Index: llvm/include/llvm/CodeGen/GlobalISel/LegalizerHelper.h =================================================================== --- llvm/include/llvm/CodeGen/GlobalISel/LegalizerHelper.h +++ llvm/include/llvm/CodeGen/GlobalISel/LegalizerHelper.h @@ -154,6 +154,10 @@ /// def by inserting a G_BITCAST from \p CastTy void bitcastDst(MachineInstr &MI, LLT CastTy, unsigned OpIdx); + /// Widen \p OrigReg to \p WideTy by merging to a wider type, padding with + /// G_IMPLICIT_DEF, and producing dead results. + Register widenWithUnmerge(LLT WideTy, Register OrigReg); + private: LegalizeResult widenScalarMergeValues(MachineInstr &MI, unsigned TypeIdx, LLT WideTy); Index: llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp =================================================================== --- llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp +++ llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp @@ -1296,10 +1296,8 @@ void LegalizerHelper::moreElementsVectorDst(MachineInstr &MI, LLT WideTy, unsigned OpIdx) { MachineOperand &MO = MI.getOperand(OpIdx); - Register DstExt = MRI.createGenericVirtualRegister(WideTy); MIRBuilder.setInsertPt(MIRBuilder.getMBB(), ++MIRBuilder.getInsertPt()); - MIRBuilder.buildExtract(MO, DstExt, 0); - MO.setReg(DstExt); + MO.setReg(widenWithUnmerge(WideTy, MO.getReg())); } void LegalizerHelper::moreElementsVectorSrc(MachineInstr &MI, LLT MoreTy, @@ -1467,6 +1465,40 @@ return Legalized; } +Register LegalizerHelper::widenWithUnmerge(LLT WideTy, Register OrigReg) { + Register WideReg = MRI.createGenericVirtualRegister(WideTy); + LLT OrigTy = MRI.getType(OrigReg); + LLT LCMTy = getLCMType(WideTy, OrigTy); + + const int NumMergeParts = LCMTy.getSizeInBits() / WideTy.getSizeInBits(); + const int NumUnmergeParts = LCMTy.getSizeInBits() / OrigTy.getSizeInBits(); + + Register UnmergeSrc = WideReg; + + // Create a merge to the LCM type, padding with undef + // %0:_(<3 x s32>) = G_FOO => <4 x s32> + // => + // %1:_(<4 x s32>) = G_FOO + // %2:_(<4 x s32>) = G_IMPLICIT_DEF + // %3:_(<12 x s32>) = G_CONCAT_VECTORS %1, %2, %2 + // %0:_(<3 x s32>), %4:_, %5:_, %6:_ = G_UNMERGE_VALUES %3 + if (NumMergeParts > 1) { + Register Undef = MIRBuilder.buildUndef(WideTy).getReg(0); + SmallVector MergeParts(NumMergeParts, Undef); + MergeParts[0] = WideReg; + UnmergeSrc = MIRBuilder.buildMerge(LCMTy, MergeParts).getReg(0); + } + + // Unmerge to the original register and pad with dead defs. + SmallVector UnmergeResults(NumUnmergeParts); + UnmergeResults[0] = OrigReg; + for (int I = 1; I != NumUnmergeParts; ++I) + UnmergeResults[I] = MRI.createGenericVirtualRegister(OrigTy); + + MIRBuilder.buildUnmerge(UnmergeResults, UnmergeSrc); + return WideReg; +} + LegalizerHelper::LegalizeResult LegalizerHelper::widenScalarUnmergeValues(MachineInstr &MI, unsigned TypeIdx, LLT WideTy) { Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-and.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-and.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-and.mir @@ -361,15 +361,18 @@ ; CHECK-LABEL: name: test_and_v3s16 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CHECK: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CHECK: [[AND:%[0-9]+]]:_(<4 x s16>) = G_AND [[INSERT]], [[INSERT1]] - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[AND]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT2]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[AND]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF @@ -403,27 +406,32 @@ ; CHECK-LABEL: name: test_and_v5s16 ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[DEF]](<6 x s16>), 0 ; CHECK: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[DEF1]](<6 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>), [[UV2:%[0-9]+]]:_(<5 x s16>), [[UV3:%[0-9]+]]:_(<5 x s16>), [[UV4:%[0-9]+]]:_(<5 x s16>), [[UV5:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s16>) ; CHECK: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<5 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<5 x s16>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT]](<6 x s16>) - ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>) + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF2]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV6:%[0-9]+]]:_(<5 x s16>), [[UV7:%[0-9]+]]:_(<5 x s16>), [[UV8:%[0-9]+]]:_(<5 x s16>), [[UV9:%[0-9]+]]:_(<5 x s16>), [[UV10:%[0-9]+]]:_(<5 x s16>), [[UV11:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<30 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV]](<5 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV6]](<5 x s16>), 0 + ; CHECK: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT]](<6 x s16>) + ; CHECK: [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>) ; CHECK: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV]](<3 x s16>), 0 - ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV12]](<3 x s16>), 0 + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV14]](<3 x s16>), 0 ; CHECK: [[AND:%[0-9]+]]:_(<4 x s16>) = G_AND [[INSERT2]], [[INSERT3]] - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[AND]](<4 x s16>), 0 - ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV1]](<3 x s16>), 0 - ; CHECK: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV3]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[AND]](<4 x s16>), [[DEF3]](<4 x s16>), [[DEF3]](<4 x s16>) + ; CHECK: [[UV16:%[0-9]+]]:_(<3 x s16>), [[UV17:%[0-9]+]]:_(<3 x s16>), [[UV18:%[0-9]+]]:_(<3 x s16>), [[UV19:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV13]](<3 x s16>), 0 + ; CHECK: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV15]](<3 x s16>), 0 ; CHECK: [[AND1:%[0-9]+]]:_(<4 x s16>) = G_AND [[INSERT4]], [[INSERT5]] - ; CHECK: [[EXTRACT3:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[AND1]](<4 x s16>), 0 - ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT2]](<3 x s16>), [[EXTRACT3]](<3 x s16>) - ; CHECK: [[EXTRACT4:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS3:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[AND1]](<4 x s16>), [[DEF3]](<4 x s16>), [[DEF3]](<4 x s16>) + ; CHECK: [[UV20:%[0-9]+]]:_(<3 x s16>), [[UV21:%[0-9]+]]:_(<3 x s16>), [[UV22:%[0-9]+]]:_(<3 x s16>), [[UV23:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<12 x s16>) + ; CHECK: [[CONCAT_VECTORS4:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV16]](<3 x s16>), [[UV20]](<3 x s16>) + ; CHECK: [[CONCAT_VECTORS5:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS4]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV24:%[0-9]+]]:_(<5 x s16>), [[UV25:%[0-9]+]]:_(<5 x s16>), [[UV26:%[0-9]+]]:_(<5 x s16>), [[UV27:%[0-9]+]]:_(<5 x s16>), [[UV28:%[0-9]+]]:_(<5 x s16>), [[UV29:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS5]](<30 x s16>) ; CHECK: [[DEF4:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT6:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF4]], [[EXTRACT4]](<5 x s16>), 0 + ; CHECK: [[INSERT6:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF4]], [[UV24]](<5 x s16>), 0 ; CHECK: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT6]](<8 x s16>) %0:_(<5 x s16>) = G_IMPLICIT_DEF %1:_(<5 x s16>) = G_IMPLICIT_DEF @@ -463,8 +471,9 @@ ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[ANYEXT6]], [[ANYEXT7]] ; CHECK: [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[AND3]](s32) ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s8>) = G_BUILD_VECTOR [[TRUNC]](s8), [[TRUNC1]](s8), [[TRUNC2]](s8), [[TRUNC3]](s8) - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s8>) = G_EXTRACT [[BUILD_VECTOR]](<4 x s8>), 0 - ; CHECK: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[EXTRACT]](<3 x s8>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<4 x s8>), [[DEF2]](<4 x s8>), [[DEF2]](<4 x s8>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s8>), [[UV9:%[0-9]+]]:_(<3 x s8>), [[UV10:%[0-9]+]]:_(<3 x s8>), [[UV11:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>) + ; CHECK: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV8]](<3 x s8>) ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT8]](<3 x s32>) %0:_(<3 x s8>) = G_IMPLICIT_DEF %1:_(<3 x s8>) = G_IMPLICIT_DEF Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ashr.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ashr.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ashr.mir @@ -863,14 +863,17 @@ ; GFX9: [[ASHR:%[0-9]+]]:_(<2 x s16>) = G_ASHR [[EXTRACT2]], [[EXTRACT4]](<2 x s16>) ; GFX9: [[ASHR1:%[0-9]+]]:_(s16) = G_ASHR [[EXTRACT3]], [[EXTRACT5]](s16) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT6:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT6]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT4]], [[ASHR]](<2 x s16>), 0 - ; GFX9: [[EXTRACT7:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT5]](<4 x s16>), 0 - ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT7]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT5]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT7:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT6]], [[ASHR1]](s16), 32 - ; GFX9: [[EXTRACT8:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT7]](<4 x s16>), 0 - ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT8]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT7]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT8]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-extract.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-extract.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-extract.mir @@ -522,11 +522,14 @@ ; CHECK-LABEL: name: extract_s8_v3s8_offset16 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF ; CHECK: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s8>) = G_EXTRACT [[TRUNC]](<4 x s8>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(s8), [[UV1:%[0-9]+]]:_(s8), [[UV2:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s8>) - ; CHECK: [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[UV]](s8) - ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s16) = G_ANYEXT [[UV1]](s8) - ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s16) = G_ANYEXT [[UV2]](s8) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; CHECK: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF1]](<4 x s32>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC1]](<4 x s8>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s8>), [[UV1:%[0-9]+]]:_(<3 x s8>), [[UV2:%[0-9]+]]:_(<3 x s8>), [[UV3:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>) + ; CHECK: [[UV4:%[0-9]+]]:_(s8), [[UV5:%[0-9]+]]:_(s8), [[UV6:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[UV]](<3 x s8>) + ; CHECK: [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[UV4]](s8) + ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s16) = G_ANYEXT [[UV5]](s8) + ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s16) = G_ANYEXT [[UV6]](s8) ; CHECK: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[ANYEXT]](s16) ; CHECK: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[ANYEXT1]](s16) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -538,12 +541,12 @@ ; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; CHECK: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>) - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT]](<4 x s16>), 32 - ; CHECK: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT2]](s16) + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>) + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; CHECK: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 + ; CHECK: [[EXTRACT1:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT]](<4 x s16>), 32 + ; CHECK: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT1]](s16) ; CHECK: $vgpr0 = COPY [[ANYEXT3]](s32) %0:_(<3 x s8>) = G_IMPLICIT_DEF %1:_(s8) = G_EXTRACT %0, 16 @@ -558,13 +561,16 @@ ; CHECK-LABEL: name: extract_s8_v5s1_offset4 ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s32>) = G_IMPLICIT_DEF ; CHECK: [[TRUNC:%[0-9]+]]:_(<6 x s1>) = G_TRUNC [[DEF]](<6 x s32>) - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<5 x s1>) = G_EXTRACT [[TRUNC]](<6 x s1>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(s1), [[UV1:%[0-9]+]]:_(s1), [[UV2:%[0-9]+]]:_(s1), [[UV3:%[0-9]+]]:_(s1), [[UV4:%[0-9]+]]:_(s1) = G_UNMERGE_VALUES [[EXTRACT]](<5 x s1>) - ; CHECK: [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[UV]](s1) - ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s16) = G_ANYEXT [[UV1]](s1) - ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s16) = G_ANYEXT [[UV2]](s1) - ; CHECK: [[ANYEXT3:%[0-9]+]]:_(s16) = G_ANYEXT [[UV3]](s1) - ; CHECK: [[ANYEXT4:%[0-9]+]]:_(s16) = G_ANYEXT [[UV4]](s1) + ; CHECK: [[DEF1:%[0-9]+]]:_(<6 x s32>) = G_IMPLICIT_DEF + ; CHECK: [[TRUNC1:%[0-9]+]]:_(<6 x s1>) = G_TRUNC [[DEF1]](<6 x s32>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s1>) = G_CONCAT_VECTORS [[TRUNC]](<6 x s1>), [[TRUNC1]](<6 x s1>), [[TRUNC1]](<6 x s1>), [[TRUNC1]](<6 x s1>), [[TRUNC1]](<6 x s1>) + ; CHECK: [[UV:%[0-9]+]]:_(<5 x s1>), [[UV1:%[0-9]+]]:_(<5 x s1>), [[UV2:%[0-9]+]]:_(<5 x s1>), [[UV3:%[0-9]+]]:_(<5 x s1>), [[UV4:%[0-9]+]]:_(<5 x s1>), [[UV5:%[0-9]+]]:_(<5 x s1>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s1>) + ; CHECK: [[UV6:%[0-9]+]]:_(s1), [[UV7:%[0-9]+]]:_(s1), [[UV8:%[0-9]+]]:_(s1), [[UV9:%[0-9]+]]:_(s1), [[UV10:%[0-9]+]]:_(s1) = G_UNMERGE_VALUES [[UV]](<5 x s1>) + ; CHECK: [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[UV6]](s1) + ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s16) = G_ANYEXT [[UV7]](s1) + ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s16) = G_ANYEXT [[UV8]](s1) + ; CHECK: [[ANYEXT3:%[0-9]+]]:_(s16) = G_ANYEXT [[UV9]](s1) + ; CHECK: [[ANYEXT4:%[0-9]+]]:_(s16) = G_ANYEXT [[UV10]](s1) ; CHECK: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[ANYEXT]](s16) ; CHECK: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[ANYEXT1]](s16) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 @@ -581,12 +587,12 @@ ; CHECK: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; CHECK: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT4]], [[SHL2]] ; CHECK: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) - ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>) - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; CHECK: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<5 x s16>), 0 - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT]](<6 x s16>), 64 - ; CHECK: [[ANYEXT5:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT2]](s16) + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>) + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<6 x s16>), 0 + ; CHECK: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<5 x s16>), 0 + ; CHECK: [[EXTRACT1:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT]](<6 x s16>), 64 + ; CHECK: [[ANYEXT5:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT1]](s16) ; CHECK: $vgpr0 = COPY [[ANYEXT5]](s32) %0:_(<5 x s1>) = G_IMPLICIT_DEF %1:_(s1) = G_EXTRACT %0, 4 @@ -850,10 +856,12 @@ ; CHECK-LABEL: name: extract_s16_v3s16_offset0 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT]](<4 x s16>), 0 - ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT1]](s16) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; CHECK: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT]](<4 x s16>), 0 + ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16) ; CHECK: $vgpr0 = COPY [[ANYEXT]](s32) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(s16) = G_EXTRACT %0, 0 @@ -933,10 +941,12 @@ ; CHECK-LABEL: name: extract_v2s16_v3s16_offset0 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 - ; CHECK: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 + ; CHECK: $vgpr0 = COPY [[EXTRACT]](<2 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<2 x s16>) = G_EXTRACT %0, 0 $vgpr0 = COPY %1 @@ -949,10 +959,12 @@ ; CHECK-LABEL: name: extract_v2s16_v5s16_offset0 ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[DEF]](<6 x s16>), 0 - ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<5 x s16>), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<6 x s16>), 0 - ; CHECK: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>) + ; CHECK: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>), [[UV2:%[0-9]+]]:_(<5 x s16>), [[UV3:%[0-9]+]]:_(<5 x s16>), [[UV4:%[0-9]+]]:_(<5 x s16>), [[UV5:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV]](<5 x s16>), 0 + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<6 x s16>), 0 + ; CHECK: $vgpr0 = COPY [[EXTRACT]](<2 x s16>) %0:_(<5 x s16>) = G_IMPLICIT_DEF %1:_(<2 x s16>) = G_EXTRACT %0, 0 $vgpr0 = COPY %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fabs.