 x s32>), %bb.1 - ; CHECK-NEXT: [[PHI1:%[0-9]+]]:_(<16 x s32>) = G_PHI [[DEF]](<16 x s32>), %bb.0, [[BUILD_VECTOR1]](<16 x s32>), %bb.1 - ; CHECK-NEXT: [[PHI2:%[0-9]+]]:_(<16 x s32>) = G_PHI [[DEF]](<16 x s32>), %bb.0, [[BUILD_VECTOR2]](<16 x s32>), %bb.1 - ; CHECK-NEXT: [[PHI3:%[0-9]+]]:_(<16 x s32>) = G_PHI [[DEF]](<16 x s32>), %bb.0, [[BUILD_VECTOR3]](<16 x s32>), %bb.1 + ; CHECK-NEXT: [[PHI:%[0-9]+]]:_(<16 x s32>) = G_PHI [[UV]](<16 x s32>), %bb.0, [[BUILD_VECTOR]](<16 x s32>), %bb.1 + ; CHECK-NEXT: [[PHI1:%[0-9]+]]:_(<16 x s32>) = G_PHI [[UV1]](<16 x s32>), %bb.0, [[BUILD_VECTOR1]](<16 x s32>), %bb.1 + ; CHECK-NEXT: [[PHI2:%[0-9]+]]:_(<16 x s32>) = G_PHI [[UV2]](<16 x s32>), %bb.0, [[BUILD_VECTOR2]](<16 x s32>), %bb.1 + ; CHECK-NEXT: [[PHI3:%[0-9]+]]:_(<16 x s32>) = G_PHI [[UV3]](<16 x s32>), %bb.0, [[BUILD_VECTOR3]](<16 x s32>), %bb.1 ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<64 x s32>) = G_CONCAT_VECTORS [[PHI]](<16 x s32>), [[PHI1]](<16 x s32>), [[PHI2]](<16 x s32>), [[PHI3]](<16 x s32>) ; CHECK-NEXT: S_SETPC_B64 undef $sgpr30_sgpr31, implicit [[CONCAT_VECTORS]](<64 x s32>) bb.0: diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect.mir index c50187f594901..1565986516860 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect.mir @@ -42,8 +42,6 @@ ret void } - define void @non_power_of_2() { ret void } - define amdgpu_kernel void @load_constant_v4i16_from_8_align8(ptr addrspace(4) %ptr0) { ret void } @@ -186,23 +184,6 @@ body: | %1:_(s32) = G_LOAD %0 :: (load (s32) from %ir.tmp1) ... ---- -name: non_power_of_2 -legalized: true - -body: | - bb.0: - ; CHECK-LABEL: name: non_power_of_2 - ; CHECK: [[DEF:%[0-9]+]]:sgpr(s448) = G_IMPLICIT_DEF - ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:sgpr(s32) = G_EXTRACT [[DEF]](s448), 0 - ; CHECK-NEXT: $sgpr0 = COPY [[EXTRACT]](s32) - ; CHECK-NEXT: SI_RETURN_TO_EPILOG $sgpr0 - %0:_(s448) = G_IMPLICIT_DEF - %1:_(s32) = G_EXTRACT %0:_(s448), 0 - $sgpr0 = COPY %1:_(s32) - SI_RETURN_TO_EPILOG $sgpr0 -... - --- name: load_constant_v4i16_from_8_align8 legalized: true diff --git a/llvm/test/CodeGen/AMDGPU/freeze.ll b/llvm/test/CodeGen/AMDGPU/freeze.ll new file mode 100644 index 0000000000000..22427ee344d91 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/freeze.ll @@ -0,0 +1,1856 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc -global-isel=0 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GFX10,GFX10-SDAG %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GFX10,GFX10-GISEL %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefixes=GFX11,GFX11-SDAG %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefixes=GFX11,GFX11-GISEL %s + +define void @freeze_v2i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-LABEL: freeze_v2i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: freeze_v2i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_store_b64 v[2:3], v[0:1], off +; GFX11-NEXT: s_setpc_b64 s[30:31] + %a = load <2 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <2 x i32> %a + store <2 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v3i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-LABEL: freeze_v3i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: global_load_dwordx3 v[4:6], v[0:1], off +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: global_store_dwordx3 v[2:3], v[4:6], off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: freeze_v3i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: global_load_b96 v[4:6], v[0:1], off +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_store_b96 v[2:3], v[4:6], off +; GFX11-NEXT: s_setpc_b64 s[30:31] + %a = load <3 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <3 x i32> %a + store <3 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v4i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-LABEL: freeze_v4i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: freeze_v4i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-NEXT: s_setpc_b64 s[30:31] + %a = load <4 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <4 x i32> %a + store <4 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v5i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v5i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x1 +; GFX10-SDAG-NEXT: global_load_dword v8, v[0:1], off offset:16 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dword v[2:3], v8, off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v5i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x1 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dword v8, v[0:1], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dword v[2:3], v8, off offset:16 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v5i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x1 +; GFX11-SDAG-NEXT: global_load_b32 v8, v[0:1], off offset:16 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b32 v[2:3], v8, off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v5i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x1 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b32 v0, v[0:1], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b32 v[2:3], v0, off offset:16 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <5 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <5 x i32> %a + store <5 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v6i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v6i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x1 +; GFX10-SDAG-NEXT: global_load_dwordx2 v[8:9], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx2 v[2:3], v[8:9], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v6i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x1 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx2 v[8:9], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx2 v[2:3], v[8:9], off offset:16 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v6i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x1 +; GFX11-SDAG-NEXT: global_load_b64 v[8:9], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b64 v[2:3], v[8:9], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v6i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x1 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b64 v[0:1], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b64 v[2:3], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <6 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <6 x i32> %a + store <6 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v7i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v7i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x1 +; GFX10-SDAG-NEXT: global_load_dwordx3 v[8:10], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx3 v[2:3], v[8:10], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v7i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x1 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx3 v[8:10], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx3 v[2:3], v[8:10], off offset:16 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v7i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x1 +; GFX11-SDAG-NEXT: global_load_b96 v[8:10], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b96 v[2:3], v[8:10], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v7i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x1 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b96 v[8:10], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b96 v[2:3], v[8:10], off offset:16 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <7 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <7 x i32> %a + store <7 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v8i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v8i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x1 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v8i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x1 