From 9236125ec88f619799879beef5a6d2855c87bced Mon Sep 17 00:00:00 2001 From: Matt Arsenault Date: Thu, 10 Jun 2021 19:32:41 -0400 Subject: [PATCH] GlobalISel: Preserve LLT when bitcasting loads and stores This also avoids improperly legalizing some truncating vector stores. --- llvm/include/llvm/CodeGen/MachineMemOperand.h | 5 + llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp | 13 + .../AMDGPU/GlobalISel/artifact-combiner-zext.mir | 2 +- .../GlobalISel/legalize-extract-vector-elt.mir | 8 +- .../AMDGPU/GlobalISel/legalize-load-constant.mir | 138 +++---- .../AMDGPU/GlobalISel/legalize-load-flat.mir | 84 ++-- .../AMDGPU/GlobalISel/legalize-load-global.mir | 457 +++++++++++---------- .../AMDGPU/GlobalISel/legalize-load-local.mir | 140 ++++--- .../AMDGPU/GlobalISel/legalize-load-private.mir | 16 +- .../AMDGPU/GlobalISel/legalize-store-global.mir | 122 +++--- .../CodeGen/AMDGPU/GlobalISel/legalize-store.mir | 168 ++------ 11 files changed, 552 insertions(+), 601 deletions(-) diff --git a/llvm/include/llvm/CodeGen/MachineMemOperand.h b/llvm/include/llvm/CodeGen/MachineMemOperand.h index 8c1fb1d..07b8e5e 100644 --- a/llvm/include/llvm/CodeGen/MachineMemOperand.h +++ b/llvm/include/llvm/CodeGen/MachineMemOperand.h @@ -329,6 +329,11 @@ public: void setValue(const PseudoSourceValue *NewSV) { PtrInfo.V = NewSV; } void setOffset(int64_t NewOffset) { PtrInfo.Offset = NewOffset; } + /// Reset the tracked memory type. + void setType(LLT NewTy) { + MemoryType = NewTy; + } + /// Profile - Gather unique data for the object. /// void Profile(FoldingSetNodeID &ID) const; diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp index 8a54b87..26a6486 100644 --- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp +++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp @@ -2983,9 +2983,15 @@ LegalizerHelper::bitcast(MachineInstr &MI, unsigned TypeIdx, LLT CastTy) { case TargetOpcode::G_LOAD: { if (TypeIdx != 0) return UnableToLegalize; + MachineMemOperand &MMO = **MI.memoperands_begin(); + + // Not sure how to interpret a bitcast of an extending load. + if (MMO.getMemoryType().getSizeInBits() != CastTy.getSizeInBits()) + return UnableToLegalize; Observer.changingInstr(MI); bitcastDst(MI, CastTy, 0); + MMO.setType(CastTy); Observer.changedInstr(MI); return Legalized; } @@ -2993,8 +2999,15 @@ LegalizerHelper::bitcast(MachineInstr &MI, unsigned TypeIdx, LLT CastTy) { if (TypeIdx != 0) return UnableToLegalize; + MachineMemOperand &MMO = **MI.memoperands_begin(); + + // Not sure how to interpret a bitcast of a truncating store. + if (MMO.getMemoryType().getSizeInBits() != CastTy.getSizeInBits()) + return UnableToLegalize; + Observer.changingInstr(MI); bitcastSrc(MI, CastTy, 0); + MMO.setType(CastTy); Observer.changedInstr(MI); return Legalized; } diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/artifact-combiner-zext.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/artifact-combiner-zext.mir index 9f53130..5b62d03 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/artifact-combiner-zext.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/artifact-combiner-zext.mir @@ -255,7 +255,7 @@ body: | ; CHECK-LABEL: name: test_zext_v2s8_to_v2s32_of_sext_v2s8_to_v2s16 ; CHECK: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CHECK: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) + ; CHECK: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-extract-vector-elt.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-extract-vector-elt.mir index 0c42cd8..88a0437 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-extract-vector-elt.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-extract-vector-elt.mir @@ -1476,19 +1476,19 @@ body: | ; CHECK-LABEL: name: extract_vector_elt_33_v64p3 ; CHECK: [[COPY:%[0-9]+]]:_(p1) = COPY $sgpr0_sgpr1 - ; CHECK: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x p3>), align 4, addrspace 4) + ; CHECK: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 4, addrspace 4) ; CHECK: [[BITCAST:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD]](<16 x s32>) ; CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 64 ; CHECK: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; CHECK: [[LOAD1:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<16 x p3>) from unknown-address + 64, align 4, addrspace 4) + ; CHECK: [[LOAD1:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<16 x s32>) from unknown-address + 64, align 4, addrspace 4) ; CHECK: [[BITCAST1:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD1]](<16 x s32>) ; CHECK: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 128 ; CHECK: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64) - ; CHECK: [[LOAD2:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD1]](p1) :: (load (<16 x p3>) from unknown-address + 128, align 4, addrspace 4) + ; CHECK: [[LOAD2:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD1]](p1) :: (load (<16 x s32>) from unknown-address + 128, align 4, addrspace 4) ; CHECK: [[BITCAST2:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD2]](<16 x s32>) ; CHECK: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 192 ; CHECK: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64) - ; CHECK: [[LOAD3:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD2]](p1) :: (load (<16 x p3>) from unknown-address + 192, align 4, addrspace 4) + ; CHECK: [[LOAD3:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD2]](p1) :: (load (<16 x s32>) from unknown-address + 192, align 4, addrspace 4) ; CHECK: [[BITCAST3:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD3]](<16 x s32>) ; CHECK: [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0 ; CHECK: [[UV:%[0-9]+]]:_(p3), [[UV1:%[0-9]+]]:_(p3), [[UV2:%[0-9]+]]:_(p3), [[UV3:%[0-9]+]]:_(p3), [[UV4:%[0-9]+]]:_(p3), [[UV5:%[0-9]+]]:_(p3), [[UV6:%[0-9]+]]:_(p3), [[UV7:%[0-9]+]]:_(p3), [[UV8:%[0-9]+]]:_(p3), [[UV9:%[0-9]+]]:_(p3), [[UV10:%[0-9]+]]:_(p3), [[UV11:%[0-9]+]]:_(p3), [[UV12:%[0-9]+]]:_(p3), [[UV13:%[0-9]+]]:_(p3), [[UV14:%[0-9]+]]:_(p3), [[UV15:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST]](<16 x p3>) diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir index 0485479..60037381 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir @@ -1085,17 +1085,17 @@ body: | ; CI-LABEL: name: test_load_constant_s96_align16 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI-LABEL: name: test_load_constant_s96_align16 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-LABEL: name: test_load_constant_s96_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -1111,17 +1111,17 @@ body: | ; CI-LABEL: name: test_load_constant_s96_align8 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 8, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 8, addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI-LABEL: name: test_load_constant_s96_align8 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 8, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 8, addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-LABEL: name: test_load_constant_s96_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 8, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 8, addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -1137,17 +1137,17 @@ body: | ; CI-LABEL: name: test_load_constant_s96_align4 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI-LABEL: name: test_load_constant_s96_align4 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-LABEL: name: test_load_constant_s96_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -1651,17 +1651,17 @@ body: | ; CI-LABEL: name: test_load_constant_s128_align16 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI-LABEL: name: test_load_constant_s128_align16 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-LABEL: name: test_load_constant_s128_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -1677,17 +1677,17 @@ body: | ; CI-LABEL: name: test_load_constant_s128_align4 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), align 4, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI-LABEL: name: test_load_constant_s128_align4 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), align 4, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-LABEL: name: test_load_constant_s128_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (s128), align 4, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -2020,17 +2020,17 @@ body: | ; CI-LABEL: name: test_load_constant_s256_align32 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (s256), align 16, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), align 16, addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; VI-LABEL: name: test_load_constant_s256_align32 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (s256), align 16, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), align 16, addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; GFX9-LABEL: name: test_load_constant_s256_align32 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (s256), align 16, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), align 16, addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -2849,7 +2849,7 @@ body: | ; CI-LABEL: name: test_load_constant_v2s8_align4 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), align 4, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2870,7 +2870,7 @@ body: | ; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_load_constant_v2s8_align4 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), align 4, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2889,7 +2889,7 @@ body: | ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-LABEL: name: test_load_constant_v2s8_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), align 4, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), align 4, addrspace 4) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2921,7 +2921,7 @@ body: | ; CI-LABEL: name: test_load_constant_v2s8_align2 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2942,7 +2942,7 @@ body: | ; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_load_constant_v2s8_align2 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2961,7 +2961,7 @@ body: | ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-LABEL: name: test_load_constant_v2s8_align2 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<2 x s8>), addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s16), addrspace 4) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -3200,7 +3200,7 @@ body: | ; CI-LABEL: name: test_load_constant_v4s8_align4 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<4 x s8>), addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -3216,7 +3216,7 @@ body: | ; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI-LABEL: name: test_load_constant_v4s8_align4 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<4 x s8>), addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -3232,7 +3232,7 @@ body: | ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-LABEL: name: test_load_constant_v4s8_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (<4 x s8>), addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p4) :: (load (s32), addrspace 4) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -3410,7 +3410,7 @@ body: | ; CI-LABEL: name: test_load_constant_v8s8_align8 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s8>), addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3437,7 +3437,7 @@ body: | ; CI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; VI-LABEL: name: test_load_constant_v8s8_align8 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s8>), addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3464,7 +3464,7 @@ body: | ; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; GFX9-LABEL: name: test_load_constant_v8s8_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s8>), addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3506,7 +3506,7 @@ body: | ; CI-LABEL: name: test_load_constant_v16s8_align16 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s8>), addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3551,7 +3551,7 @@ body: | ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; VI-LABEL: name: test_load_constant_v16s8_align16 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s8>), addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3596,7 +3596,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; GFX9-LABEL: name: test_load_constant_v16s8_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s8>), addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3660,7 +3660,7 @@ body: | ; CI-LABEL: name: test_load_constant_v32s8_align32 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<32 x s8>), addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4) ; CI: [[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) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3741,7 +3741,7 @@ body: | ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; VI-LABEL: name: test_load_constant_v32s8_align32 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<32 x s8>), addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4) ; VI: [[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) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3822,7 +3822,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; GFX9-LABEL: name: test_load_constant_v32s8_align32 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<32 x s8>), addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4) ; GFX9: [[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) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -4778,17 +4778,17 @@ body: | ; CI-LABEL: name: test_load_constant_v8s16_align8 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s16>), align 8, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; VI-LABEL: name: test_load_constant_v8s16_align8 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s16>), align 8, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9-LABEL: name: test_load_constant_v8s16_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s16>), align 8, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -7234,17 +7234,17 @@ body: | ; CI-LABEL: name: test_load_constant_v2s128_align32 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s128>), addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; VI-LABEL: name: test_load_constant_v2s128_align32 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s128>), addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; GFX9-LABEL: name: test_load_constant_v2s128_align32 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s128>), addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p4) :: (load (<8 x s32>), addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -7260,17 +7260,17 @@ body: | ; CI-LABEL: name: test_load_constant_v2p1_align16 