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fabs.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fabs.mir @@ -209,13 +209,15 @@ ; SI-LABEL: name: test_fabs_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) @@ -231,21 +233,23 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] ; SI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; SI: [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; SI: [[FABS:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BITCAST2]] ; SI: [[FABS1:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BITCAST3]] - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FABS]](<2 x s16>), [[FABS1]](<2 x s16>), [[DEF1]](<2 x s16>) - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FABS]](<2 x s16>), [[FABS1]](<2 x s16>), [[DEF2]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<6 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fabs_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) @@ -261,34 +265,36 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] ; VI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; VI: [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; VI: [[FABS:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BITCAST2]] ; VI: [[FABS1:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BITCAST3]] - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FABS]](<2 x s16>), [[FABS1]](<2 x s16>), [[DEF1]](<2 x s16>) - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FABS]](<2 x s16>), [[FABS1]](<2 x s16>), [[DEF2]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<6 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fabs_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY]](s32), [[COPY1]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) - ; GFX9: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF1]](s32) - ; GFX9: [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF2]](s32) + ; GFX9: [[DEF3:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; GFX9: [[FABS:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BUILD_VECTOR_TRUNC]] ; GFX9: [[FABS1:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BUILD_VECTOR_TRUNC1]] - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FABS]](<2 x s16>), [[FABS1]](<2 x s16>), [[DEF2]](<2 x s16>) - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FABS]](<2 x s16>), [[FABS1]](<2 x s16>), [[DEF3]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<6 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_FABS %0 S_NOP 0, implicit %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fadd.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fadd.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fadd.mir @@ -330,26 +330,29 @@ ; SI-LABEL: name: test_fadd_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -374,31 +377,34 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fadd_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[TRUNC]], [[TRUNC3]] @@ -414,44 +420,47 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fadd_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY]](s32), [[COPY1]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) - ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF2]](s32) - ; GFX9: [[DEF3:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; GFX9: [[DEF3:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF3]](s32) + ; GFX9: [[DEF4:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[DEF2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[DEF3]](s32) ; GFX9: [[FADD:%[0-9]+]]:_(<2 x s16>) = G_FADD [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]] ; GFX9: [[FADD1:%[0-9]+]]:_(<2 x s16>) = G_FADD [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]] - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FADD]](<2 x s16>), [[FADD1]](<2 x s16>), [[DEF3]](<2 x s16>) - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FADD]](<2 x s16>), [[FADD1]](<2 x s16>), [[DEF4]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<6 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF %2:_(<3 x s16>) = G_FADD %0, %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fcanonicalize.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fcanonicalize.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fcanonicalize.mir @@ -222,15 +222,17 @@ ; SI-LABEL: name: test_fcanonicalize_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -252,20 +254,22 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fcanonicalize_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[FCANONICALIZE:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC]] @@ -281,31 +285,33 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fcanonicalize_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY]](s32), [[COPY1]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) - ; GFX9: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF1]](s32) - ; GFX9: [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF2]](s32) + ; GFX9: [[DEF3:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; GFX9: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC]] ; GFX9: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC1]] - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FCANONICALIZE]](<2 x s16>), [[FCANONICALIZE1]](<2 x s16>), [[DEF2]](<2 x s16>) - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FCANONICALIZE]](<2 x s16>), [[FCANONICALIZE1]](<2 x s16>), [[DEF3]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<6 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_FCANONICALIZE %0 S_NOP 0, implicit %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fcos.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fcos.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fcos.mir @@ -326,15 +326,17 @@ ; SI-LABEL: name: test_fcos_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -363,20 +365,22 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fcos_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH3118 @@ -399,20 +403,22 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fcos_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH3118 @@ -426,11 +432,11 @@ ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT2]](s16) - ; GFX9: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF1]](s32) - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF2]](s32) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_FCOS %0 S_NOP 0, implicit %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir @@ -1307,26 +1307,29 @@ ; SI-LABEL: name: test_fdiv_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -1385,31 +1388,34 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fdiv_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -1440,31 +1446,34 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fdiv_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; GFX9: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX9: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; GFX9: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX9: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -1489,33 +1498,36 @@ ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT5]](s16) - ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF2]](s32) - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; GFX9: [[DEF3:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF3]](s32) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-UNSAFE-LABEL: name: test_fdiv_v3s16 ; GFX9-UNSAFE: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-UNSAFE: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX9-UNSAFE: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-UNSAFE: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; GFX9-UNSAFE: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9-UNSAFE: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9-UNSAFE: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9-UNSAFE: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-UNSAFE: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9-UNSAFE: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9-UNSAFE: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-UNSAFE: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9-UNSAFE: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9-UNSAFE: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9-UNSAFE: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9-UNSAFE: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9-UNSAFE: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9-UNSAFE: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX9-UNSAFE: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9-UNSAFE: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9-UNSAFE: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; GFX9-UNSAFE: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9-UNSAFE: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9-UNSAFE: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; GFX9-UNSAFE: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX9-UNSAFE: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9-UNSAFE: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX9-UNSAFE: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX9-UNSAFE: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; GFX9-UNSAFE: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; GFX9-UNSAFE: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; GFX9-UNSAFE: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX9-UNSAFE: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX9-UNSAFE: [[INT:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC3]](s16) @@ -1528,33 +1540,36 @@ ; GFX9-UNSAFE: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL1]](s16) ; GFX9-UNSAFE: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9-UNSAFE: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL2]](s16) - ; GFX9-UNSAFE: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9-UNSAFE: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF2]](s32) - ; GFX9-UNSAFE: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) - ; GFX9-UNSAFE: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9-UNSAFE: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; GFX9-UNSAFE: [[DEF3:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9-UNSAFE: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF3]](s32) + ; GFX9-UNSAFE: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) + ; GFX9-UNSAFE: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; GFX9-UNSAFE: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX10-LABEL: name: test_fdiv_v3s16 ; GFX10: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX10: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX10: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX10: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; GFX10: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; GFX10: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX10: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX10: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX10: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX10: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX10: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX10: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX10: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX10: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX10: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; GFX10: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX10: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX10: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX10: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX10: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX10: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; GFX10: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX10: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; GFX10: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX10: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX10: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; GFX10: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX10: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX10: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; GFX10: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX10: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX10: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; GFX10: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; GFX10: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; GFX10: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX10: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX10: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -1579,11 +1594,11 @@ ; GFX10: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16) ; GFX10: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX10: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT5]](s16) - ; GFX10: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX10: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF2]](s32) - ; GFX10: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) - ; GFX10: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX10: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; GFX10: [[DEF3:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX10: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF3]](s32) + ; GFX10: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) + ; GFX10: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; GFX10: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF %2:_(<3 x s16>) = G_FDIV %0, %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ffloor.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ffloor.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ffloor.mir @@ -322,15 +322,17 @@ ; SI-LABEL: name: test_ffloor_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -352,20 +354,22 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_ffloor_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[FFLOOR:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC]] @@ -381,20 +385,22 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_ffloor_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[FFLOOR:%[0-9]+]]:_(s16) = G_FFLOOR [[TRUNC]] @@ -404,11 +410,11 @@ ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FFLOOR1]](s16) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FFLOOR2]](s16) - ; GFX9: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF1]](s32) - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF2]](s32) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_FFLOOR %0 S_NOP 0, implicit %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fma.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fma.