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v8i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x1 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v8i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x1 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <8 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <8 x i32> %a + store <8 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v9i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v9i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x2 +; GFX10-SDAG-NEXT: global_load_dword v12, v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dword v[2:3], v12, off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v9i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x2 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dword v12, v[0:1], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dword v[2:3], v12, off offset:32 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v9i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x2 +; GFX11-SDAG-NEXT: global_load_b32 v12, v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b32 v[2:3], v12, off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v9i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x2 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b32 v0, v[0:1], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b32 v[2:3], v0, off offset:32 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <9 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <9 x i32> %a + store <9 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v10i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-LABEL: freeze_v10i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_clause 0x2 +; GFX10-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-NEXT: global_load_dwordx2 v[12:13], v[0:1], off offset:32 +; GFX10-NEXT: s_waitcnt vmcnt(2) +; GFX10-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-NEXT: s_waitcnt vmcnt(1) +; GFX10-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: global_store_dwordx2 v[2:3], v[12:13], off offset:32 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: freeze_v10i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_clause 0x2 +; GFX11-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off offset:32 +; GFX11-NEXT: s_waitcnt vmcnt(2) +; GFX11-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_store_b64 v[2:3], v[0:1], off offset:32 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %a = load <10 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <10 x i32> %a + store <10 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v11i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-LABEL: freeze_v11i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_clause 0x2 +; GFX10-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-NEXT: global_load_dwordx3 v[12:14], v[0:1], off offset:32 +; GFX10-NEXT: s_waitcnt vmcnt(2) +; GFX10-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-NEXT: s_waitcnt vmcnt(1) +; GFX10-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: global_store_dwordx3 v[2:3], v[12:14], off offset:32 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: freeze_v11i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_clause 0x2 +; GFX11-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-NEXT: global_load_b96 v[12:14], v[0:1], off offset:32 +; GFX11-NEXT: s_waitcnt vmcnt(2) +; GFX11-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_store_b96 v[2:3], v[12:14], off offset:32 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %a = load <11 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <11 x i32> %a + store <11 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v12i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-LABEL: freeze_v12i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_clause 0x2 +; GFX10-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-NEXT: s_waitcnt vmcnt(2) +; GFX10-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-NEXT: s_waitcnt vmcnt(1) +; GFX10-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: freeze_v12i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_clause 0x2 +; GFX11-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-NEXT: s_waitcnt vmcnt(2) +; GFX11-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-NEXT: s_waitcnt vmcnt(1) +; GFX11-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %a = load <12 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <12 x i32> %a + store <12 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} +define void @freeze_v13i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v13i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x3 +; GFX10-SDAG-NEXT: global_load_dword v16, v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dword v[2:3], v16, off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v13i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x3 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dword v16, v[0:1], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dword v[2:3], v16, off offset:48 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v13i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x3 +; GFX11-SDAG-NEXT: global_load_b32 v16, v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b32 v[2:3], v16, off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v13i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x3 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b32 v0, v[0:1], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b32 v[2:3], v0, off offset:48 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <13 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <13 x i32> %a + store <13 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v14i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v14i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x3 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx2 v[16:17], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx2 v[2:3], v[16:17], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v14i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x3 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx2 v[16:17], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx2 v[2:3], v[16:17], off offset:48 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v14i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x3 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b64 v[16:17], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b64 v[2:3], v[16:17], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v14i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x3 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b64 v[0:1], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b64 v[2:3], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <14 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <14 x i32> %a + store <14 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v15i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v15i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x3 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx3 v[16:18], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx3 v[2:3], v[16:18], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v15i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x3 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx3 v[16:18], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx3 v[2:3], v[16:18], off offset:48 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v15i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x3 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b96 v[16:18], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b96 v[2:3], v[16:18], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v15i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x3 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b96 