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI-LABEL: name: test_load_constant_v2p1_align16 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-LABEL: name: test_load_constant_v2p1_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -7286,17 +7286,17 @@ body: | ; CI-LABEL: name: test_load_constant_v2p1_align8 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 8, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI-LABEL: name: test_load_constant_v2p1_align8 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 8, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-LABEL: name: test_load_constant_v2p1_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 8, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 8, addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -7312,17 +7312,17 @@ body: | ; CI-LABEL: name: test_load_constant_v2p1_align4 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 4, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI-LABEL: name: test_load_constant_v2p1_align4 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 4, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-LABEL: name: test_load_constant_v2p1_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x p1>), align 4, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -9138,11 +9138,11 @@ body: | ; CI-LABEL: name: test_extload_constant_v2s96_from_24_align4 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 1) + ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) - ; CI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; CI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; CI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -9150,11 +9150,11 @@ body: | ; CI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; VI-LABEL: name: test_extload_constant_v2s96_from_24_align4 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; VI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) - ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -9162,11 +9162,11 @@ body: | ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-LABEL: name: test_extload_constant_v2s96_from_24_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 4, addrspace 1) + ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 4, addrspace 1) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) - ; GFX9: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; GFX9: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -9188,11 +9188,11 @@ body: | ; CI-LABEL: name: test_extload_constant_v2s96_from_24_align16 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 1) + ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 1) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) - ; CI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; CI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; CI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -9200,11 +9200,11 @@ body: | ; CI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; VI-LABEL: name: test_extload_constant_v2s96_from_24_align16 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; VI: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) - ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -9212,11 +9212,11 @@ body: | ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-LABEL: name: test_extload_constant_v2s96_from_24_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (s96), align 16, addrspace 1) + ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p4) :: (load (<3 x s32>), align 16, addrspace 1) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64) - ; GFX9: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; GFX9: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; GFX9: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -9238,17 +9238,17 @@ body: | ; CI-LABEL: name: test_load_constant_s512_align32 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (s512), align 32, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512) ; VI-LABEL: name: test_load_constant_s512_align32 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (s512), align 32, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512) ; GFX9-LABEL: name: test_load_constant_s512_align32 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (s512), align 32, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(s512) = G_BITCAST [[LOAD]](<16 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](s512) %0:_(p4) = COPY $vgpr0_vgpr1 @@ -9264,17 +9264,17 @@ body: | ; CI-LABEL: name: test_load_constant_v4s128_align32 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s128>), align 32, addrspace 4) + ; CI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4) ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>) ; VI-LABEL: name: test_load_constant_v4s128_align32 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s128>), align 32, addrspace 4) + ; VI: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>) ; GFX9-LABEL: name: test_load_constant_v4s128_align32 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s128>), align 32, addrspace 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p4) :: (load (<16 x s32>), align 32, addrspace 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s128>) = G_BITCAST [[LOAD]](<16 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15 = COPY [[BITCAST]](<4 x s128>) %0:_(p4) = COPY $vgpr0_vgpr1 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir index af33169..c11148f 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir @@ -807,17 +807,17 @@ body: | ; CI-LABEL: name: test_load_flat_s96_align16 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 16) + ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI-LABEL: name: test_load_flat_s96_align16 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 16) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-LABEL: name: test_load_flat_s96_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 16) + ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 16) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) %0:_(p0) = COPY $vgpr0_vgpr1 @@ -833,17 +833,17 @@ body: | ; CI-LABEL: name: test_load_flat_s96_align8 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 8) + ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI-LABEL: name: test_load_flat_s96_align8 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 8) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-LABEL: name: test_load_flat_s96_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 8) + ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 8) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) %0:_(p0) = COPY $vgpr0_vgpr1 @@ -859,17 +859,17 @@ body: | ; CI-LABEL: name: test_load_flat_s96_align4 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 4) + ; CI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4) ; CI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI-LABEL: name: test_load_flat_s96_align4 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 4) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-LABEL: name: test_load_flat_s96_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (s96), align 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p0) :: (load (<3 x s32>), align 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) %0:_(p0) = COPY $vgpr0_vgpr1 @@ -1373,17 +1373,17 @@ body: | ; CI-LABEL: name: test_load_flat_s128_align16 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128)) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>)) ; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI-LABEL: name: test_load_flat_s128_align16 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128)) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>)) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-LABEL: name: test_load_flat_s128_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128)) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>)) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) %0:_(p0) = COPY $vgpr0_vgpr1 @@ -1399,17 +1399,17 @@ body: | ; CI-LABEL: name: test_load_flat_s128_align4 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128), align 4) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4) ; CI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI-LABEL: name: test_load_flat_s128_align4 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128), align 4) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-LABEL: name: test_load_flat_s128_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (s128), align 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) %0:_(p0) = COPY $vgpr0_vgpr1 @@ -2583,7 +2583,7 @@ body: | ; CI-LABEL: name: test_load_flat_v2s8_align4 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>), align 4) + ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2604,7 +2604,7 @@ body: | ; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_load_flat_v2s8_align4 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>), align 4) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2623,7 +2623,7 @@ body: | ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-LABEL: name: test_load_flat_v2s8_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>), align 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16), align 4) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2655,7 +2655,7 @@ body: | ; CI-LABEL: name: test_load_flat_v2s8_align2 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>)) + ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16)) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2676,7 +2676,7 @@ body: | ; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_load_flat_v2s8_align2 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>)) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16)) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2695,7 +2695,7 @@ body: | ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-LABEL: name: test_load_flat_v2s8_align2 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<2 x s8>)) + ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s16)) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2934,7 +2934,7 @@ body: | ; CI-LABEL: name: test_load_flat_v4s8_align4 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<4 x s8>)) + ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32)) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2950,7 +2950,7 @@ body: | ; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI-LABEL: name: test_load_flat_v4s8_align4 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<4 x s8>)) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32)) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -2966,7 +2966,7 @@ body: | ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-LABEL: name: test_load_flat_v4s8_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (<4 x s8>)) + ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p0) :: (load (s32)) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -3144,7 +3144,7 @@ body: | ; CI-LABEL: name: test_load_flat_v8s8_align8 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s8>)) + ; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>)) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3171,7 +3171,7 @@ body: | ; CI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; VI-LABEL: name: test_load_flat_v8s8_align8 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s8>)) + ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>)) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3198,7 +3198,7 @@ body: | ; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; GFX9-LABEL: name: test_load_flat_v8s8_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s8>)) + ; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x s32>)) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3240,7 +3240,7 @@ body: | ; CI-LABEL: name: test_load_flat_v16s8_align16 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<16 x s8>)) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>)) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3285,7 +3285,7 @@ body: | ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; VI-LABEL: name: test_load_flat_v16s8_align16 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<16 x s8>)) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>)) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -3330,7 +3330,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; GFX9-LABEL: name: test_load_flat_v16s8_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<16 x s8>)) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>)) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -4524,17 +4524,17 @@ body: | ; CI-LABEL: name: test_load_flat_v8s16_align8 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s16>), align 8) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8) ; CI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; VI-LABEL: name: test_load_flat_v8s16_align8 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s16>), align 8) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9-LABEL: name: test_load_flat_v8s16_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<8 x s16>), align 8) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8) ; GFX9: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) %0:_(p0) = COPY $vgpr0_vgpr1 @@ -6832,17 +6832,17 @@ body: | ; CI-LABEL: name: test_load_flat_v2p1_align16 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>)) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>)) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI-LABEL: name: test_load_flat_v2p1_align16 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>)) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>)) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-LABEL: name: test_load_flat_v2p1_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>)) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>)) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) %0:_(p0) = COPY $vgpr0_vgpr1 @@ -6858,17 +6858,17 @@ body: | ; CI-LABEL: name: test_load_flat_v2p1_align8 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 8) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI-LABEL: name: test_load_flat_v2p1_align8 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 8) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-LABEL: name: test_load_flat_v2p1_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 8) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 8) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) %0:_(p0) = COPY $vgpr0_vgpr1 @@ -6884,17 +6884,17 @@ body: | ; CI-LABEL: name: test_load_flat_v2p1_align4 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 4) + ; CI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4) ; CI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI-LABEL: name: test_load_flat_v2p1_align4 