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fma.mir @@ -393,37 +393,41 @@ ; SI-LABEL: name: test_fma_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF2]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; SI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT2]](<3 x s16>), 0 - ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) - ; SI: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 + ; SI: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; SI: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) ; SI: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; SI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; SI: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) - ; SI: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; SI: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) ; SI: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; SI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -451,42 +455,46 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) - ; SI: [[EXTRACT3:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT3]](<3 x s16>) + ; SI: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS3]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fma_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF2]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; VI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT2]](<3 x s16>), 0 - ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) - ; VI: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 + ; VI: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; VI: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) ; VI: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; VI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; VI: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) - ; VI: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; VI: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) ; VI: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; VI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; VI: [[FMA:%[0-9]+]]:_(s16) = G_FMA [[TRUNC]], [[TRUNC3]], [[TRUNC6]] @@ -502,57 +510,61 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) - ; VI: [[EXTRACT3:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT3]](<3 x s16>) + ; VI: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS3]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fma_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF2]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY]](s32), [[COPY1]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) - ; GFX9: [[DEF3:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF3]](s32) - ; GFX9: [[DEF4:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; GFX9: [[DEF4:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF4]](s32) + ; GFX9: [[DEF5:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[DEF3]](s32) - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT2]](<3 x s16>), 0 - ; GFX9: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) - ; GFX9: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[DEF4]](s32) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 + ; GFX9: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; GFX9: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) ; GFX9: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) - ; GFX9: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; GFX9: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) ; GFX9: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) ; GFX9: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[COPY7]](s32) ; GFX9: [[COPY8:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY8]](s32), [[DEF3]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY8]](s32), [[DEF4]](s32) ; GFX9: [[FMA:%[0-9]+]]:_(<2 x s16>) = G_FMA [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]], [[BUILD_VECTOR_TRUNC4]] ; GFX9: [[FMA1:%[0-9]+]]:_(<2 x s16>) = G_FMA [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]], [[BUILD_VECTOR_TRUNC5]] - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FMA]](<2 x s16>), [[FMA1]](<2 x s16>), [[DEF4]](<2 x s16>) - ; GFX9: [[EXTRACT3:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT3]](<3 x s16>) + ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FMA]](<2 x s16>), [[FMA1]](<2 x s16>), [[DEF5]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS3]](<6 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF %2:_(<3 x s16>) = G_IMPLICIT_DEF Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmaxnum.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmaxnum.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmaxnum.mir @@ -506,8 +506,9 @@ ; GFX9: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV3]] ; GFX9: [[FMAXNUM_IEEE1:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]] ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[FMAXNUM_IEEE]](<2 x s16>), [[FMAXNUM_IEEE1]](<2 x s16>) - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fminnum.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fminnum.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fminnum.mir @@ -506,8 +506,9 @@ ; GFX9: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV3]] ; GFX9: [[FMINNUM_IEEE1:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]] ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[FMINNUM_IEEE]](<2 x s16>), [[FMINNUM_IEEE1]](<2 x s16>) - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmul.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmul.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmul.mir @@ -329,26 +329,29 @@ ; SI-LABEL: name: test_fmul_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -373,31 +376,34 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fmul_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC3]] @@ -413,44 +419,47 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fmul_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY]](s32), [[COPY1]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) - ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF2]](s32) - ; GFX9: [[DEF3:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; GFX9: [[DEF3:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF3]](s32) + ; GFX9: [[DEF4:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[DEF2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[DEF3]](s32) ; GFX9: [[FMUL:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]] ; GFX9: [[FMUL1:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]] - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FMUL]](<2 x s16>), [[FMUL1]](<2 x s16>), [[DEF3]](<2 x s16>) - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FMUL]](<2 x s16>), [[FMUL1]](<2 x s16>), [[DEF4]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<6 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF %2:_(<3 x s16>) = G_FMUL %0, %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fneg.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fneg.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fneg.mir @@ -207,13 +207,15 @@ ; SI-LABEL: name: test_fneg_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) @@ -229,21 +231,23 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] ; SI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; SI: [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; SI: [[FNEG:%[0-9]+]]:_(<2 x s16>) = G_FNEG [[BITCAST2]] ; SI: [[FNEG1:%[0-9]+]]:_(<2 x s16>) = G_FNEG [[BITCAST3]] - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FNEG]](<2 x s16>), [[FNEG1]](<2 x s16>), [[DEF1]](<2 x s16>) - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FNEG]](<2 x s16>), [[FNEG1]](<2 x s16>), [[DEF2]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<6 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fneg_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 ; VI: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) @@ -259,34 +263,36 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] ; VI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; VI: [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; VI: [[FNEG:%[0-9]+]]:_(<2 x s16>) = G_FNEG [[BITCAST2]] ; VI: [[FNEG1:%[0-9]+]]:_(<2 x s16>) = G_FNEG [[BITCAST3]] - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FNEG]](<2 x s16>), [[FNEG1]](<2 x s16>), [[DEF1]](<2 x s16>) - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FNEG]](<2 x s16>), [[FNEG1]](<2 x s16>), [[DEF2]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<6 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fneg_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY]](s32), [[COPY1]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) - ; GFX9: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF1]](s32) - ; GFX9: [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF2]](s32) + ; GFX9: [[DEF3:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF ; GFX9: [[FNEG:%[0-9]+]]:_(<2 x s16>) = G_FNEG [[BUILD_VECTOR_TRUNC]] ; GFX9: [[FNEG1:%[0-9]+]]:_(<2 x s16>) = G_FNEG [[BUILD_VECTOR_TRUNC1]] - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FNEG]](<2 x s16>), [[FNEG1]](<2 x s16>), [[DEF2]](<2 x s16>) - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FNEG]](<2 x s16>), [[FNEG1]](<2 x s16>), [[DEF3]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<6 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_FNEG %0 S_NOP 0, implicit %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fpext.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fpext.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fpext.mir @@ -70,15 +70,17 @@ ; CHECK-LABEL: name: test_fpext_v3f16_to_v3f32 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; CHECK: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; CHECK: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; CHECK: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; CHECK: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-freeze.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-freeze.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-freeze.mir @@ -490,8 +490,9 @@ ; CHECK: [[ANYEXT:%[0-9]+]]:_(<4 x s32>) = G_ANYEXT [[INSERT]](<4 x s1>) ; CHECK: [[FREEZE:%[0-9]+]]:_(<4 x s32>) = G_FREEZE [[ANYEXT]] ; CHECK: [[TRUNC:%[0-9]+]]:_(<4 x s1>) = G_TRUNC [[FREEZE]](<4 x s32>) - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s1>) = G_EXTRACT [[TRUNC]](<4 x s1>), 0 - ; CHECK: [[ANYEXT1:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[EXTRACT]](<3 x s1>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s1>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s1>), [[DEF1]](<4 x s1>), [[DEF1]](<4 x s1>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s1>), [[UV1:%[0-9]+]]:_(<3 x s1>), [[UV2:%[0-9]+]]:_(<3 x s1>), [[UV3:%[0-9]+]]:_(<3 x s1>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s1>) + ; CHECK: [[ANYEXT1:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV]](<3 x s1>) ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT1]](<3 x s32>) %0:_(<3 x s1>) = G_IMPLICIT_DEF %1:_(<3 x s1>) = G_FREEZE %0 @@ -530,8 +531,9 @@ ; CHECK: [[ANYEXT:%[0-9]+]]:_(<4 x s32>) = G_ANYEXT [[INSERT]](<4 x s8>) ; CHECK: [[FREEZE:%[0-9]+]]:_(<4 x s32>) = G_FREEZE [[ANYEXT]] ; CHECK: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[FREEZE]](<4 x s32>) - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s8>) = G_EXTRACT [[TRUNC1]](<4 x s8>), 0 - ; CHECK: [[ANYEXT1:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[EXTRACT]](<3 x s8>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[TRUNC1]](<4 x s8>), [[DEF]](<4 x s8>), [[DEF]](<4 x s8>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s8>), [[UV1:%[0-9]+]]:_(<3 x s8>), [[UV2:%[0-9]+]]:_(<3 x s8>), [[UV3:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>) + ; CHECK: [[ANYEXT1:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV]](<3 x s8>) ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT1]](<3 x s32>) %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2 %1:_(<3 x s8>) = G_TRUNC %0 @@ -573,13 +575,14 @@ ; CHECK: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 ; CHECK: [[FREEZE:%[0-9]+]]:_(<4 x s16>) = G_FREEZE [[INSERT]] - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[FREEZE]](<4 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 - ; CHECK: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[FREEZE]](<4 x s16>), [[DEF2]](<4 x s16>), [[DEF2]](<4 x s16>) + ; CHECK: [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[UV3]](<3 x s16>), 0 + ; CHECK: [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) @@ -628,15 +631,16 @@ ; CHECK: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<5 x s16>), 0 ; CHECK: [[FREEZE:%[0-9]+]]:_(<6 x s16>) = G_FREEZE [[INSERT]] - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[FREEZE]](<6 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<5 x s16>), 0 - ; CHECK: [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>) - ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[FREEZE]](<6 x s16>), [[DEF2]](<6 x s16>), [[DEF2]](<6 x s16>), [[DEF2]](<6 x s16>), [[DEF2]](<6 x s16>) + ; CHECK: [[UV5:%[0-9]+]]:_(<5 x s16>), [[UV6:%[0-9]+]]:_(<5 x s16>), [[UV7:%[0-9]+]]:_(<5 x s16>), [[UV8:%[0-9]+]]:_(<5 x s16>), [[UV9:%[0-9]+]]:_(<5 x s16>), [[UV10:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<30 x s16>) + ; CHECK: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[UV5]](<5 x s16>), 0 + ; CHECK: [[UV11:%[0-9]+]]:_(<2 x s16>), [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) + ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsin.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsin.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsin.mir @@ -326,15 +326,17 @@ ; SI-LABEL: name: test_fsin_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -363,20 +365,22 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fsin_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[C1:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH3118 @@ -399,20 +403,22 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fsin_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[C1:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH3118 @@ -426,11 +432,11 @@ ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT2]](s16) - ; GFX9: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF1]](s32) - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF2]](s32) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_FSIN %0 S_NOP 0, implicit %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsqrt.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsqrt.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsqrt.mir @@ -246,15 +246,17 @@ ; SI-LABEL: name: test_fsqrt_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -276,20 +278,22 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fsqrt_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; VI: [[FSQRT:%[0-9]+]]:_(s16) = G_FSQRT [[TRUNC]] @@ -305,20 +309,22 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fsqrt_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[FSQRT:%[0-9]+]]:_(s16) = G_FSQRT [[TRUNC]] @@ -328,11 +334,11 @@ ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FSQRT1]](s16) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FSQRT2]](s16) - ; GFX9: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF1]](s32) - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT1]](<3 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF2]](s32) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS1]](<4 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_FSQRT %0 S_NOP 0, implicit %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsub.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsub.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsub.mir @@ -374,26 +374,29 @@ ; SI-LABEL: name: test_fsub_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[FNEG:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC3]] @@ -421,31 +424,34 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_fsub_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[FNEG:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC3]] @@ -464,31 +470,34 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_fsub_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; GFX9: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; GFX9: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; GFX9: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; GFX9: [[FNEG:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC3]] @@ -501,11 +510,11 @@ ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD1]](s16) ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) ; GFX9: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD2]](s16) - ; GFX9: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF2]](s32) - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; GFX9: [[DEF3:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF3]](s32) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF %2:_(<3 x s16>) = G_FSUB %0, %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-implicit-def.