v[16:18], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b96 v[2:3], v[16:18], off offset:48 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <15 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <15 x i32> %a + store <15 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v16i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v16i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x3 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v16i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x3 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:48 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v16i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x3 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[16:19], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[16:19], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v16i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x3 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b128 v[16:19], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[16:19], off offset:48 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <16 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <16 x i32> %a + store <16 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v17i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v17i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x4 +; GFX10-SDAG-NEXT: global_load_dword v20, v[0:1], off offset:64 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX10-SDAG-NEXT: global_store_dword v[2:3], v20, off offset:64 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v17i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x4 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: global_load_dword v20, v[0:1], off offset:64 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dword v[2:3], v20, off offset:64 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v17i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x4 +; GFX11-SDAG-NEXT: global_load_b32 v20, v[0:1], off offset:64 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[16:19], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX11-SDAG-NEXT: global_store_b32 v[2:3], v20, off offset:64 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[16:19], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v17i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x4 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b128 v[16:19], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: global_load_b32 v0, v[0:1], off offset:64 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[16:19], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b32 v[2:3], v0, off offset:64 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <17 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <17 x i32> %a + store <17 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v18i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v18i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x4 +; GFX10-SDAG-NEXT: global_load_dwordx2 v[20:21], v[0:1], off offset:64 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX10-SDAG-NEXT: global_store_dwordx2 v[2:3], v[20:21], off offset:64 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v18i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x4 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: global_load_dwordx2 v[20:21], v[0:1], off offset:64 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx2 v[2:3], v[20:21], off offset:64 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v18i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x4 +; GFX11-SDAG-NEXT: global_load_b64 v[20:21], v[0:1], off offset:64 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[16:19], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX11-SDAG-NEXT: global_store_b64 v[2:3], v[20:21], off offset:64 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[16:19], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v18i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x4 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b128 v[16:19], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: global_load_b64 v[0:1], v[0:1], off offset:64 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[16:19], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b64 v[2:3], v[0:1], off offset:64 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <18 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <18 x i32> %a + store <18 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v19i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v19i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x4 +; GFX10-SDAG-NEXT: global_load_dwordx3 v[20:22], v[0:1], off offset:64 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX10-SDAG-NEXT: global_store_dwordx3 v[2:3], v[20:22], off offset:64 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v19i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x4 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: global_load_dwordx3 v[20:22], v[0:1], off offset:64 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx3 v[2:3], v[20:22], off offset:64 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v19i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x4 +; GFX11-SDAG-NEXT: global_load_b96 v[20:22], v[0:1], off offset:64 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[16:19], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX11-SDAG-NEXT: global_store_b96 v[2:3], v[20:22], off offset:64 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[16:19], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v19i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x4 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b128 v[16:19], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: global_load_b96 v[20:22], v[0:1], off offset:64 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[16:19], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b96 v[2:3], v[20:22], off offset:64 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <19 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <19 x i32> %a + store <19 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v20i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v20i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x4 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:64 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[16:19], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:64 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[16:19], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v20i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x4 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:64 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:64 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v20i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x4 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:64 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[16:19], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[20:23], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:64 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[16:19], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[20:23], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v20i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x4 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b128 v[16:19], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: global_load_b128 v[20:23], v[0:1], off offset:64 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[16:19], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[20:23], off offset:64 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <20 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <20 x i32> %a + store <20 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v21i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v21i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x5 +; GFX10-SDAG-NEXT: global_load_dword v24, v[0:1], off offset:80 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:64 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[16:19], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(5) +; GFX10-SDAG-NEXT: global_store_dword v[2:3], v24, off offset:80 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:64 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[16:19], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v21i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x5 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:64 +; GFX10-GISEL-NEXT: global_load_dword v24, v[0:1], off offset:80 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(5) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:64 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dword v[2:3], v24, off offset:80 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v21i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x5 +; GFX11-SDAG-NEXT: global_load_b32 v24, v[0:1], off offset:80 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:64 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[16:19], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[20:23], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(5) +; GFX11-SDAG-NEXT: global_store_b32 v[2:3], v24, off offset:80 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:64 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[16:19], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[20:23], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v21i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x5 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b128 v[16:19], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: global_load_b128 v[20:23], v[0:1], off offset:64 +; GFX11-GISEL-NEXT: global_load_b32 v0, v[0:1], off offset:80 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(5) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[16:19], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[20:23], off offset:64 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b32 v[2:3], v0, off offset:80 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <21 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <21 x i32> %a + store <21 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v22i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v22i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x5 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:64 +; GFX10-SDAG-NEXT: global_load_dwordx2 v[24:25], v[0:1], off offset:80 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[16:19], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(5) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:64 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX10-SDAG-NEXT: global_store_dwordx2 v[2:3], v[24:25], off offset:80 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[16:19], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v22i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x5 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:64 +; GFX10-GISEL-NEXT: global_load_dwordx2 v[24:25], v[0:1], off offset:80 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(5) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:64 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx2 v[2:3], v[24:25], off offset:80 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v22i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x5 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:64 +; GFX11-SDAG-NEXT: global_load_b64 v[24:25], v[0:1], off offset:80 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[16:19], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[20:23], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(5) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:64 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX11-SDAG-NEXT: global_store_b64 v[2:3], v[24:25], off offset:80 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[16:19], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[20:23], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v22i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x5 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b128 v[16:19], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: global_load_b128 v[20:23], v[0:1], off offset:64 +; GFX11-GISEL-NEXT: global_load_b64 v[0:1], v[0:1], off offset:80 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(5) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[16:19], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[20:23], off offset:64 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b64 v[2:3], v[0:1], off offset:80 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <22 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <22 x i32> %a + store <22 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v30i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v30i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x7 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:96 +; GFX10-SDAG-NEXT: global_load_dwordx2 v[32:33], v[0:1], off offset:112 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:64 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:80 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[24:27], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[28:31], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(7) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:96 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(6) +; GFX10-SDAG-NEXT: global_store_dwordx2 v[2:3], v[32:33], off offset:112 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(5) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:64 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:80 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[24:27], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[28:31], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v30i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x7 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:64 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[24:27], v[0:1], off offset:80 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[28:31], v[0:1], off offset:96 +; GFX10-GISEL-NEXT: global_load_dwordx2 v[32:33], v[0:1], off offset:112 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(7) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(6) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(5) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:64 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[24:27], off offset:80 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[28:31], off offset:96 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx2 v[2:3], v[32:33], off offset:112 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v30i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x7 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:96 +; GFX11-SDAG-NEXT: global_load_b64 v[32:33], v[0:1], off offset:112 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:64 +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off offset:80 +; GFX11-SDAG-NEXT: global_load_b128 v[16:19], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[20:23], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[24:27], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[28:31], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(7) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:96 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(6) +; GFX11-SDAG-NEXT: global_store_b64 v[2:3], v[32:33], off offset:112 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(5) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:64 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off offset:80 