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 4) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-LABEL: name: test_load_flat_v2p1_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<2 x p1>), align 4) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4) ; GFX9: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) %0:_(p0) = COPY $vgpr0_vgpr1 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir index 29a3b22c..872f27d 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir @@ -1640,27 +1640,27 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-HSA-LABEL: name: test_load_global_s96_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-MESA-LABEL: name: test_load_global_s96_align16 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI-LABEL: name: test_load_global_s96_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-HSA-LABEL: name: test_load_global_s96_align16 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-MESA-LABEL: name: test_load_global_s96_align16 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -1686,27 +1686,27 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-HSA-LABEL: name: test_load_global_s96_align8 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 8, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-MESA-LABEL: name: test_load_global_s96_align8 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 8, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI-LABEL: name: test_load_global_s96_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 8, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-HSA-LABEL: name: test_load_global_s96_align8 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 8, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-MESA-LABEL: name: test_load_global_s96_align8 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 8, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -1732,27 +1732,27 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-HSA-LABEL: name: test_load_global_s96_align4 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-MESA-LABEL: name: test_load_global_s96_align4 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; VI-LABEL: name: test_load_global_s96_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-HSA-LABEL: name: test_load_global_s96_align4 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-MESA-LABEL: name: test_load_global_s96_align4 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -1807,7 +1807,7 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-HSA-LABEL: name: test_load_global_s96_align2 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 2, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-MESA-LABEL: name: test_load_global_s96_align2 @@ -1890,7 +1890,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-HSA-LABEL: name: test_load_global_s96_align2 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 2, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-MESA-LABEL: name: test_load_global_s96_align2 @@ -2024,7 +2024,7 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-HSA-LABEL: name: test_load_global_s96_align1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 1, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; CI-MESA-LABEL: name: test_load_global_s96_align1 @@ -2187,7 +2187,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-HSA-LABEL: name: test_load_global_s96_align1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 1, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-MESA-LABEL: name: test_load_global_s96_align1 @@ -2489,32 +2489,32 @@ body: | ; SI-LABEL: name: test_load_global_s128_align16 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-HSA-LABEL: name: test_load_global_s128_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-MESA-LABEL: name: test_load_global_s128_align16 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI-LABEL: name: test_load_global_s128_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-HSA-LABEL: name: test_load_global_s128_align16 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-MESA-LABEL: name: test_load_global_s128_align16 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -2534,32 +2534,32 @@ body: | ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128) ; SI-LABEL: name: test_load_global_s128_align4 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-HSA-LABEL: name: test_load_global_s128_align4 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-MESA-LABEL: name: test_load_global_s128_align4 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI-LABEL: name: test_load_global_s128_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-HSA-LABEL: name: test_load_global_s128_align4 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-MESA-LABEL: name: test_load_global_s128_align4 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -2677,7 +2677,7 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-HSA-LABEL: name: test_load_global_s128_align1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 1, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-MESA-LABEL: name: test_load_global_s128_align1 @@ -2886,7 +2886,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-HSA-LABEL: name: test_load_global_s128_align1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (s128), align 1, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-MESA-LABEL: name: test_load_global_s128_align1 @@ -3004,32 +3004,32 @@ body: | ; SI-LABEL: name: test_load_global_s256_align32 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; CI-HSA-LABEL: name: test_load_global_s256_align32 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; CI-MESA-LABEL: name: test_load_global_s256_align32 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; VI-LABEL: name: test_load_global_s256_align32 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; GFX9-HSA-LABEL: name: test_load_global_s256_align32 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) ; GFX9-MESA-LABEL: name: test_load_global_s256_align32 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (s256), align 16, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 16, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s256) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](s256) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -4178,7 +4178,7 @@ body: | ; SI-LABEL: name: test_load_global_v2s8_align4 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4199,7 +4199,7 @@ body: | ; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-HSA-LABEL: name: test_load_global_v2s8_align4 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4220,7 +4220,7 @@ body: | ; CI-HSA: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-MESA-LABEL: name: test_load_global_v2s8_align4 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1) ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4241,7 +4241,7 @@ body: | ; CI-MESA: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_load_global_v2s8_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4260,7 +4260,7 @@ body: | ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-HSA-LABEL: name: test_load_global_v2s8_align4 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4279,7 +4279,7 @@ body: | ; GFX9-HSA: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-MESA-LABEL: name: test_load_global_v2s8_align4 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 4, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1) ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4311,7 +4311,7 @@ body: | ; SI-LABEL: name: test_load_global_v2s8_align2 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4332,7 +4332,7 @@ body: | ; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-HSA-LABEL: name: test_load_global_v2s8_align2 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4353,7 +4353,7 @@ body: | ; CI-HSA: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-MESA-LABEL: name: test_load_global_v2s8_align2 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4374,7 +4374,7 @@ body: | ; CI-MESA: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_load_global_v2s8_align2 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4393,7 +4393,7 @@ body: | ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-HSA-LABEL: name: test_load_global_v2s8_align2 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4412,7 +4412,7 @@ body: | ; GFX9-HSA: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-MESA-LABEL: name: test_load_global_v2s8_align2 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1) ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4462,7 +4462,7 @@ body: | ; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-HSA-LABEL: name: test_load_global_v2s8_align1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 1, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4517,7 +4517,7 @@ body: | ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-HSA-LABEL: name: test_load_global_v2s8_align1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<2 x s8>), align 1, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4715,11 +4715,26 @@ body: | ; SI: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; CI-HSA-LABEL: name: test_load_global_v3s8_align1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(s24) = G_LOAD [[COPY]](p1) :: (load (<3 x s8>), align 1, addrspace 1) - ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<3 x s8>) = G_BITCAST [[LOAD]](s24) + ; CI-HSA: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1) + ; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2 + ; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1) + ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; CI-HSA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32) + ; CI-HSA: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]] + ; CI-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 + ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32) + ; CI-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32) + ; CI-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 + ; CI-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32) + ; CI-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[OR]](s32) + ; CI-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; CI-HSA: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) + ; CI-HSA: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32) + ; CI-HSA: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[BUILD_VECTOR]](<3 x s32>) ; CI-HSA: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF - ; CI-HSA: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) - ; CI-HSA: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC]], [[BITCAST]](<3 x s8>), 0 + ; CI-HSA: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) + ; CI-HSA: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0 ; CI-HSA: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; CI-MESA-LABEL: name: test_load_global_v3s8_align1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 @@ -4759,11 +4774,31 @@ body: | ; VI: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; GFX9-HSA-LABEL: name: test_load_global_v3s8_align1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s24) = G_LOAD [[COPY]](p1) :: (load (<3 x s8>), align 1, addrspace 1) - ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<3 x s8>) = G_BITCAST [[LOAD]](s24) - ; GFX9-HSA: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF - ; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) - ; GFX9-HSA: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC]], [[BITCAST]](<3 x s8>), 0 + ; GFX9-HSA: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1) + ; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2 + ; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 2, addrspace 1) + ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; GFX9-HSA: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32) + ; GFX9-HSA: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]] + ; GFX9-HSA: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 + ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32) + ; GFX9-HSA: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32) + ; GFX9-HSA: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 + ; GFX9-HSA: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32) + ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s32) = COPY [[OR]](s32) + ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; GFX9-HSA: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32) + ; GFX9-HSA: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) + ; GFX9-HSA: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9-HSA: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[DEF]](s32) + ; GFX9-HSA: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; GFX9-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[DEF1]](<2 x s16>) + ; GFX9-HSA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>) + ; GFX9-HSA: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[UV]](<3 x s16>) + ; GFX9-HSA: [[DEF2:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; GFX9-HSA: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF2]](<4 x s32>) + ; GFX9-HSA: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0 ; GFX9-HSA: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; GFX9-MESA-LABEL: name: test_load_global_v3s8_align1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 @@ -4803,7 +4838,7 @@ body: | ; SI-LABEL: name: test_load_global_v4s8_align4 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4819,7 +4854,7 @@ body: | ; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI-HSA-LABEL: name: test_load_global_v4s8_align4 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4835,7 +4870,7 @@ body: | ; CI-HSA: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI-MESA-LABEL: name: test_load_global_v4s8_align4 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4851,7 +4886,7 @@ body: | ; CI-MESA: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI-LABEL: name: test_load_global_v4s8_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4867,7 +4902,7 @@ body: | ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-HSA-LABEL: name: test_load_global_v4s8_align4 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4885,7 +4920,7 @@ body: | ; GFX9-HSA: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-MESA-LABEL: name: test_load_global_v4s8_align4 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4955,7 +4990,7 @@ body: | ; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI-HSA-LABEL: name: test_load_global_v4s8_align2 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), align 2, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -5015,7 +5050,7 @@ body: | ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-HSA-LABEL: name: test_load_global_v4s8_align2 