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-implicit-def.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-implicit-def.mir @@ -377,11 +377,14 @@ ; CHECK-LABEL: name: test_implicit_def_v3s1 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF ; CHECK: [[TRUNC:%[0-9]+]]:_(<4 x s1>) = G_TRUNC [[DEF]](<4 x s32>) - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s1>) = G_EXTRACT [[TRUNC]](<4 x s1>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(s1), [[UV1:%[0-9]+]]:_(s1), [[UV2:%[0-9]+]]:_(s1) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s1>) - ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s1) - ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[UV1]](s1) - ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[UV2]](s1) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; CHECK: [[TRUNC1:%[0-9]+]]:_(<4 x s1>) = G_TRUNC [[DEF1]](<4 x s32>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s1>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s1>), [[TRUNC1]](<4 x s1>), [[TRUNC1]](<4 x s1>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s1>), [[UV1:%[0-9]+]]:_(<3 x s1>), [[UV2:%[0-9]+]]:_(<3 x s1>), [[UV3:%[0-9]+]]:_(<3 x s1>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s1>) + ; CHECK: [[UV4:%[0-9]+]]:_(s1), [[UV5:%[0-9]+]]:_(s1), [[UV6:%[0-9]+]]:_(s1) = G_UNMERGE_VALUES [[UV]](<3 x s1>) + ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV4]](s1) + ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[UV5]](s1) + ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[UV6]](s1) ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32) ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) %0:_(<3 x s1>) = G_IMPLICIT_DEF @@ -411,11 +414,14 @@ ; CHECK-LABEL: name: test_implicit_def_v3s8 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF ; CHECK: [[TRUNC:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF]](<4 x s32>) - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s8>) = G_EXTRACT [[TRUNC]](<4 x s8>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(s8), [[UV1:%[0-9]+]]:_(s8), [[UV2:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s8>) - ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s8) - ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[UV1]](s8) - ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[UV2]](s8) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; CHECK: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[DEF1]](<4 x s32>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[TRUNC]](<4 x s8>), [[TRUNC1]](<4 x s8>), [[TRUNC1]](<4 x s8>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s8>), [[UV1:%[0-9]+]]:_(<3 x s8>), [[UV2:%[0-9]+]]:_(<3 x s8>), [[UV3:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>) + ; CHECK: [[UV4:%[0-9]+]]:_(s8), [[UV5:%[0-9]+]]:_(s8), [[UV6:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[UV]](<3 x s8>) + ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV4]](s8) + ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[UV5]](s8) + ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[UV6]](s8) ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32) ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) %0:_(<3 x s8>) = G_IMPLICIT_DEF @@ -442,9 +448,10 @@ ; CHECK-LABEL: name: test_implicit_def_v3s16 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %1(<4 x s16>), %1(<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<4 x s16>) = G_IMPLICIT_DEF @@ -471,9 +478,11 @@ ; CHECK-LABEL: name: test_implicit_def_v5s16 ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[DEF]](<6 x s16>), 0 - ; CHECK: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<5 x s16>), 0 + ; CHECK: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>), [[UV2:%[0-9]+]]:_(<5 x s16>), [[UV3:%[0-9]+]]:_(<5 x s16>), [[UV4:%[0-9]+]]:_(<5 x s16>), [[UV5:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s16>) + ; CHECK: [[DEF2:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[INSERT:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF2]], [[UV]](<5 x s16>), 0 ; CHECK: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<8 x s16>) %0:_(<5 x s16>) = G_IMPLICIT_DEF %1:_(<8 x s16>) = G_IMPLICIT_DEF Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-insert.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-insert.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-insert.mir @@ -843,8 +843,9 @@ ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[TRUNC]](s16), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_EXTRACT %0, 0 @@ -869,8 +870,9 @@ ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[TRUNC]](s16), 16 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_EXTRACT %0, 0 @@ -895,8 +897,9 @@ ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[TRUNC]](s16), 32 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_EXTRACT %0, 0 @@ -920,8 +923,9 @@ ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[COPY1]](<2 x s16>), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_EXTRACT %0, 0 @@ -944,8 +948,9 @@ ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[COPY1]](<2 x s16>), 16 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_EXTRACT %0, 0 @@ -968,8 +973,9 @@ ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[COPY1]](s32), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_EXTRACT %0, 0 @@ -992,8 +998,9 @@ ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[COPY1]](s32), 16 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_EXTRACT %0, 0 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.load.2d.d16.ll =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.load.2d.d16.ll +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.load.2d.d16.ll @@ -768,13 +768,14 @@ ; UNPACKED: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; UNPACKED: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; UNPACKED: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; UNPACKED: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; UNPACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %16(<4 x s16>), %16(<4 x s16>) + ; UNPACKED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; UNPACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; UNPACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; UNPACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 - ; UNPACKED: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32 - ; UNPACKED: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>) - ; UNPACKED: $vgpr1 = COPY [[EXTRACT2]](<2 x s16>) + ; UNPACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; UNPACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 + ; UNPACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32 + ; UNPACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>) + ; UNPACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>) ; UNPACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1 ; PACKED-LABEL: name: image_load_v3f16_dmask_0000 ; PACKED: bb.1 (%ir-block.0): @@ -790,13 +791,14 @@ ; PACKED: [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0 ; PACKED: [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1 ; PACKED: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; PACKED: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; PACKED: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %16(<4 x s16>), %16(<4 x s16>) + ; PACKED: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; PACKED: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; PACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; PACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 - ; PACKED: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32 - ; PACKED: $vgpr0 = COPY [[EXTRACT1]](<2 x s16>) - ; PACKED: $vgpr1 = COPY [[EXTRACT2]](<2 x s16>) + ; PACKED: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; PACKED: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 + ; PACKED: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 32 + ; PACKED: $vgpr0 = COPY [[EXTRACT]](<2 x s16>) + ; PACKED: $vgpr1 = COPY [[EXTRACT1]](<2 x s16>) ; PACKED: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1 %tex = call <3 x half> @llvm.amdgcn.image.load.2d.v3f16.i32(i32 0, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0) ret <3 x half> %tex Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.s.buffer.load.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.s.buffer.load.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.s.buffer.load.mir @@ -30,8 +30,10 @@ ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 ; GCN: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GCN: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load 12, align 4) - ; GCN: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), 0 - ; GCN: S_ENDPGM 0, implicit [[EXTRACT]](<3 x s32>) + ; GCN: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; GCN: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>) + ; GCN: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) + ; GCN: S_ENDPGM 0, implicit [[UV]](<3 x s32>) %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 %1:_(s32) = G_CONSTANT i32 0 %2:_(<3 x s32>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0 @@ -49,8 +51,10 @@ ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 ; GCN: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GCN: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x p3>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load 12, align 4) - ; GCN: [[EXTRACT:%[0-9]+]]:_(<3 x p3>) = G_EXTRACT [[AMDGPU_S_BUFFER_LOAD]](<4 x p3>), 0 - ; GCN: S_ENDPGM 0, implicit [[EXTRACT]](<3 x p3>) + ; GCN: [[DEF:%[0-9]+]]:_(<4 x p3>) = G_IMPLICIT_DEF + ; GCN: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x p3>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x p3>), [[DEF]](<4 x p3>), [[DEF]](<4 x p3>) + ; GCN: [[UV:%[0-9]+]]:_(<3 x p3>), [[UV1:%[0-9]+]]:_(<3 x p3>), [[UV2:%[0-9]+]]:_(<3 x p3>), [[UV3:%[0-9]+]]:_(<3 x p3>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x p3>) + ; GCN: S_ENDPGM 0, implicit [[UV]](<3 x p3>) %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 %1:_(s32) = G_CONSTANT i32 0 %2:_(<3 x p3>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0 @@ -68,8 +72,10 @@ ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 ; GCN: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GCN: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load 12, align 4) - ; GCN: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), 0 - ; GCN: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[EXTRACT]](<3 x s32>) + ; GCN: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; GCN: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>) + ; GCN: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) + ; GCN: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[UV]](<3 x s32>) ; GCN: S_ENDPGM 0, implicit [[BITCAST]](<6 x s16>) %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 %1:_(s32) = G_CONSTANT i32 0 @@ -88,8 +94,10 @@ ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 ; GCN: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GCN: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<8 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load 24, align 4) - ; GCN: [[EXTRACT:%[0-9]+]]:_(<6 x s32>) = G_EXTRACT [[AMDGPU_S_BUFFER_LOAD]](<8 x s32>), 0 - ; GCN: S_ENDPGM 0, implicit [[EXTRACT]](<6 x s32>) + ; GCN: [[DEF:%[0-9]+]]:_(<8 x s32>) = G_IMPLICIT_DEF + ; GCN: [[CONCAT_VECTORS:%[0-9]+]]:_(<24 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<8 x s32>), [[DEF]](<8 x s32>), [[DEF]](<8 x s32>) + ; GCN: [[UV:%[0-9]+]]:_(<6 x s32>), [[UV1:%[0-9]+]]:_(<6 x s32>), [[UV2:%[0-9]+]]:_(<6 x s32>), [[UV3:%[0-9]+]]:_(<6 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<24 x s32>) + ; GCN: S_ENDPGM 0, implicit [[UV]](<6 x s32>) %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 %1:_(s32) = G_CONSTANT i32 0 %2:_(<6 x s32>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0 @@ -107,8 +115,10 @@ ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 ; GCN: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GCN: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s64>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load 24, align 4) - ; GCN: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[AMDGPU_S_BUFFER_LOAD]](<4 x s64>), 0 - ; GCN: S_ENDPGM 0, implicit [[EXTRACT]](<3 x s64>) + ; GCN: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF + ; GCN: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s64>), [[DEF]](<4 x s64>), [[DEF]](<4 x s64>) + ; GCN: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) + ; GCN: S_ENDPGM 0, implicit [[UV]](<3 x s64>) %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 %1:_(s32) = G_CONSTANT i32 0 %2:_(<3 x s64>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0 @@ -126,29 +136,31 @@ ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 ; GCN: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GCN: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load 12, align 4) - ; GCN: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), 0 - ; GCN: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s32>) + ; GCN: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; GCN: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>) + ; GCN: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) + ; GCN: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](<3 x s32>) ; GCN: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 - ; GCN: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32) + ; GCN: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C1]](s32) ; GCN: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; GCN: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32) + ; GCN: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C2]](s32) ; GCN: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 - ; GCN: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C3]](s32) - ; GCN: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32) + ; GCN: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C3]](s32) + ; GCN: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV4]](s32) ; GCN: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) ; GCN: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) ; GCN: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) - ; GCN: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32) - ; GCN: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32) - ; GCN: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C3]](s32) - ; GCN: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV1]](s32) + ; GCN: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C1]](s32) + ; GCN: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C2]](s32) + ; GCN: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C3]](s32) + ; GCN: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV5]](s32) ; GCN: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) ; GCN: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) ; GCN: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) - ; GCN: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32) - ; GCN: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32) - ; GCN: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C3]](s32) - ; GCN: [[COPY9:%[0-9]+]]:_(s32) = COPY [[UV2]](s32) + ; GCN: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C1]](s32) + ; GCN: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C2]](s32) + ; GCN: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C3]](s32) + ; GCN: [[COPY9:%[0-9]+]]:_(s32) = COPY [[UV6]](s32) ; GCN: [[COPY10:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) ; GCN: [[COPY11:%[0-9]+]]:_(s32) = COPY [[LSHR7]](s32) ; GCN: [[COPY12:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32) @@ -195,8 +207,8 @@ ; GCN: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C2]](s32) ; GCN: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND10]], [[SHL5]] ; GCN: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32) - ; GCN: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; GCN: S_ENDPGM 0, implicit [[CONCAT_VECTORS]](<12 x s16>) + ; GCN: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; GCN: S_ENDPGM 0, implicit [[CONCAT_VECTORS1]](<12 x s16>) %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 %1:_(s32) = G_CONSTANT i32 0 %2:_(<12 x s8>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0 @@ -215,8 +227,10 @@ ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 ; GCN: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; GCN: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load 12, align 4) - ; GCN: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), 0 - ; GCN: S_ENDPGM 0, implicit [[EXTRACT]](<3 x s32>) + ; GCN: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; GCN: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>) + ; GCN: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) + ; GCN: S_ENDPGM 0, implicit [[UV]](<3 x s32>) %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3 %1:_(s32) = G_CONSTANT i32 0 %2:_(<3 x s32>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir @@ -5884,37 +5884,42 @@ ; CI-LABEL: name: test_load_constant_v3s16_align8 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load 6, align 8, addrspace 4) - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; VI-LABEL: name: test_load_constant_v3s16_align8 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load 6, align 8, addrspace 4) - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; GFX9-LABEL: name: test_load_constant_v3s16_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load 6, align 8, addrspace 4) - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; CI-MESA-LABEL: name: test_load_constant_v3s16_align8 ; CI-MESA: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load 6, align 8, addrspace 4) - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CI-MESA: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; GFX9-MESA-LABEL: name: test_load_constant_v3s16_align8 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load 6, align 8, addrspace 4) - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; GFX9-MESA: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) %0:_(p4) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 8, addrspace 4) @@ -5992,15 +5997,18 @@ ; CI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load 2 + 4, addrspace 4) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; CI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_constant_v3s16_align2 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 @@ -6022,15 +6030,18 @@ ; VI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load 2 + 4, addrspace 4) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-LABEL: name: test_load_constant_v3s16_align2 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 @@ -6046,15 +6057,18 @@ ; GFX9: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load 2 + 4, addrspace 4) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-MESA-LABEL: name: test_load_constant_v3s16_align2 ; CI-MESA: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 @@ -6076,15 +6090,18 @@ ; CI-MESA: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load 2 + 4, addrspace 4) ; CI-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; CI-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-MESA-LABEL: name: test_load_constant_v3s16_align2 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 @@ -6100,15 +6117,18 @@ ; GFX9-MESA: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p4) :: (load 2 + 4, addrspace 4) ; GFX9-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; GFX9-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p4) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 2, addrspace 4) @@ -6172,15 +6192,18 @@ ; CI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; CI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; CI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_constant_v3s16_align1 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 @@ -6225,15 +6248,18 @@ ; VI: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; VI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL3]] ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-LABEL: name: test_load_constant_v3s16_align1 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 @@ -6275,15 +6301,18 @@ ; GFX9: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; GFX9: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL2]] ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR2]](s16), 32 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-MESA-LABEL: name: test_load_constant_v3s16_align1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 @@ -6334,15 +6363,18 @@ ; CI-MESA: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; CI-MESA: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; CI-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-MESA-LABEL: name: test_load_constant_v3s16_align1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 @@ -6384,15 +6416,18 @@ ; GFX9-MESA: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; GFX9-MESA: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL2]] ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR2]](s16), 32 - ; GFX9-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p4) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 1, addrspace 4) @@ -8262,37 +8297,42 @@ ; CI-LABEL: name: test_load_constant_v3s64_align32 ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load 24, align 32, addrspace 4) - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; CI: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; CI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) ; VI-LABEL: name: test_load_constant_v3s64_align32 ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; VI: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load 24, align 32, addrspace 4) - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) ; GFX9-LABEL: name: test_load_constant_v3s64_align32 ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load 24, align 32, addrspace 4) - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; GFX9: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) ; CI-MESA-LABEL: name: test_load_constant_v3s64_align32 ; CI-MESA: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load 24, align 32, addrspace 4) - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) ; GFX9-MESA-LABEL: name: test_load_constant_v3s64_align32 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load 24, align 32, addrspace 4) - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) %0:_(p4) = COPY $vgpr0_vgpr1 %1:_(<3 x s64>) = G_LOAD %0 :: (load 24, align 32, addrspace 4) Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir @@ -5924,37 +5924,42 @@ ; CI-LABEL: name: test_load_flat_v3s16_align8 ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load 6, align 8) - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; VI-LABEL: name: test_load_flat_v3s16_align8 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; VI: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load 6, align 8) - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; GFX9-LABEL: name: test_load_flat_v3s16_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load 6, align 8) - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; CI-MESA-LABEL: name: test_load_flat_v3s16_align8 ; CI-MESA: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load 6, align 8) - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CI-MESA: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; GFX9-MESA-LABEL: name: test_load_flat_v3s16_align8 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load 6, align 8) - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; GFX9-MESA: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) %0:_(p0) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 8, addrspace 0) @@ -6032,15 +6037,18 @@ ; CI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load 2 + 4) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; CI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_flat_v3s16_align2 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 @@ -6062,15 +6070,18 @@ ; VI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load 2 + 4) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-LABEL: name: test_load_flat_v3s16_align2 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 @@ -6086,15 +6097,18 @@ ; GFX9: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load 2 + 4) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-MESA-LABEL: name: test_load_flat_v3s16_align2 ; CI-MESA: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 @@ -6116,15 +6130,18 @@ ; CI-MESA: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load 2 + 4) ; CI-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; CI-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-MESA-LABEL: name: test_load_flat_v3s16_align2 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 @@ -6140,15 +6157,18 @@ ; GFX9-MESA: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p0) :: (load 2 + 4) ; GFX9-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; GFX9-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p0) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 2, addrspace 0) @@ -6212,15 +6232,18 @@ ; CI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; CI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; CI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_flat_v3s16_align1 ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 @@ -6265,15 +6288,18 @@ ; VI: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; VI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL3]] ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-LABEL: name: test_load_flat_v3s16_align1 ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 @@ -6315,15 +6341,18 @@ ; GFX9: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; GFX9: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL2]] ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR2]](s16), 32 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-MESA-LABEL: name: test_load_flat_v3s16_align1 ; CI-MESA: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 @@ -6374,15 +6403,18 @@ ; CI-MESA: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; CI-MESA: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; CI-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-MESA-LABEL: name: test_load_flat_v3s16_align1 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1 @@ -6424,15 +6456,18 @@ ; GFX9-MESA: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; GFX9-MESA: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL2]] ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR2]](s16), 32 - ; GFX9-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p0) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 1, addrspace 0) Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir @@ -1203,8 +1203,10 @@ ; SI-LABEL: name: test_load_global_s96_align16 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load 12, align 16, addrspace 1) - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0 - ; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[EXTRACT]](<3 x s32>) + ; SI: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) + ; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[UV]](<3 x s32>) ; 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 @@ -5929,44 +5931,50 @@ ; SI-LABEL: name: test_load_global_v3s16_align8 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 6, align 8, addrspace 1) - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; SI: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; CI-HSA-LABEL: name: test_load_global_v3s16_align8 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 6, align 8, addrspace 1) - ; CI-HSA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; CI-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-HSA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-HSA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-HSA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CI-HSA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CI-HSA: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; CI-MESA-LABEL: name: test_load_global_v3s16_align8 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 6, align 8, addrspace 1) - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CI-MESA: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; VI-LABEL: name: test_load_global_v3s16_align8 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 6, align 8, addrspace 1) - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v3s16_align8 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 6, align 8, addrspace 1) - ; GFX9-HSA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; GFX9-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9-HSA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9-HSA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-HSA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9-HSA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; GFX9-HSA: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; GFX9-MESA-LABEL: name: test_load_global_v3s16_align8 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load 6, align 8, addrspace 1) - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; GFX9-MESA: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 8, addrspace 1) @@ -6056,15 +6064,18 @@ ; SI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load 2 + 4, addrspace 1) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; SI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-HSA-LABEL: name: test_load_global_v3s16_align2 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 @@ -6092,15 +6103,18 @@ ; CI-MESA: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load 2 + 4, addrspace 1) ; CI-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; CI-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_global_v3s16_align2 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 @@ -6122,15 +6136,18 @@ ; VI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load 2 + 4, addrspace 1) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v3s16_align2 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 @@ -6152,15 +6169,18 @@ ; GFX9-MESA: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load 2 + 4, addrspace 1) ; GFX9-MESA: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; GFX9-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 2, addrspace 1) @@ -6224,15 +6244,18 @@ ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; SI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; SI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-HSA-LABEL: name: test_load_global_v3s16_align1 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 @@ -6289,15 +6312,18 @@ ; CI-MESA: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; CI-MESA: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; CI-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_global_v3s16_align1 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 @@ -6342,15 +6368,18 @@ ; VI: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; VI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL3]] ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-HSA-LABEL: name: test_load_global_v3s16_align1 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 @@ -6398,15 +6427,18 @@ ; GFX9-MESA: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; GFX9-MESA: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL2]] ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9-MESA: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9-MESA: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9-MESA: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9-MESA: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR2]](s16), 32 - ; GFX9-MESA: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9-MESA: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9-MESA: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9-MESA: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 1, addrspace 1) @@ -7036,8 +7068,10 @@ ; SI-LABEL: name: test_load_global_v3s32_align16 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load 12, align 16, addrspace 1) - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0 - ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[EXTRACT]](<3 x s32>) + ; SI: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) + ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[UV]](<3 x s32>) ; CI-HSA-LABEL: name: test_load_global_v3s32_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load 12, align 16, addrspace 1) @@ -8167,44 +8201,50 @@ ; SI-LABEL: name: test_load_global_v3s64_align32 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load 24, align 32, addrspace 1) - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; SI: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; SI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) ; CI-HSA-LABEL: name: test_load_global_v3s64_align32 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load 24, align 32, addrspace 1) - ; CI-HSA: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; CI-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; CI-HSA: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; CI-HSA: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; CI-HSA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; CI-HSA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; CI-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) ; CI-MESA-LABEL: name: test_load_global_v3s64_align32 ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-MESA: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load 24, align 32, addrspace 1) - ; CI-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; CI-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; CI-MESA: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; CI-MESA: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; CI-MESA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; CI-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) ; VI-LABEL: name: test_load_global_v3s64_align32 ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; VI: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load 24, align 32, addrspace 1) - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; VI: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) ; GFX9-HSA-LABEL: name: test_load_global_v3s64_align32 ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-HSA: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load 24, align 32, addrspace 1) - ; GFX9-HSA: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; GFX9-HSA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; GFX9-HSA: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; GFX9-HSA: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; GFX9-HSA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; GFX9-HSA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; GFX9-HSA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) ; GFX9-MESA-LABEL: name: test_load_global_v3s64_align32 ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; GFX9-MESA: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load 24, align 32, addrspace 1) - ; GFX9-MESA: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0 + ; GFX9-MESA: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[LOAD]](<4 x s64>), %2(<4 x s64>), %2(<4 x s64>) + ; GFX9-MESA: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>) ; GFX9-MESA: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF - ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0 + ; GFX9-MESA: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[UV]](<3 x s64>), 0 ; GFX9-MESA: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>) %0:_(p1) = COPY $vgpr0_vgpr1 %1:_(<3 x s64>) = G_LOAD %0 :: (load 24, align 32, addrspace 1) @@ -12643,23 +12683,25 @@ ; SI-LABEL: name: test_extload_global_v2s96_from_24_align16 ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; SI: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load 12, align 16, addrspace 1) - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0 - ; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[EXTRACT]](<3 x s32>) + ; SI: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) + ; SI: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[UV]](<3 x s32>) ; SI: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12 ; SI: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64) ; SI: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load 8 + 12, align 4, addrspace 1) ; SI: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 8 ; SI: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C1]](s64) ; SI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load 4 + 20, addrspace 1) - ; SI: [[DEF:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<3 x s32>) = G_INSERT [[DEF]], [[LOAD1]](<2 x s32>), 0 + ; SI: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF + ; SI: [[INSERT:%[0-9]+]]:_(<3 x s32>) = G_INSERT [[DEF1]], [[LOAD1]](<2 x s32>), 0 ; SI: [[INSERT1:%[0-9]+]]:_(<3 x s32>) = G_INSERT [[INSERT]], [[LOAD2]](s32), 64 ; SI: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[INSERT1]](<3 x s32>) ; SI: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s96>) = G_BUILD_VECTOR [[BITCAST]](s96), [[BITCAST1]](s96) - ; SI: [[EXTRACT1:%[0-9]+]]:_(s96) = G_EXTRACT [[BUILD_VECTOR]](<2 x s96>), 0 - ; SI: [[EXTRACT2:%[0-9]+]]:_(s96) = G_EXTRACT [[BUILD_VECTOR]](<2 x s96>), 96 - ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[EXTRACT1]](s96) - ; SI: $vgpr3_vgpr4_vgpr5 = COPY [[EXTRACT2]](s96) + ; SI: [[EXTRACT:%[0-9]+]]:_(s96) = G_EXTRACT [[BUILD_VECTOR]](<2 x s96>), 0 + ; SI: [[EXTRACT1:%[0-9]+]]:_(s96) = G_EXTRACT [[BUILD_VECTOR]](<2 x s96>), 96 + ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[EXTRACT]](s96) + ; SI: $vgpr3_vgpr4_vgpr5 = COPY [[EXTRACT1]](s96) ; CI-HSA-LABEL: name: test_extload_global_v2s96_from_24_align16 ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1 ; CI-HSA: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load 12, align 16, addrspace 1) Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir @@ -6061,37 +6061,42 @@ ; SI-LABEL: name: test_load_local_v3s16_align8 ; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; SI: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load 6, align 8, addrspace 3) - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; SI: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; CI-LABEL: name: test_load_local_v3s16_align8 ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; CI: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load 6, align 8, addrspace 3) - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; CI-DS128-LABEL: name: test_load_local_v3s16_align8 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; CI-DS128: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load 6, align 8, addrspace 3) - ; CI-DS128: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; CI-DS128: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-DS128: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-DS128: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-DS128: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CI-DS128: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CI-DS128: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; VI-LABEL: name: test_load_local_v3s16_align8 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; VI: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load 6, align 8, addrspace 3) - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) ; GFX9-LABEL: name: test_load_local_v3s16_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 ; GFX9: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load 6, align 8, addrspace 3) - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[LOAD]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[LOAD]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>) %0:_(p3) = COPY $vgpr0 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 8, addrspace 3) @@ -6126,15 +6131,18 @@ ; SI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load 2 + 4, addrspace 3) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; SI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-LABEL: name: test_load_local_v3s16_align2 ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 @@ -6156,15 +6164,18 @@ ; CI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load 2 + 4, addrspace 3) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; CI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-DS128-LABEL: name: test_load_local_v3s16_align2 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 @@ -6186,15 +6197,18 @@ ; CI-DS128: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load 2 + 4, addrspace 3) ; CI-DS128: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; CI-DS128: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-DS128: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI-DS128: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-DS128: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-DS128: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-DS128: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI-DS128: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI-DS128: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI-DS128: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI-DS128: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI-DS128: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-DS128: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI-DS128: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI-DS128: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; CI-DS128: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI-DS128: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI-DS128: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-DS128: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI-DS128: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI-DS128: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_local_v3s16_align2 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 @@ -6216,15 +6230,18 @@ ; VI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load 2 + 4, addrspace 3) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-LABEL: name: test_load_local_v3s16_align2 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 @@ -6240,15 +6257,18 @@ ; GFX9: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p3) :: (load 2 + 4, addrspace 3) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p3) = COPY $vgpr0 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 2, addrspace 3) @@ -6312,15 +6332,18 @@ ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; SI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; SI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-LABEL: name: test_load_local_v3s16_align1 ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 @@ -6371,15 +6394,18 @@ ; CI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; CI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; CI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-DS128-LABEL: name: test_load_local_v3s16_align1 ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 @@ -6430,15 +6456,18 @@ ; CI-DS128: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; CI-DS128: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; CI-DS128: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-DS128: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI-DS128: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI-DS128: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI-DS128: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI-DS128: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI-DS128: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI-DS128: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI-DS128: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI-DS128: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI-DS128: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-DS128: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI-DS128: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI-DS128: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; CI-DS128: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI-DS128: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI-DS128: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI-DS128: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI-DS128: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI-DS128: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_local_v3s16_align1 ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 @@ -6483,15 +6512,18 @@ ; VI: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; VI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL3]] ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-LABEL: name: test_load_local_v3s16_align1 ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0 @@ -6533,15 +6565,18 @@ ; GFX9: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; GFX9: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL2]] ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR2]](s16), 32 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p3) = COPY $vgpr0 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 1, addrspace 3) Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir @@ -5027,15 +5027,18 @@ ; SI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load 2 + 4, align 4, addrspace 5) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[LOAD]](<2 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; SI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-LABEL: name: test_load_private_v3s16_align8 ; CI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 @@ -5045,15 +5048,18 @@ ; CI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load 2 + 4, align 4, addrspace 5) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32) ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[LOAD]](<2 x s16>), 0 - ; CI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; CI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_private_v3s16_align8 ; VI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 @@ -5063,15 +5069,18 @@ ; VI: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load 2 + 4, align 4, addrspace 5) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[LOAD]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-LABEL: name: test_load_private_v3s16_align8 ; GFX9: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 @@ -5081,15 +5090,18 @@ ; GFX9: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load 2 + 4, align 4, addrspace 5) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD1]](s32) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[LOAD]](<2 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p5) = COPY $vgpr0 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 8, addrspace 5) @@ -5124,15 +5136,18 @@ ; SI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load 2 + 4, addrspace 5) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; SI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-LABEL: name: test_load_private_v3s16_align2 ; CI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 @@ -5154,15 +5169,18 @@ ; CI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load 2 + 4, addrspace 5) ; CI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; CI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_private_v3s16_align2 ; VI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 @@ -5184,15 +5202,18 @@ ; VI: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load 2 + 4, addrspace 5) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-LABEL: name: test_load_private_v3s16_align2 ; GFX9: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 @@ -5208,15 +5229,18 @@ ; GFX9: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load 2 + 4, addrspace 5) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LOAD2]](s32) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[TRUNC]](s16), 32 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p5) = COPY $vgpr0 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 2, addrspace 5) @@ -5280,15 +5304,18 @@ ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; SI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; SI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CI-LABEL: name: test_load_private_v3s16_align1 ; CI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 @@ -5339,15 +5366,18 @@ ; CI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32) ; CI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[TRUNC5]] ; CI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; CI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; CI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; CI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; CI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; CI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; CI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; VI-LABEL: name: test_load_private_v3s16_align1 ; VI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 @@ -5392,15 +5422,18 @@ ; VI: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; VI: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL3]] ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BITCAST]](<2 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR3]](s16), 32 - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; VI: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; GFX9-LABEL: name: test_load_private_v3s16_align1 ; GFX9: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0 @@ -5442,15 +5475,18 @@ ; GFX9: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16) ; GFX9: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL2]] ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), %2(<4 x s16>), %2(<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[BUILD_VECTOR_TRUNC]](<2 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT2]], [[OR2]](s16), 32 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT3]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT2]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) %0:_(p5) = COPY $vgpr0 %1:_(<3 x s16>) = G_LOAD %0 :: (load 6, align 1, addrspace 5) Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-lshr.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-lshr.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-lshr.mir @@ -726,14 +726,17 @@ ; GFX9: [[LSHR:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[EXTRACT2]], [[EXTRACT4]](<2 x s16>) ; GFX9: [[LSHR1:%[0-9]+]]:_(s16) = G_LSHR [[EXTRACT3]], [[EXTRACT5]](s16) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT6:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT6]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT4]], [[LSHR]](<2 x s16>), 0 - ; GFX9: [[EXTRACT7:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT5]](<4 x s16>), 0 - ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT7]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT5]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT7:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT6]], [[LSHR1]](s16), 32 - ; GFX9: [[EXTRACT8:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT7]](<4 x s16>), 0 - ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT8]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT7]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT8]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-or.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-or.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-or.mir @@ -361,15 +361,18 @@ ; CHECK-LABEL: name: test_or_v3s16 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CHECK: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CHECK: [[OR:%[0-9]+]]:_(<4 x s16>) = G_OR [[INSERT]], [[INSERT1]] - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[OR]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT2]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[OR]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF @@ -403,27 +406,32 @@ ; CHECK-LABEL: name: test_or_v5s16 ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[DEF]](<6 x s16>), 0 ; CHECK: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[DEF1]](<6 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>), [[UV2:%[0-9]+]]:_(<5 x s16>), [[UV3:%[0-9]+]]:_(<5 x s16>), [[UV4:%[0-9]+]]:_(<5 x s16>), [[UV5:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s16>) ; CHECK: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<5 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<5 x s16>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT]](<6 x s16>) - ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>) + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF2]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV6:%[0-9]+]]:_(<5 x s16>), [[UV7:%[0-9]+]]:_(<5 x s16>), [[UV8:%[0-9]+]]:_(<5 x s16>), [[UV9:%[0-9]+]]:_(<5 x s16>), [[UV10:%[0-9]+]]:_(<5 x s16>), [[UV11:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<30 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV]](<5 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV6]](<5 x s16>), 0 + ; CHECK: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT]](<6 x s16>) + ; CHECK: [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>) ; CHECK: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV]](<3 x s16>), 0 - ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV12]](<3 x s16>), 0 + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV14]](<3 x s16>), 0 ; CHECK: [[OR:%[0-9]+]]:_(<4 x s16>) = G_OR [[INSERT2]], [[INSERT3]] - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[OR]](<4 x s16>), 0 - ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV1]](<3 x s16>), 0 - ; CHECK: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV3]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[OR]](<4 x s16>), [[DEF3]](<4 x s16>), [[DEF3]](<4 x s16>) + ; CHECK: [[UV16:%[0-9]+]]:_(<3 x s16>), [[UV17:%[0-9]+]]:_(<3 x s16>), [[UV18:%[0-9]+]]:_(<3 x s16>), [[UV19:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV13]](<3 x s16>), 0 + ; CHECK: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV15]](<3 x s16>), 0 ; CHECK: [[OR1:%[0-9]+]]:_(<4 x s16>) = G_OR [[INSERT4]], [[INSERT5]] - ; CHECK: [[EXTRACT3:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[OR1]](<4 x s16>), 0 - ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT2]](<3 x s16>), [[EXTRACT3]](<3 x s16>) - ; CHECK: [[EXTRACT4:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS3:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[OR1]](<4 x s16>), [[DEF3]](<4 x s16>), [[DEF3]](<4 x s16>) + ; CHECK: [[UV20:%[0-9]+]]:_(<3 x s16>), [[UV21:%[0-9]+]]:_(<3 x s16>), [[UV22:%[0-9]+]]:_(<3 x s16>), [[UV23:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<12 x s16>) + ; CHECK: [[CONCAT_VECTORS4:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV16]](<3 x s16>), [[UV20]](<3 x s16>) + ; CHECK: [[CONCAT_VECTORS5:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS4]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV24:%[0-9]+]]:_(<5 x s16>), [[UV25:%[0-9]+]]:_(<5 x s16>), [[UV26:%[0-9]+]]:_(<5 x s16>), [[UV27:%[0-9]+]]:_(<5 x s16>), [[UV28:%[0-9]+]]:_(<5 x s16>), [[UV29:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS5]](<30 x s16>) ; CHECK: [[DEF4:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT6:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF4]], [[EXTRACT4]](<5 x s16>), 0 + ; CHECK: [[INSERT6:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF4]], [[UV24]](<5 x s16>), 0 ; CHECK: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT6]](<8 x s16>) %0:_(<5 x s16>) = G_IMPLICIT_DEF %1:_(<5 x s16>) = G_IMPLICIT_DEF @@ -463,8 +471,9 @@ ; CHECK: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ANYEXT6]], [[ANYEXT7]] ; CHECK: [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[OR3]](s32) ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s8>) = G_BUILD_VECTOR [[TRUNC]](s8), [[TRUNC1]](s8), [[TRUNC2]](s8), [[TRUNC3]](s8) - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s8>) = G_EXTRACT [[BUILD_VECTOR]](<4 x s8>), 0 - ; CHECK: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[EXTRACT]](<3 x s8>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<4 x s8>), [[DEF2]](<4 x s8>), [[DEF2]](<4 x s8>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s8>), [[UV9:%[0-9]+]]:_(<3 x s8>), [[UV10:%[0-9]+]]:_(<3 x s8>), [[UV11:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>) + ; CHECK: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV8]](<3 x s8>) ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT8]](<3 x s32>) %0:_(<3 x s8>) = G_IMPLICIT_DEF %1:_(<3 x s8>) = G_IMPLICIT_DEF Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-phi.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-phi.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-phi.mir @@ -174,9 +174,10 @@ ; CHECK: G_BR %bb.2 ; CHECK: bb.2: ; CHECK: [[PHI:%[0-9]+]]:_(<4 x s16>) = G_PHI [[INSERT]](<4 x s16>), %bb.0, [[INSERT3]](<4 x s16>), %bb.1 - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[PHI]](<4 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[PHI]](<4 x s16>), %7(<4 x s16>), %7(<4 x s16>) + ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) ; CHECK: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT2]](<3 x s16>), 0 + ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[UV4]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>) ; CHECK: S_SETPC_B64 undef $sgpr30_sgpr31 bb.0: Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddsat.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddsat.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddsat.mir @@ -582,9 +582,11 @@ ; GFX6: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX6: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; GFX6: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX6: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX6: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX6: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX6: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX6: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX6: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX8-LABEL: name: saddsat_v3s16 ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX8: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) @@ -644,9 +646,11 @@ ; GFX8: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX8: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; GFX8: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX8: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX8: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX8: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX8: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX8: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX8: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX9-LABEL: name: saddsat_v3s16 ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) @@ -701,9 +705,11 @@ ; GFX9: [[ADD1:%[0-9]+]]:_(<2 x s16>) = G_ADD [[BUILD_VECTOR_TRUNC1]], [[SMIN3]] ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[ADD]](<2 x s16>), [[ADD1]](<2 x s16>), [[DEF2]](<2 x s16>) ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX9: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0 %3:_(<3 x s16>) = G_SADDSAT %1, %2 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-select.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-select.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-select.mir @@ -291,8 +291,9 @@ ; CHECK: [[SELECT3:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[ANYEXT6]], [[ANYEXT7]] ; CHECK: [[TRUNC5:%[0-9]+]]:_(s8) = G_TRUNC [[SELECT3]](s16) ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s8>) = G_BUILD_VECTOR [[TRUNC2]](s8), [[TRUNC3]](s8), [[TRUNC4]](s8), [[TRUNC5]](s8) - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s8>) = G_EXTRACT [[BUILD_VECTOR]](<4 x s8>), 0 - ; CHECK: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[EXTRACT]](<3 x s8>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<4 x s8>), [[DEF]](<4 x s8>), [[DEF]](<4 x s8>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s8>), [[UV9:%[0-9]+]]:_(<3 x s8>), [[UV10:%[0-9]+]]:_(<3 x s8>), [[UV11:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>) + ; CHECK: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV8]](<3 x s8>) ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT8]](<3 x s32>) %0:_(s32) = COPY $vgpr0 %1:_(s32) = G_CONSTANT i32 0 @@ -399,8 +400,9 @@ ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 ; CHECK: [[SELECT:%[0-9]+]]:_(<4 x s16>) = G_SELECT [[ICMP]](s1), [[INSERT]], [[INSERT1]] - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[SELECT]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[SELECT]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(s32) = COPY $vgpr0 %1:_(<4 x s16>) = COPY $vgpr1_vgpr2 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shl.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shl.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shl.mir @@ -710,14 +710,17 @@ ; GFX9: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[EXTRACT2]], [[EXTRACT4]](<2 x s16>) ; GFX9: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[EXTRACT3]], [[EXTRACT5]](s16) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT6:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT6]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 ; GFX9: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT4]], [[SHL]](<2 x s16>), 0 - ; GFX9: [[EXTRACT7:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT5]](<4 x s16>), 0 - ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT7]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT5]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT7:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT6]], [[SHL1]](s16), 32 - ; GFX9: [[EXTRACT8:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT7]](<4 x s16>), 0 - ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT8]](<3 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT7]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT8]](<4 x s16>) %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-smax.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-smax.