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[16:19], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[20:23], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[24:27], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[28:31], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v30i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x7 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b128 v[16:19], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: global_load_b128 v[20:23], v[0:1], off offset:64 +; GFX11-GISEL-NEXT: global_load_b128 v[24:27], v[0:1], off offset:80 +; GFX11-GISEL-NEXT: global_load_b128 v[28:31], v[0:1], off offset:96 +; GFX11-GISEL-NEXT: global_load_b64 v[0:1], v[0:1], off offset:112 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(7) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(6) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(5) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[16:19], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[20:23], off offset:64 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[24:27], off offset:80 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[28:31], off offset:96 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b64 v[2:3], v[0:1], off offset:112 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <30 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <30 x i32> %a + store <30 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v31i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v31i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x7 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:96 +; GFX10-SDAG-NEXT: global_load_dwordx3 v[32:34], v[0:1], off offset:112 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:64 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:80 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[24:27], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[28:31], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(7) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:96 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(6) +; GFX10-SDAG-NEXT: global_store_dwordx3 v[2:3], v[32:34], off offset:112 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(5) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:64 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:80 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[24:27], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[28:31], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v31i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x7 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:64 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[24:27], v[0:1], off offset:80 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[28:31], v[0:1], off offset:96 +; GFX10-GISEL-NEXT: global_load_dwordx3 v[32:34], v[0:1], off offset:112 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(7) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(6) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(5) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:64 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[24:27], off offset:80 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[28:31], off offset:96 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx3 v[2:3], v[32:34], off offset:112 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v31i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x7 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:96 +; GFX11-SDAG-NEXT: global_load_b96 v[32:34], v[0:1], off offset:112 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:64 +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off offset:80 +; GFX11-SDAG-NEXT: global_load_b128 v[16:19], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[20:23], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[24:27], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[28:31], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(7) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:96 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(6) +; GFX11-SDAG-NEXT: global_store_b96 v[2:3], v[32:34], off offset:112 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(5) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:64 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off offset:80 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[16:19], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[20:23], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[24:27], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[28:31], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v31i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x7 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b128 v[16:19], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: global_load_b128 v[20:23], v[0:1], off offset:64 +; GFX11-GISEL-NEXT: global_load_b128 v[24:27], v[0:1], off offset:80 +; GFX11-GISEL-NEXT: global_load_b128 v[28:31], v[0:1], off offset:96 +; GFX11-GISEL-NEXT: global_load_b96 v[32:34], v[0:1], off offset:112 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(7) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(6) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(5) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[16:19], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[20:23], off offset:64 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[24:27], off offset:80 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[28:31], off offset:96 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b96 v[2:3], v[32:34], off offset:112 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <31 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <31 x i32> %a + store <31 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_v32i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_v32i32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x7 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:96 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:112 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:64 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:80 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:32 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[24:27], v[0:1], off offset:48 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[28:31], v[0:1], off +; GFX10-SDAG-NEXT: global_load_dwordx4 v[32:35], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(7) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:96 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(6) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:112 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(5) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:64 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:80 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:32 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[24:27], off offset:48 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[28:31], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[32:35], off offset:16 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_v32i32: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x7 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[12:15], v[0:1], off offset:32 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[16:19], v[0:1], off offset:48 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[20:23], v[0:1], off offset:64 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[24:27], v[0:1], off offset:80 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[28:31], v[0:1], off offset:96 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[32:35], v[0:1], off offset:112 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(7) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(6) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(5) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[12:15], off