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), align 2, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 2, addrspace 1) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -5087,7 +5122,7 @@ body: | ; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI-HSA-LABEL: name: test_load_global_v4s8_align1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), align 1, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -5141,7 +5176,7 @@ body: | ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-HSA-LABEL: name: test_load_global_v4s8_align1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<4 x s8>), align 1, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), align 1, addrspace 1) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -5191,7 +5226,7 @@ body: | ; SI-LABEL: name: test_load_global_v8s8_align8 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5218,7 +5253,7 @@ body: | ; SI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; CI-HSA-LABEL: name: test_load_global_v8s8_align8 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1) ; CI-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5245,7 +5280,7 @@ body: | ; CI-HSA: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; CI-MESA-LABEL: name: test_load_global_v8s8_align8 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1) ; CI-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5272,7 +5307,7 @@ body: | ; CI-MESA: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; VI-LABEL: name: test_load_global_v8s8_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5299,7 +5334,7 @@ body: | ; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; GFX9-HSA-LABEL: name: test_load_global_v8s8_align8 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1) ; GFX9-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5330,7 +5365,7 @@ body: | ; GFX9-HSA: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>) ; GFX9-MESA-LABEL: name: test_load_global_v8s8_align8 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s8>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s32>), addrspace 1) ; GFX9-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5372,7 +5407,7 @@ body: | ; SI-LABEL: name: test_load_global_v16s8_align16 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5417,7 +5452,7 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; CI-HSA-LABEL: name: test_load_global_v16s8_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5462,7 +5497,7 @@ body: | ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; CI-MESA-LABEL: name: test_load_global_v16s8_align16 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5507,7 +5542,7 @@ body: | ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; VI-LABEL: name: test_load_global_v16s8_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5552,7 +5587,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<16 x s8>) ; GFX9-HSA-LABEL: name: test_load_global_v16s8_align16 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-HSA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5605,7 +5640,7 @@ body: | ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>) ; GFX9-MESA-LABEL: name: test_load_global_v16s8_align16 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s8>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-MESA: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5669,7 +5704,7 @@ body: | ; SI-LABEL: name: test_load_global_v32s8_align32 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; SI: [[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) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5750,7 +5785,7 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; CI-HSA-LABEL: name: test_load_global_v32s8_align32 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; CI-HSA: [[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) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5831,7 +5866,7 @@ body: | ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; CI-MESA-LABEL: name: test_load_global_v32s8_align32 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; CI-MESA: [[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) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5912,7 +5947,7 @@ body: | ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; VI-LABEL: name: test_load_global_v32s8_align32 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; VI: [[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) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -5993,7 +6028,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS]](<32 x s8>) ; GFX9-HSA-LABEL: name: test_load_global_v32s8_align32 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; GFX9-HSA: [[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) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -6090,7 +6125,7 @@ body: | ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[CONCAT_VECTORS8]](<32 x s8>) ; GFX9-MESA-LABEL: name: test_load_global_v32s8_align32 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<32 x s8>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; GFX9-MESA: [[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) = G_UNMERGE_VALUES [[LOAD]](<8 x s32>) ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -7461,7 +7496,7 @@ body: | ; SI-LABEL: name: test_load_global_v5s16_align16 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; SI: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF @@ -7504,7 +7539,7 @@ body: | ; SI: $vgpr2 = COPY [[BITCAST7]](<2 x s16>) ; CI-HSA-LABEL: name: test_load_global_v5s16_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; CI-HSA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF @@ -7547,7 +7582,7 @@ body: | ; CI-HSA: $vgpr2 = COPY [[BITCAST7]](<2 x s16>) ; CI-MESA-LABEL: name: test_load_global_v5s16_align16 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; CI-MESA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF @@ -7590,7 +7625,7 @@ body: | ; CI-MESA: $vgpr2 = COPY [[BITCAST7]](<2 x s16>) ; VI-LABEL: name: test_load_global_v5s16_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; VI: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF @@ -7633,7 +7668,7 @@ body: | ; VI: $vgpr2 = COPY [[BITCAST7]](<2 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v5s16_align16 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; GFX9-HSA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF @@ -7663,7 +7698,7 @@ body: | ; GFX9-HSA: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC2]](<2 x s16>) ; GFX9-MESA-LABEL: name: test_load_global_v5s16_align16 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF @@ -8571,27 +8606,27 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-HSA-LABEL: name: test_load_global_v6s16_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 16, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-MESA-LABEL: name: test_load_global_v6s16_align16 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 16, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; VI-LABEL: name: test_load_global_v6s16_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 16, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v6s16_align16 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 16, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-MESA-LABEL: name: test_load_global_v6s16_align16 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 16, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -8617,27 +8652,27 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-HSA-LABEL: name: test_load_global_v6s16_align8 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 8, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-MESA-LABEL: name: test_load_global_v6s16_align8 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 8, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; VI-LABEL: name: test_load_global_v6s16_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 8, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v6s16_align8 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 8, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-MESA-LABEL: name: test_load_global_v6s16_align8 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 8, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 8, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -8663,27 +8698,27 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-HSA-LABEL: name: test_load_global_v6s16_align4 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-MESA-LABEL: name: test_load_global_v6s16_align4 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; VI-LABEL: name: test_load_global_v6s16_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v6s16_align4 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-MESA-LABEL: name: test_load_global_v6s16_align4 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 4, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -8738,7 +8773,7 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-HSA-LABEL: name: test_load_global_v6s16_align2 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 2, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-MESA-LABEL: name: test_load_global_v6s16_align2 @@ -8821,7 +8856,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v6s16_align2 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 2, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-MESA-LABEL: name: test_load_global_v6s16_align2 @@ -8955,7 +8990,7 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-HSA-LABEL: name: test_load_global_v6s16_align1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 1, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; CI-MESA-LABEL: name: test_load_global_v6s16_align1 @@ -9118,7 +9153,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v6s16_align1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<6 x s16>), align 1, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>) ; GFX9-MESA-LABEL: name: test_load_global_v6s16_align1 @@ -9213,7 +9248,7 @@ body: | ; SI-LABEL: name: test_load_global_v7s16_align16 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; SI: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF @@ -9265,7 +9300,7 @@ body: | ; SI: $vgpr3 = COPY [[BITCAST9]](<2 x s16>) ; CI-HSA-LABEL: name: test_load_global_v7s16_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; CI-HSA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF @@ -9317,7 +9352,7 @@ body: | ; CI-HSA: $vgpr3 = COPY [[BITCAST9]](<2 x s16>) ; CI-MESA-LABEL: name: test_load_global_v7s16_align16 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; CI-MESA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF @@ -9369,7 +9404,7 @@ body: | ; CI-MESA: $vgpr3 = COPY [[BITCAST9]](<2 x s16>) ; VI-LABEL: name: test_load_global_v7s16_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; VI: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF @@ -9421,7 +9456,7 @@ body: | ; VI: $vgpr3 = COPY [[BITCAST9]](<2 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v7s16_align16 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; GFX9-HSA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF @@ -9456,7 +9491,7 @@ body: | ; GFX9-HSA: $vgpr3 = COPY [[BUILD_VECTOR_TRUNC3]](<2 x s16>) ; GFX9-MESA-LABEL: name: test_load_global_v7s16_align16 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF ; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF @@ -11147,32 +11182,32 @@ body: | ; SI-LABEL: name: test_load_global_v8s16_align16 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; CI-HSA-LABEL: name: test_load_global_v8s16_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; CI-MESA-LABEL: name: test_load_global_v8s16_align16 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; VI-LABEL: name: test_load_global_v8s16_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v8s16_align16 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9-MESA-LABEL: name: test_load_global_v8s16_align16 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -11188,32 +11223,32 @@ body: | ; SI-LABEL: name: test_load_global_v8s16_align8 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; CI-HSA-LABEL: name: test_load_global_v8s16_align8 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; CI-MESA-LABEL: name: test_load_global_v8s16_align8 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; VI-LABEL: name: test_load_global_v8s16_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v8s16_align8 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) ; GFX9-MESA-LABEL: name: test_load_global_v8s16_align8 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 8, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<8 x s16>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -12793,32 +12828,32 @@ body: | ; SI-LABEL: name: test_load_global_v2sp1_align16 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-HSA-LABEL: name: test_load_global_v2sp1_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-MESA-LABEL: name: test_load_global_v2sp1_align16 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI-LABEL: name: test_load_global_v2sp1_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-HSA-LABEL: name: test_load_global_v2sp1_align16 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-MESA-LABEL: name: test_load_global_v2sp1_align16 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -14708,32 +14743,32 @@ body: | ; SI-LABEL: name: test_load_global_v2s128_align32 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; CI-HSA-LABEL: name: test_load_global_v2s128_align32 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; CI-MESA-LABEL: name: test_load_global_v2s128_align32 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; VI-LABEL: name: test_load_global_v2s128_align32 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; GFX9-HSA-LABEL: name: test_load_global_v2s128_align32 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) ; GFX9-MESA-LABEL: name: test_load_global_v2s128_align32 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x s128>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x s128>) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<2 x s128>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -14749,32 +14784,32 @@ body: | ; SI-LABEL: name: test_load_global_v2p1_align16 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-HSA-LABEL: name: test_load_global_v2p1_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-MESA-LABEL: name: test_load_global_v2p1_align16 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI-LABEL: name: test_load_global_v2p1_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-HSA-LABEL: name: test_load_global_v2p1_align16 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-MESA-LABEL: name: test_load_global_v2p1_align16 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -14790,32 +14825,32 @@ body: | ; SI-LABEL: name: test_load_global_v2p1_align8 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-HSA-LABEL: name: test_load_global_v2p1_align8 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-MESA-LABEL: name: test_load_global_v2p1_align8 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI-LABEL: name: test_load_global_v2p1_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-HSA-LABEL: name: test_load_global_v2p1_align8 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-MESA-LABEL: name: test_load_global_v2p1_align8 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 8, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 8, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -14831,32 +14866,32 @@ body: | ; SI-LABEL: name: test_load_global_v2p1_align4 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-HSA-LABEL: name: test_load_global_v2p1_align4 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-MESA-LABEL: name: test_load_global_v2p1_align4 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; VI-LABEL: name: test_load_global_v2p1_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-HSA-LABEL: name: test_load_global_v2p1_align4 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-MESA-LABEL: name: test_load_global_v2p1_align4 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -14974,7 +15009,7 @@ body: | ; SI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-HSA-LABEL: name: test_load_global_v2p1_align1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 1, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; CI-MESA-LABEL: name: test_load_global_v2p1_align1 @@ -15183,7 +15218,7 @@ body: | ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-HSA-LABEL: name: test_load_global_v2p1_align1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 1, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 1, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-MESA-LABEL: name: test_load_global_v2p1_align1 @@ -15301,32 +15336,32 @@ body: | ; SI-LABEL: name: test_load_global_v4p1_align8 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1) ; SI: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; CI-HSA-LABEL: name: test_load_global_v4p1_align8 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; CI-MESA-LABEL: name: test_load_global_v4p1_align8 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; VI-LABEL: name: test_load_global_v4p1_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; GFX9-HSA-LABEL: name: test_load_global_v4p1_align8 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) ; GFX9-MESA-LABEL: name: test_load_global_v4p1_align8 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x p1>), align 8, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[COPY]](p1) :: (load (<8 x s32>), align 8, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(<4 x p1>) = G_BITCAST [[LOAD]](<8 x s32>) ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BITCAST]](<4 x p1>) %0:_(p1) = COPY $vgpr0_vgpr1 @@ -16692,11 +16727,11 @@ body: | ; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY26]](s96) ; CI-HSA-LABEL: name: test_global_v2s96_align1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 1, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 1, addrspace 1) + ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 1) ; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17006,11 +17041,11 @@ body: | ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY26]](s96) ; GFX9-HSA-LABEL: name: test_global_v2s96_align1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 1, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 1, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 1, addrspace 1) + ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 1) ; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17258,11 +17293,11 @@ body: | ; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY14]](s96) ; CI-HSA-LABEL: name: test_global_v2s96_align2 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 2, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 2, addrspace 1) + ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 1) ; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17420,11 +17455,11 @@ body: | ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY14]](s96) ; GFX9-HSA-LABEL: name: test_global_v2s96_align2 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 2, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 2, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 2, addrspace 1) + ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 1) ; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17542,11 +17577,11 @@ body: | ; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-HSA-LABEL: name: test_global_v2s96_align4 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17554,11 +17589,11 @@ body: | ; CI-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-MESA-LABEL: name: test_global_v2s96_align4 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; CI-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; CI-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; CI-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17566,11 +17601,11 @@ body: | ; CI-MESA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; VI-LABEL: name: test_global_v2s96_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; VI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17578,11 +17613,11 @@ body: | ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-HSA-LABEL: name: test_global_v2s96_align4 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17590,11 +17625,11 @@ body: | ; GFX9-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-MESA-LABEL: name: test_global_v2s96_align4 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 4, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 4, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; GFX9-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17634,11 +17669,11 @@ body: | ; SI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-HSA-LABEL: name: test_global_v2s96_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; CI-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; CI-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; CI-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17646,11 +17681,11 @@ body: | ; CI-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-MESA-LABEL: name: test_global_v2s96_align16 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; CI-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; CI-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; CI-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; CI-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; CI-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; CI-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; CI-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17658,11 +17693,11 @@ body: | ; CI-MESA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; VI-LABEL: name: test_global_v2s96_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; VI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; VI: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; VI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; VI: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17670,11 +17705,11 @@ body: | ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-HSA-LABEL: name: test_global_v2s96_align16 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; GFX9-HSA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-HSA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-HSA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; GFX9-HSA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; GFX9-HSA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-HSA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-HSA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17682,11 +17717,11 @@ body: | ; GFX9-HSA: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-MESA-LABEL: name: test_global_v2s96_align16 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (s96), align 16, addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1) ; GFX9-MESA: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-MESA: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; GFX9-MESA: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) - ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (s96) from unknown-address + 12, align 4, addrspace 1) + ; GFX9-MESA: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 1) ; GFX9-MESA: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-MESA: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-MESA: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17708,7 +17743,7 @@ body: | ; SI-LABEL: name: test_load_global_v32s1_align4 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 @@ -17808,7 +17843,7 @@ body: | ; SI: $vgpr0 = COPY [[TRUNC]](<32 x s1>) ; CI-HSA-LABEL: name: test_load_global_v32s1_align4 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 @@ -17908,7 +17943,7 @@ body: | ; CI-HSA: $vgpr0 = COPY [[TRUNC]](<32 x s1>) ; CI-MESA-LABEL: name: test_load_global_v32s1_align4 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 @@ -18008,7 +18043,7 @@ body: | ; CI-MESA: $vgpr0 = COPY [[TRUNC]](<32 x s1>) ; VI-LABEL: name: test_load_global_v32s1_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 @@ -18108,7 +18143,7 @@ body: | ; VI: $vgpr0 = COPY [[TRUNC]](<32 x s1>) ; GFX9-HSA-LABEL: name: test_load_global_v32s1_align4 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 @@ -18224,7 +18259,7 @@ body: | ; GFX9-HSA: $vgpr0 = COPY [[TRUNC]](<32 x s1>) ; GFX9-MESA-LABEL: name: test_load_global_v32s1_align4 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<32 x s1>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 @@ -18351,7 +18386,7 @@ body: | ; SI-LABEL: name: test_load_global_v8s4_align4 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) + ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 @@ -18379,7 +18414,7 @@ body: | ; SI: $vgpr0 = COPY [[TRUNC]](<8 x s4>) ; CI-HSA-LABEL: name: test_load_global_v8s4_align4 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) + ; CI-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; CI-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; CI-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 @@ -18407,7 +18442,7 @@ body: | ; CI-HSA: $vgpr0 = COPY [[TRUNC]](<8 x s4>) ; CI-MESA-LABEL: name: test_load_global_v8s4_align4 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) + ; CI-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; CI-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; CI-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 @@ -18435,7 +18470,7 @@ body: | ; CI-MESA: $vgpr0 = COPY [[TRUNC]](<8 x s4>) ; VI-LABEL: name: test_load_global_v8s4_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 @@ -18463,7 +18498,7 @@ body: | ; VI: $vgpr0 = COPY [[TRUNC]](<8 x s4>) ; GFX9-HSA-LABEL: name: test_load_global_v8s4_align4 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) + ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; GFX9-HSA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; GFX9-HSA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-HSA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 @@ -18495,7 +18530,7 @@ body: | ; GFX9-HSA: $vgpr0 = COPY [[TRUNC]](<8 x s4>) ; GFX9-MESA-LABEL: name: test_load_global_v8s4_align4 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (<8 x s4>), addrspace 1) + ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s32), addrspace 1) ; GFX9-MESA: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 ; GFX9-MESA: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-MESA: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir index 137970d..e332da1 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir @@ -2717,7 +2717,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align16 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 1, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX10-LABEL: name: test_load_local_s96_align16 @@ -2939,7 +2939,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align8 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 8, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 8, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX10-LABEL: name: test_load_local_s96_align8 @@ -3025,7 +3025,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align4 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 4, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 4, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX10-LABEL: name: test_load_local_s96_align4 @@ -3256,7 +3256,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align2 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 2, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 2, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX10-LABEL: name: test_load_local_s96_align2 @@ -3740,7 +3740,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED-LABEL: name: test_load_local_s96_align1 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 1, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96) ; GFX10-LABEL: name: test_load_local_s96_align1 @@ -4419,7 +4419,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align16 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 1, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX10-LABEL: name: test_load_local_s128_align16 @@ -4655,32 +4655,32 @@ body: | ; CI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; CI-DS128-LABEL: name: test_load_local_s128_align8 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; CI-DS128: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) + ; CI-DS128: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3) ; CI-DS128: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; CI-DS128: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; VI-LABEL: name: test_load_local_s128_align8 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) + ; VI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3) ; VI: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; VI: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-LABEL: name: test_load_local_s128_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) + ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3) ; GFX9: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align8 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX10-LABEL: name: test_load_local_s128_align8 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX10: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) + ; GFX10: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3) ; GFX10: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX10: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX10-UNALIGNED-LABEL: name: test_load_local_s128_align8 ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 8, addrspace 3) + ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 8, addrspace 3) ; GFX10-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX10-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) %0:_(p3) = COPY $vgpr0 @@ -4759,7 +4759,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align4 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 4, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 4, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX10-LABEL: name: test_load_local_s128_align4 @@ -5060,7 +5060,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align2 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 2, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 2, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX10-LABEL: name: test_load_local_s128_align2 @@ -5685,7 +5685,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX9-UNALIGNED-LABEL: name: test_load_local_s128_align1 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (s128), align 1, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](s128) ; GFX10-LABEL: name: test_load_local_s128_align1 @@ -7484,7 +7484,7 @@ body: | ; SI-LABEL: name: test_load_local_v2s8_align2 ; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) + ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -7505,7 +7505,7 @@ body: | ; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-LABEL: name: test_load_local_v2s8_align2 ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) + ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -7526,7 +7526,7 @@ body: | ; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-DS128-LABEL: name: test_load_local_v2s8_align2 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; CI-DS128: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) + ; CI-DS128: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-DS128: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-DS128: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -7547,7 +7547,7 @@ body: | ; CI-DS128: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_load_local_v2s8_align2 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -7566,7 +7566,7 @@ body: | ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-LABEL: name: test_load_local_v2s8_align2 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) + ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -7585,7 +7585,7 @@ body: | ; GFX9: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s8_align2 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -7604,7 +7604,7 @@ body: | ; GFX9-UNALIGNED: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX10-LABEL: name: test_load_local_v2s8_align2 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) + ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -7623,7 +7623,7 @@ body: | ; GFX10: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX10-UNALIGNED-LABEL: name: test_load_local_v2s8_align2 ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), addrspace 3) + ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), addrspace 3) ; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX10-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -7713,7 +7713,7 @@ body: | ; GFX9: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s8_align1 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<2 x s8>), align 1, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3) ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -8030,11 +8030,31 @@ body: | ; GFX9: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v3s8_align1 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s24) = G_LOAD [[COPY]](p3) :: (load (<3 x s8>), align 1, addrspace 3) - ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(<3 x s8>) = G_BITCAST [[LOAD]](s24) - ; GFX9-UNALIGNED: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF - ; GFX9-UNALIGNED: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) - ; GFX9-UNALIGNED: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC]], [[BITCAST]](<3 x s8>), 0 + ; GFX9-UNALIGNED: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p3) :: (load (s16), align 1, addrspace 3) + ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 2 + ; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p3) :: (load (s8) from unknown-address + 2, addrspace 3) + ; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; GFX9-UNALIGNED: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[LOAD]], [[C1]](s32) + ; GFX9-UNALIGNED: [[OR:%[0-9]+]]:_(s32) = G_OR [[SHL]], [[ZEXTLOAD]] + ; GFX9-UNALIGNED: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 + ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C2]](s32) + ; GFX9-UNALIGNED: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C1]](s32) + ; GFX9-UNALIGNED: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 + ; GFX9-UNALIGNED: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[OR]], [[C3]](s32) + ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s32) = COPY [[OR]](s32) + ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; GFX9-UNALIGNED: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32) + ; GFX9-UNALIGNED: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) + ; GFX9-UNALIGNED: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9-UNALIGNED: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[DEF]](s32) + ; GFX9-UNALIGNED: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; GFX9-UNALIGNED: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[DEF1]](<2 x s16>) + ; GFX9-UNALIGNED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>) + ; GFX9-UNALIGNED: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[UV]](<3 x s16>) + ; GFX9-UNALIGNED: [[DEF2:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; GFX9-UNALIGNED: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF2]](<4 x s32>) + ; GFX9-UNALIGNED: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[TRUNC1]], [[TRUNC]](<3 x s8>), 0 ; GFX9-UNALIGNED: $vgpr0 = COPY [[INSERT]](<4 x s8>) ; GFX10-LABEL: name: test_load_local_v3s8_align1 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 @@ -8097,7 +8117,7 @@ body: | ; SI-LABEL: name: test_load_local_v4s8_align4 ; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) + ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -8113,7 +8133,7 @@ body: | ; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI-LABEL: name: test_load_local_v4s8_align4 ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) + ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -8129,7 +8149,7 @@ body: | ; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI-DS128-LABEL: name: test_load_local_v4s8_align4 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; CI-DS128: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) + ; CI-DS128: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3) ; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-DS128: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI-DS128: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -8145,7 +8165,7 @@ body: | ; CI-DS128: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI-LABEL: name: test_load_local_v4s8_align4 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -8161,7 +8181,7 @@ body: | ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-LABEL: name: test_load_local_v4s8_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) + ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -8179,7 +8199,7 @@ body: | ; GFX9: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v4s8_align4 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3) ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -8197,7 +8217,7 @@ body: | ; GFX9-UNALIGNED: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX10-LABEL: name: test_load_local_v4s8_align4 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) + ; GFX10: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX10: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -8215,7 +8235,7 @@ body: | ; GFX10: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX10-UNALIGNED-LABEL: name: test_load_local_v4s8_align4 ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (<4 x s8>), addrspace 3) + ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p3) :: (load (s32), addrspace 3) ; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX10-UNALIGNED: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -8244,7 +8264,7 @@ body: | ; SI-LABEL: name: test_load_local_v8s8_align8 ; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; SI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) + ; SI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3) ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -8271,7 +8291,7 @@ body: | ; SI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; CI-LABEL: name: test_load_local_v8s8_align8 ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) + ; CI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3) ; CI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -8298,7 +8318,7 @@ body: | ; CI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; CI-DS128-LABEL: name: test_load_local_v8s8_align8 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; CI-DS128: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) + ; CI-DS128: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3) ; CI-DS128: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI-DS128: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -8325,7 +8345,7 @@ body: | ; CI-DS128: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; VI-LABEL: name: test_load_local_v8s8_align8 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) + ; VI: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3) ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -8352,7 +8372,7 @@ body: | ; VI: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<8 x s8>) ; GFX9-LABEL: name: test_load_local_v8s8_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) + ; GFX9: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3) ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -8383,7 +8403,7 @@ body: | ; GFX9: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v8s8_align8 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3) ; GFX9-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -8414,7 +8434,7 @@ body: | ; GFX9-UNALIGNED: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>) ; GFX10-LABEL: name: test_load_local_v8s8_align8 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX10: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) + ; GFX10: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3) ; GFX10: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -8445,7 +8465,7 @@ body: | ; GFX10: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS2]](<8 x s8>) ; GFX10-UNALIGNED-LABEL: name: test_load_local_v8s8_align8 ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<8 x s8>), addrspace 3) + ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x s32>), addrspace 3) ; GFX10-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>) ; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX10-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -8818,7 +8838,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<16 x s8>) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v16s8_align16 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<16 x s8>), align 1, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 1, addrspace 3) ; GFX9-UNALIGNED: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9-UNALIGNED: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32) @@ -14824,7 +14844,7 @@ body: | ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2p1_align4 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<2 x p1>), align 4, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p3) :: (load (<4 x s32>), align 4, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(<2 x p1>) = G_BITCAST [[LOAD]](<4 x s32>) ; GFX9-UNALIGNED: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BITCAST]](<2 x p1>) ; GFX10-LABEL: name: test_load_local_v2p1_align4 @@ -16295,11 +16315,11 @@ body: | ; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY26]](s96) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align1 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 1, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 1, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) - ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (s96) from unknown-address + 12, align 1, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 1, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -16996,11 +17016,11 @@ body: | ; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY14]](s96) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align2 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 2, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 2, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) - ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (s96) from unknown-address + 12, align 2, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 2, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17277,11 +17297,11 @@ body: | ; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align4 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 4, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 4, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) - ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (s96) from unknown-address + 12, align 4, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17387,7 +17407,7 @@ body: | ; CI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; CI-DS128-LABEL: name: test_load_local_v2s96_align16 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; CI-DS128: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) + ; CI-DS128: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3) ; CI-DS128: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; CI-DS128: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; CI-DS128: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) @@ -17404,7 +17424,7 @@ body: | ; CI-DS128: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; VI-LABEL: name: test_load_local_v2s96_align16 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) + ; VI: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3) ; VI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; VI: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) @@ -17421,7 +17441,7 @@ body: | ; VI: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-LABEL: name: test_load_local_v2s96_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) + ; GFX9: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3) ; GFX9: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) @@ -17438,11 +17458,11 @@ body: | ; GFX9: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX9-UNALIGNED-LABEL: name: test_load_local_v2s96_align16 ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX9-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX9-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) - ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (s96) from unknown-address + 12, align 4, addrspace 3) + ; GFX9-UNALIGNED: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p3) :: (load (<3 x s32>) from unknown-address + 12, align 4, addrspace 3) ; GFX9-UNALIGNED: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD1]](<3 x s32>) ; GFX9-UNALIGNED: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96) ; GFX9-UNALIGNED: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96) @@ -17450,7 +17470,7 @@ body: | ; GFX9-UNALIGNED: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX10-LABEL: name: test_load_local_v2s96_align16 ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX10: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) + ; GFX10: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3) ; GFX10: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX10: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) @@ -17467,7 +17487,7 @@ body: | ; GFX10: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96) ; GFX10-UNALIGNED-LABEL: name: test_load_local_v2s96_align16 ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 - ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (s96), align 16, addrspace 3) + ; GFX10-UNALIGNED: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p3) :: (load (<3 x s32>), align 16, addrspace 3) ; GFX10-UNALIGNED: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>) ; GFX10-UNALIGNED: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 12 ; GFX10-UNALIGNED: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32) diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir index 47443d8..1856200 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir @@ -4526,7 +4526,7 @@ body: | ; SI-LABEL: name: test_load_private_v2s8_align2 ; SI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 - ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<2 x s8>), addrspace 5) + ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4547,7 +4547,7 @@ body: | ; SI: $vgpr0 = COPY [[ANYEXT]](s32) ; CI-LABEL: name: test_load_private_v2s8_align2 ; CI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 - ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<2 x s8>), addrspace 5) + ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4568,7 +4568,7 @@ body: | ; CI: $vgpr0 = COPY [[ANYEXT]](s32) ; VI-LABEL: name: test_load_private_v2s8_align2 ; VI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<2 x s8>), addrspace 5) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4587,7 +4587,7 @@ body: | ; VI: $vgpr0 = COPY [[ANYEXT]](s32) ; GFX9-LABEL: name: test_load_private_v2s8_align2 ; GFX9: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 - ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<2 x s8>), addrspace 5) + ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s16), addrspace 5) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4857,7 +4857,7 @@ body: | ; SI-LABEL: name: test_load_private_v4s8_align4 ; SI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 - ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<4 x s8>), addrspace 5) + ; SI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4873,7 +4873,7 @@ body: | ; SI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; CI-LABEL: name: test_load_private_v4s8_align4 ; CI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 - ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<4 x s8>), addrspace 5) + ; CI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5) ; CI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; CI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4889,7 +4889,7 @@ body: | ; CI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; VI-LABEL: name: test_load_private_v4s8_align4 ; VI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 - ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<4 x s8>), addrspace 5) + ; VI: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -4905,7 +4905,7 @@ body: | ; VI: $vgpr0 = COPY [[TRUNC]](<4 x s8>) ; GFX9-LABEL: name: test_load_private_v4s8_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 - ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (<4 x s8>), addrspace 5) + ; GFX9: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), addrspace 5) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[LOAD]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir index 8a16437..11d8d28 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir @@ -3618,7 +3618,7 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 1, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1) ; VI-LABEL: name: test_store_global_v8s16_align1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 @@ -3696,7 +3696,7 @@ body: | ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 1, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (<8 x s16>), align 1, addrspace 1) @@ -3750,7 +3750,7 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 2, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1) ; VI-LABEL: name: test_store_global_v8s16_align2 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 @@ -3793,7 +3793,7 @@ body: | ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 2, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (<8 x s16>), align 2, addrspace 1) @@ -3809,22 +3809,22 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 4, addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; CI-LABEL: name: test_store_global_v8s16_align4 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 4, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; VI-LABEL: name: test_store_global_v8s16_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 4, addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; GFX9-LABEL: name: test_store_global_v8s16_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 4, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (<8 x s16>), align 4, addrspace 1) @@ -3840,22 +3840,22 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 8, addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) ; CI-LABEL: name: test_store_global_v8s16_align8 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 8, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) ; VI-LABEL: name: test_store_global_v8s16_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 8, addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) ; GFX9-LABEL: name: test_store_global_v8s16_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), align 8, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (<8 x s16>), align 8, addrspace 1) @@ -3871,22 +3871,22 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; CI-LABEL: name: test_store_global_v8s16_align16 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; VI-LABEL: name: test_store_global_v8s16_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; GFX9-LABEL: name: test_store_global_v8s16_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<8 x s16>) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<8 x s16>), addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<8 x s16>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (<8 x s16>), align 16, addrspace 1) @@ -3975,7 +3975,7 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 1, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1) ; VI-LABEL: name: test_store_global_v2p0_align1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 @@ -4053,7 +4053,7 @@ body: | ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 1, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (<2 x p0>), align 1, addrspace 1) @@ -4107,7 +4107,7 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 2, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1) ; VI-LABEL: name: test_store_global_v2p0_align2 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 @@ -4150,7 +4150,7 @@ body: | ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 2, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (<2 x p0>), align 2, addrspace 1) @@ -4166,22 +4166,22 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 4, addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; CI-LABEL: name: test_store_global_v2p0_align4 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 4, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; VI-LABEL: name: test_store_global_v2p0_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 4, addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; GFX9-LABEL: name: test_store_global_v2p0_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 4, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (<2 x p0>), align 4, addrspace 1) @@ -4197,22 +4197,22 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 8, addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) ; CI-LABEL: name: test_store_global_v2p0_align8 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 8, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) ; VI-LABEL: name: test_store_global_v2p0_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 8, addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) ; GFX9-LABEL: name: test_store_global_v2p0_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), align 8, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (<2 x p0>), align 8, addrspace 1) @@ -4228,22 +4228,22 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; CI-LABEL: name: test_store_global_v2p0_align16 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; VI-LABEL: name: test_store_global_v2p0_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; GFX9-LABEL: name: test_store_global_v2p0_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](<2 x p0>) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<2 x p0>), addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<2 x p0>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (<2 x p0>), align 16, addrspace 1) @@ -4318,7 +4318,7 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 1, addrspace 1) + ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1) ; VI-LABEL: name: test_store_global_s96_align1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 @@ -4382,7 +4382,7 @@ body: | ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 1, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 1, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s96) = COPY $vgpr2_vgpr3_vgpr4 G_STORE %1, %0 :: (store (s96), align 1, addrspace 1) @@ -4429,7 +4429,7 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 2, addrspace 1) + ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1) ; VI-LABEL: name: test_store_global_s96_align2 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 @@ -4465,7 +4465,7 @@ body: | ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 2, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s96) = COPY $vgpr2_vgpr3_vgpr4 G_STORE %1, %0 :: (store (s96), align 2, addrspace 1) @@ -4491,17 +4491,17 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 4, addrspace 1) + ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1) ; VI-LABEL: name: test_store_global_s96_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 4, addrspace 1) + ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1) ; GFX9-LABEL: name: test_store_global_s96_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 4, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s96) = COPY $vgpr2_vgpr3_vgpr4 G_STORE %1, %0 :: (store (s96), align 4, addrspace 1) @@ -4527,17 +4527,17 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 8, addrspace 1) + ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1) ; VI-LABEL: name: test_store_global_s96_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 8, addrspace 1) + ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1) ; GFX9-LABEL: name: test_store_global_s96_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 8, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s96) = COPY $vgpr2_vgpr3_vgpr4 G_STORE %1, %0 :: (store (s96), align 8, addrspace 1) @@ -4563,17 +4563,17 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; CI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 16, addrspace 1) + ; CI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1) ; VI-LABEL: name: test_store_global_s96_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 16, addrspace 1) + ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1) ; GFX9-LABEL: name: test_store_global_s96_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4 ; GFX9: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96) - ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (s96), align 16, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s96) = COPY $vgpr2_vgpr3_vgpr4 G_STORE %1, %0 :: (store (s96), align 16, addrspace 1) @@ -4662,7 +4662,7 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 1, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1) ; VI-LABEL: name: test_store_global_s128_align1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 @@ -4740,7 +4740,7 @@ body: | ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 1, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (s128), align 1, addrspace 1) @@ -4794,7 +4794,7 @@ body: | ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 2, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1) ; VI-LABEL: name: test_store_global_s128_align2 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 @@ -4837,7 +4837,7 @@ body: | ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 2, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (s128), align 2, addrspace 1) @@ -4853,22 +4853,22 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 4, addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; CI-LABEL: name: test_store_global_s128_align4 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 4, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; VI-LABEL: name: test_store_global_s128_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 4, addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) ; GFX9-LABEL: name: test_store_global_s128_align4 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 4, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (s128), align 4, addrspace 1) @@ -4884,22 +4884,22 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 8, addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) ; CI-LABEL: name: test_store_global_s128_align8 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 8, addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) ; VI-LABEL: name: test_store_global_s128_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 8, addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) ; GFX9-LABEL: name: test_store_global_s128_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), align 8, addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (s128), align 8, addrspace 1) @@ -4915,22 +4915,22 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; CI-LABEL: name: test_store_global_s128_align16 ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; CI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) + ; CI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; VI-LABEL: name: test_store_global_s128_align16 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; GFX9-LABEL: name: test_store_global_s128_align16 ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; GFX9: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) + ; GFX9: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (s128), align 16, addrspace 1) diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store.mir index db38191..1e8b5fe 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store.mir @@ -252,12 +252,12 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; VI-LABEL: name: test_truncstore_global_s128_to_s8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (s128), addrspace 1) @@ -355,7 +355,7 @@ body: | ; VI: [[COPY:%[0-9]+]]:_(s96) = COPY $vgpr0_vgpr1_vgpr2 ; VI: [[COPY1:%[0-9]+]]:_(p1) = COPY $vgpr3_vgpr4 ; VI: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY]](s96) - ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY1]](p1) :: (store (s96), align 16, addrspace 1) + ; VI: G_STORE [[BITCAST]](<3 x s32>), [[COPY1]](p1) :: (store (<3 x s32>), align 16, addrspace 1) %0:_(s96) = COPY $vgpr0_vgpr1_vgpr2 %1:_(p1) = COPY $vgpr3_vgpr4 @@ -372,12 +372,12 @@ body: | ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; SI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) + ; SI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) ; VI-LABEL: name: test_store_global_i128 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 ; VI: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY1]](s128) - ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (s128), addrspace 1) + ; VI: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(s128) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 G_STORE %1, %0 :: (store (s128), addrspace 1) @@ -456,7 +456,7 @@ body: | ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) - ; SI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s8>), addrspace 1) + ; SI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (s16), addrspace 1) ; VI-LABEL: name: test_store_global_v2s8_align2 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[DEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF @@ -470,7 +470,7 @@ body: | ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16) ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) - ; VI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s8>), addrspace 1) + ; VI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (s16), addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<2 x s8>) = G_IMPLICIT_DEF G_STORE %1, %0 :: (store (<2 x s8>), addrspace 1, align 2) @@ -498,7 +498,7 @@ body: | ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) - ; SI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s8>), align 4, addrspace 1) + ; SI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1) ; VI-LABEL: name: test_store_global_v2s8_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[DEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF @@ -512,7 +512,7 @@ body: | ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16) ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) - ; VI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s8>), align 4, addrspace 1) + ; VI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (s16), align 4, addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<2 x s8>) = G_IMPLICIT_DEF G_STORE %1, %0 :: (store (<2 x s8>), addrspace 1, align 4) @@ -864,7 +864,7 @@ body: | ; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32) ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]] - ; SI: G_STORE [[OR2]](s32), [[COPY]](p1) :: (store (<4 x s8>), addrspace 1) + ; SI: G_STORE [[OR2]](s32), [[COPY]](p1) :: (store (s32), addrspace 1) ; VI-LABEL: name: test_store_global_v4s8_align4 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 @@ -887,7 +887,7 @@ body: | ; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32) ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]] - ; VI: G_STORE [[OR2]](s32), [[COPY]](p1) :: (store (<4 x s8>), addrspace 1) + ; VI: G_STORE [[OR2]](s32), [[COPY]](p1) :: (store (s32), addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<4 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5 %2:_(<4 x s8>) = G_TRUNC %1 @@ -903,34 +903,12 @@ body: | ; SI-LABEL: name: test_truncstore_global_v2s8_to_1_align1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; SI: [[DEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF - ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<2 x s32>) - ; SI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 - ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32) - ; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]] - ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 - ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 - ; SI: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV1]](s32) - ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C2]] - ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32) - ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) - ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] - ; SI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) - ; SI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s4>), addrspace 1) + ; SI: [[DEF:%[0-9]+]]:_(<2 x s8>) = G_IMPLICIT_DEF + ; SI: G_STORE [[DEF]](<2 x s8>), [[COPY]](p1) :: (store (<2 x s4>), addrspace 1) ; VI-LABEL: name: test_truncstore_global_v2s8_to_1_align1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 - ; VI: [[DEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF - ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<2 x s32>) - ; VI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 - ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32) - ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]] - ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32) - ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]] - ; VI: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 - ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16) - ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] - ; VI: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) - ; VI: G_STORE [[ANYEXT]](s32), [[COPY]](p1) :: (store (<2 x s4>), addrspace 1) + ; VI: [[DEF:%[0-9]+]]:_(<2 x s8>) = G_IMPLICIT_DEF + ; VI: G_STORE [[DEF]](<2 x s8>), [[COPY]](p1) :: (store (<2 x s4>), addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<2 x s8>) = G_IMPLICIT_DEF G_STORE %1, %0 :: (store (<2 x s4>), addrspace 1, align 1) @@ -946,63 +924,13 @@ body: | ; SI-LABEL: name: test_truncstore_global_v3s8_to_1_align1 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 - ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) - ; SI: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; SI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 - ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32) - ; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]] - ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 - ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C1]](s32) - ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 - ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[UV1]](s32) - ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] - ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY2]](s32) - ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) - ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] - ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32) - ; SI: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]] - ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[DEF]](s32) - ; SI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] - ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C1]](s32) - ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[SHL1]](s32) - ; SI: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[TRUNC3]] - ; SI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16) - ; SI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16) - ; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32) - ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]] - ; SI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[OR2]](s32) - ; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 63 - ; SI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C4]] - ; SI: G_STORE [[AND4]](s32), [[COPY]](p1) :: (store (s8), addrspace 1) + ; SI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY1]](<3 x s32>) + ; SI: G_STORE [[TRUNC]](<3 x s8>), [[COPY]](p1) :: (store (<3 x s2>), addrspace 1) ; VI-LABEL: name: test_truncstore_global_v3s8_to_1_align1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 - ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) - ; VI: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; VI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 - ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32) - ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]] - ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32) - ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]] - ; VI: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 - ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16) - ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] - ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32) - ; VI: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]] - ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32) - ; VI: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C]] - ; VI: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C1]](s16) - ; VI: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]] - ; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16) - ; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16) - ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C2]](s32) - ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]] - ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[OR2]](s32) - ; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 63 - ; VI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C3]] - ; VI: G_STORE [[AND4]](s32), [[COPY]](p1) :: (store (s8), addrspace 1) + ; VI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY1]](<3 x s32>) + ; VI: G_STORE [[TRUNC]](<3 x s8>), [[COPY]](p1) :: (store (<3 x s2>), addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 %2:_(<3 x s8>) = G_TRUNC %1 @@ -1019,63 +947,13 @@ body: | ; SI-LABEL: name: test_truncstore_global_v3s8_to_2_align2 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 - ; SI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) - ; SI: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; SI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 - ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32) - ; SI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]] - ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 - ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C1]](s32) - ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 - ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[UV1]](s32) - ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] - ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY2]](s32) - ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL]](s32) - ; SI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] - ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32) - ; SI: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]] - ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[DEF]](s32) - ; SI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] - ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C1]](s32) - ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[SHL1]](s32) - ; SI: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[TRUNC3]] - ; SI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16) - ; SI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16) - ; SI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32) - ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]] - ; SI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[OR2]](s32) - ; SI: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4095 - ; SI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C4]] - ; SI: G_STORE [[AND4]](s32), [[COPY]](p1) :: (store (s16), addrspace 1) + ; SI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY1]](<3 x s32>) + ; SI: G_STORE [[TRUNC]](<3 x s8>), [[COPY]](p1) :: (store (<3 x s4>), addrspace 1) ; VI-LABEL: name: test_truncstore_global_v3s8_to_2_align2 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 - ; VI: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) - ; VI: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; VI: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 - ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32) - ; VI: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]] - ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32) - ; VI: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]] - ; VI: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 - ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16) - ; VI: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]] - ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32) - ; VI: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]] - ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32) - ; VI: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C]] - ; VI: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C1]](s16) - ; VI: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]] - ; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16) - ; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16) - ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C2]](s32) - ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]] - ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[OR2]](s32) - ; VI: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4095 - ; VI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C3]] - ; VI: G_STORE [[AND4]](s32), [[COPY]](p1) :: (store (s16), addrspace 1) + ; VI: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY1]](<3 x s32>) + ; VI: G_STORE [[TRUNC]](<3 x s8>), [[COPY]](p1) :: (store (<3 x s4>), addrspace 1) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4 %2:_(<3 x s8>) = G_TRUNC %1 -- 2.7.4