offset:32 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[16:19], off offset:48 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[20:23], off offset:64 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[24:27], off offset:80 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[28:31], off offset:96 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[32:35], off offset:112 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_v32i32: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x7 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:96 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off offset:112 +; GFX11-SDAG-NEXT: global_load_b128 v[12:15], v[0:1], off offset:64 +; GFX11-SDAG-NEXT: global_load_b128 v[16:19], v[0:1], off offset:80 +; GFX11-SDAG-NEXT: global_load_b128 v[20:23], v[0:1], off offset:32 +; GFX11-SDAG-NEXT: global_load_b128 v[24:27], v[0:1], off offset:48 +; GFX11-SDAG-NEXT: global_load_b128 v[28:31], v[0:1], off +; GFX11-SDAG-NEXT: global_load_b128 v[32:35], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(7) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:96 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(6) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off offset:112 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(5) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[12:15], off offset:64 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(4) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[16:19], off offset:80 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(3) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[20:23], off offset:32 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(2) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[24:27], off offset:48 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[28:31], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[32:35], off offset:16 +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_v32i32: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x7 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: global_load_b128 v[12:15], v[0:1], off offset:32 +; GFX11-GISEL-NEXT: global_load_b128 v[16:19], v[0:1], off offset:48 +; GFX11-GISEL-NEXT: global_load_b128 v[20:23], v[0:1], off offset:64 +; GFX11-GISEL-NEXT: global_load_b128 v[24:27], v[0:1], off offset:80 +; GFX11-GISEL-NEXT: global_load_b128 v[28:31], v[0:1], off offset:96 +; GFX11-GISEL-NEXT: global_load_b128 v[32:35], v[0:1], off offset:112 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(7) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(6) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(5) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[12:15], off offset:32 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(4) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[16:19], off offset:48 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(3) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[20:23], off offset:64 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(2) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[24:27], off offset:80 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[28:31], off offset:96 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[32:35], off offset:112 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load <32 x i32>, ptr addrspace(1) %ptra, align 4 + %freeze = freeze <32 x i32> %a + store <32 x i32> %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_i32(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-LABEL: freeze_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: global_load_dword v0, v[0:1], off +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: global_store_dword v[2:3], v0, off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: freeze_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: global_load_b32 v0, v[0:1], off +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_store_b32 v[2:3], v0, off +; GFX11-NEXT: s_setpc_b64 s[30:31] + %a = load i32, ptr addrspace(1) %ptra, align 4 + %freeze = freeze i32 %a + store i32 %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_i64(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-LABEL: freeze_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: freeze_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_store_b64 v[2:3], v[0:1], off +; GFX11-NEXT: s_setpc_b64 s[30:31] + %a = load i64, ptr addrspace(1) %ptra, align 4 + %freeze = freeze i64 %a + store i64 %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_float(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-LABEL: freeze_float: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: global_load_dword v0, v[0:1], off +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: global_store_dword v[2:3], v0, off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: freeze_float: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: global_load_b32 v0, v[0:1], off +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_store_b32 v[2:3], v0, off +; GFX11-NEXT: s_setpc_b64 s[30:31] + %a = load float, ptr addrspace(1) %ptra, align 4 + %freeze = freeze float %a + store float %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_i128(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-LABEL: freeze_i128: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: freeze_i128: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-NEXT: s_setpc_b64 s[30:31] + %a = load i128, ptr addrspace(1) %ptra, align 4 + %freeze = freeze i128 %a + store i128 %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +} + +define void @freeze_i256(ptr addrspace(1) %ptra, ptr addrspace(1) %ptrb) { +; GFX10-SDAG-LABEL: freeze_i256: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: s_clause 0x1 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[4:7], v[0:1], off offset:16 +; GFX10-SDAG-NEXT: global_load_dwordx4 v[8:11], v[0:1], off +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off offset:16 +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX10-SDAG-NEXT: global_store_dwordx4 v[2:3], v[8:11], off +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-GISEL-LABEL: freeze_i256: +; GFX10-GISEL: ; %bb.0: +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-GISEL-NEXT: s_clause 0x1 +; GFX10-GISEL-NEXT: global_load_dwordx4 v[4:7], v[0:1], off +; GFX10-GISEL-NEXT: global_load_dwordx4 v[8:11], v[0:1], off offset:16 +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[4:7], off +; GFX10-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX10-GISEL-NEXT: global_store_dwordx4 v[2:3], v[8:11], off offset:16 +; GFX10-GISEL-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-SDAG-LABEL: freeze_i256: +; GFX11-SDAG: ; %bb.0: +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-SDAG-NEXT: s_clause 0x1 +; GFX11-SDAG-NEXT: global_load_b128 v[4:7], v[0:1], off offset:16 +; GFX11-SDAG-NEXT: global_load_b128 v[8:11], v[0:1], off +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(1) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[4:7], off offset:16 +; GFX11-SDAG-NEXT: s_waitcnt vmcnt(0) +; GFX11-SDAG-NEXT: global_store_b128 v[2:3], v[8:11], off +; GFX11-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-GISEL-LABEL: freeze_i256: +; GFX11-GISEL: ; %bb.0: +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-GISEL-NEXT: s_clause 0x1 +; GFX11-GISEL-NEXT: global_load_b128 v[4:7], v[0:1], off +; GFX11-GISEL-NEXT: global_load_b128 v[8:11], v[0:1], off offset:16 +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(1) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[4:7], off +; GFX11-GISEL-NEXT: s_waitcnt vmcnt(0) +; GFX11-GISEL-NEXT: global_store_b128 v[2:3], v[8:11], off offset:16 +; GFX11-GISEL-NEXT: s_setpc_b64 s[30:31] + %a = load i256, ptr addrspace(1) %ptra, align 4 + %freeze = freeze i256 %a + store i256 %freeze, ptr addrspace(1) %ptrb, align 4 + ret void +}