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-smax.mir @@ -349,21 +349,24 @@ ; SI-LABEL: name: test_smax_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; SI: [[SEXT_INREG:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY]], 16 @@ -394,31 +397,34 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_smax_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[SMAX:%[0-9]+]]:_(s16) = G_SMAX [[TRUNC]], [[TRUNC3]] @@ -434,24 +440,27 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_smax_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[SMAX:%[0-9]+]]:_(<2 x s16>) = G_SMAX [[UV]], [[UV2]] - ; GFX9: [[SMAX1:%[0-9]+]]:_(<2 x s16>) = G_SMAX [[UV1]], [[UV3]] - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[SMAX]](<2 x s16>), [[SMAX1]](<2 x s16>) - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[SMAX:%[0-9]+]]:_(<2 x s16>) = G_SMAX [[UV8]], [[UV10]] + ; GFX9: [[SMAX1:%[0-9]+]]:_(<2 x s16>) = G_SMAX [[UV9]], [[UV11]] + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[SMAX]](<2 x s16>), [[SMAX1]](<2 x s16>) + ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<12 x s16>) + ; GFX9: S_NOP 0, implicit [[UV12]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF %2:_(<3 x s16>) = G_SMAX %0, %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-smin.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-smin.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-smin.mir @@ -349,21 +349,24 @@ ; SI-LABEL: name: test_smin_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; SI: [[SEXT_INREG:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY]], 16 @@ -394,31 +397,34 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_smin_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[SMIN:%[0-9]+]]:_(s16) = G_SMIN [[TRUNC]], [[TRUNC3]] @@ -434,24 +440,27 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_smin_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[SMIN:%[0-9]+]]:_(<2 x s16>) = G_SMIN [[UV]], [[UV2]] - ; GFX9: [[SMIN1:%[0-9]+]]:_(<2 x s16>) = G_SMIN [[UV1]], [[UV3]] - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[SMIN]](<2 x s16>), [[SMIN1]](<2 x s16>) - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[SMIN:%[0-9]+]]:_(<2 x s16>) = G_SMIN [[UV8]], [[UV10]] + ; GFX9: [[SMIN1:%[0-9]+]]:_(<2 x s16>) = G_SMIN [[UV9]], [[UV11]] + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[SMIN]](<2 x s16>), [[SMIN1]](<2 x s16>) + ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<12 x s16>) + ; GFX9: S_NOP 0, implicit [[UV12]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF %2:_(<3 x s16>) = G_SMIN %0, %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubsat.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubsat.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubsat.mir @@ -582,9 +582,11 @@ ; GFX6: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX6: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; GFX6: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX6: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX6: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX6: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX6: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX6: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX6: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX8-LABEL: name: ssubsat_v3s16 ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX8: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) @@ -644,9 +646,11 @@ ; GFX8: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX8: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; GFX8: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX8: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX8: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX8: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX8: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX8: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX8: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX9-LABEL: name: ssubsat_v3s16 ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) @@ -701,9 +705,11 @@ ; GFX9: [[SUB5:%[0-9]+]]:_(<2 x s16>) = G_SUB [[BUILD_VECTOR_TRUNC1]], [[SMIN3]] ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[SUB2]](<2 x s16>), [[SUB5]](<2 x s16>), [[DEF2]](<2 x s16>) ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX9: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0 %3:_(<3 x s16>) = G_SSUBSAT %1, %2 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-uaddsat.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-uaddsat.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-uaddsat.mir @@ -456,9 +456,11 @@ ; GFX6: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX6: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; GFX6: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX6: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX6: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX6: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX6: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX6: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX6: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX8-LABEL: name: uaddsat_v3s16 ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX8: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) @@ -504,9 +506,11 @@ ; GFX8: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX8: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; GFX8: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX8: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX8: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX8: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX8: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX8: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX8: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX9-LABEL: name: uaddsat_v3s16 ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) @@ -547,9 +551,11 @@ ; GFX9: [[ADD1:%[0-9]+]]:_(<2 x s16>) = G_ADD [[BUILD_VECTOR_TRUNC1]], [[UMIN1]] ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[ADD]](<2 x s16>), [[ADD1]](<2 x s16>), [[DEF2]](<2 x s16>) ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX9: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0 %3:_(<3 x s16>) = G_UADDSAT %1, %2 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-umax.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-umax.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-umax.mir @@ -351,21 +351,24 @@ ; SI-LABEL: name: test_umax_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) @@ -396,31 +399,34 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_umax_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[UMAX:%[0-9]+]]:_(s16) = G_UMAX [[TRUNC]], [[TRUNC3]] @@ -436,24 +442,27 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_umax_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[UMAX:%[0-9]+]]:_(<2 x s16>) = G_UMAX [[UV]], [[UV2]] - ; GFX9: [[UMAX1:%[0-9]+]]:_(<2 x s16>) = G_UMAX [[UV1]], [[UV3]] - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[UMAX]](<2 x s16>), [[UMAX1]](<2 x s16>) - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[UMAX:%[0-9]+]]:_(<2 x s16>) = G_UMAX [[UV8]], [[UV10]] + ; GFX9: [[UMAX1:%[0-9]+]]:_(<2 x s16>) = G_UMAX [[UV9]], [[UV11]] + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[UMAX]](<2 x s16>), [[UMAX1]](<2 x s16>) + ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<12 x s16>) + ; GFX9: S_NOP 0, implicit [[UV12]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF %2:_(<3 x s16>) = G_UMAX %0, %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-umin.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-umin.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-umin.mir @@ -351,21 +351,24 @@ ; SI-LABEL: name: test_umin_v3s16 ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 ; SI: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) @@ -396,31 +399,34 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; VI-LABEL: name: test_umin_v3s16 ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[UMIN:%[0-9]+]]:_(s16) = G_UMIN [[TRUNC]], [[TRUNC3]] @@ -436,24 +442,27 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 + ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) ; GFX9-LABEL: name: test_umin_v3s16 ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[UMIN:%[0-9]+]]:_(<2 x s16>) = G_UMIN [[UV]], [[UV2]] - ; GFX9: [[UMIN1:%[0-9]+]]:_(<2 x s16>) = G_UMIN [[UV1]], [[UV3]] - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[UMIN]](<2 x s16>), [[UMIN1]](<2 x s16>) - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT2]](<3 x s16>) + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[UMIN:%[0-9]+]]:_(<2 x s16>) = G_UMIN [[UV8]], [[UV10]] + ; GFX9: [[UMIN1:%[0-9]+]]:_(<2 x s16>) = G_UMIN [[UV9]], [[UV11]] + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[UMIN]](<2 x s16>), [[UMIN1]](<2 x s16>) + ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; GFX9: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<12 x s16>) + ; GFX9: S_NOP 0, implicit [[UV12]](<3 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF %2:_(<3 x s16>) = G_UMIN %0, %1 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-unmerge-values.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-unmerge-values.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-unmerge-values.mir @@ -60,13 +60,15 @@ bb.0: ; CHECK-LABEL: name: test_unmerge_s16_v3s16 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; CHECK: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-usubsat.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-usubsat.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-usubsat.mir @@ -416,9 +416,11 @@ ; GFX6: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX6: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; GFX6: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX6: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX6: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX6: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX6: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX6: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX6: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX8-LABEL: name: usubsat_v3s16 ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX8: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) @@ -460,9 +462,11 @@ ; GFX8: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX8: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; GFX8: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX8: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX8: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX8: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX8: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX8: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX8: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX9-LABEL: name: usubsat_v3s16 ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) @@ -498,9 +502,11 @@ ; GFX9: [[SUB1:%[0-9]+]]:_(<2 x s16>) = G_SUB [[BUILD_VECTOR_TRUNC1]], [[UMIN1]] ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[SUB]](<2 x s16>), [[SUB1]](<2 x s16>), [[DEF2]](<2 x s16>) ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 - ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) - ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>) + ; GFX9: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>), [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[UV6]](<3 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0 %3:_(<3 x s16>) = G_USUBSAT %1, %2 Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-xor.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-xor.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-xor.mir @@ -361,15 +361,18 @@ ; CHECK-LABEL: name: test_xor_v3s16 ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF1]](<4 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) ; CHECK: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 ; CHECK: [[XOR:%[0-9]+]]:_(<4 x s16>) = G_XOR [[INSERT]], [[INSERT1]] - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[XOR]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT2]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[XOR]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) %0:_(<3 x s16>) = G_IMPLICIT_DEF %1:_(<3 x s16>) = G_IMPLICIT_DEF @@ -403,27 +406,32 @@ ; CHECK-LABEL: name: test_xor_v5s16 ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[DEF]](<6 x s16>), 0 ; CHECK: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[DEF1]](<6 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>), [[UV2:%[0-9]+]]:_(<5 x s16>), [[UV3:%[0-9]+]]:_(<5 x s16>), [[UV4:%[0-9]+]]:_(<5 x s16>), [[UV5:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s16>) ; CHECK: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[EXTRACT]](<5 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<5 x s16>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT]](<6 x s16>) - ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>) + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF2]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV6:%[0-9]+]]:_(<5 x s16>), [[UV7:%[0-9]+]]:_(<5 x s16>), [[UV8:%[0-9]+]]:_(<5 x s16>), [[UV9:%[0-9]+]]:_(<5 x s16>), [[UV10:%[0-9]+]]:_(<5 x s16>), [[UV11:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<30 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV]](<5 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV6]](<5 x s16>), 0 + ; CHECK: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT]](<6 x s16>) + ; CHECK: [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>) ; CHECK: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV]](<3 x s16>), 0 - ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV12]](<3 x s16>), 0 + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV14]](<3 x s16>), 0 ; CHECK: [[XOR:%[0-9]+]]:_(<4 x s16>) = G_XOR [[INSERT2]], [[INSERT3]] - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[XOR]](<4 x s16>), 0 - ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV1]](<3 x s16>), 0 - ; CHECK: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV3]](<3 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[XOR]](<4 x s16>), [[DEF3]](<4 x s16>), [[DEF3]](<4 x s16>) + ; CHECK: [[UV16:%[0-9]+]]:_(<3 x s16>), [[UV17:%[0-9]+]]:_(<3 x s16>), [[UV18:%[0-9]+]]:_(<3 x s16>), [[UV19:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV13]](<3 x s16>), 0 + ; CHECK: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV15]](<3 x s16>), 0 ; CHECK: [[XOR1:%[0-9]+]]:_(<4 x s16>) = G_XOR [[INSERT4]], [[INSERT5]] - ; CHECK: [[EXTRACT3:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[XOR1]](<4 x s16>), 0 - ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT2]](<3 x s16>), [[EXTRACT3]](<3 x s16>) - ; CHECK: [[EXTRACT4:%[0-9]+]]:_(<5 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS3:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[XOR1]](<4 x s16>), [[DEF3]](<4 x s16>), [[DEF3]](<4 x s16>) + ; CHECK: [[UV20:%[0-9]+]]:_(<3 x s16>), [[UV21:%[0-9]+]]:_(<3 x s16>), [[UV22:%[0-9]+]]:_(<3 x s16>), [[UV23:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<12 x s16>) + ; CHECK: [[CONCAT_VECTORS4:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV16]](<3 x s16>), [[UV20]](<3 x s16>) + ; CHECK: [[CONCAT_VECTORS5:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS4]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>) + ; CHECK: [[UV24:%[0-9]+]]:_(<5 x s16>), [[UV25:%[0-9]+]]:_(<5 x s16>), [[UV26:%[0-9]+]]:_(<5 x s16>), [[UV27:%[0-9]+]]:_(<5 x s16>), [[UV28:%[0-9]+]]:_(<5 x s16>), [[UV29:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS5]](<30 x s16>) ; CHECK: [[DEF4:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT6:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF4]], [[EXTRACT4]](<5 x s16>), 0 + ; CHECK: [[INSERT6:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF4]], [[UV24]](<5 x s16>), 0 ; CHECK: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT6]](<8 x s16>) %0:_(<5 x s16>) = G_IMPLICIT_DEF %1:_(<5 x s16>) = G_IMPLICIT_DEF @@ -463,8 +471,9 @@ ; CHECK: [[XOR3:%[0-9]+]]:_(s32) = G_XOR [[ANYEXT6]], [[ANYEXT7]] ; CHECK: [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[XOR3]](s32) ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s8>) = G_BUILD_VECTOR [[TRUNC]](s8), [[TRUNC1]](s8), [[TRUNC2]](s8), [[TRUNC3]](s8) - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s8>) = G_EXTRACT [[BUILD_VECTOR]](<4 x s8>), 0 - ; CHECK: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[EXTRACT]](<3 x s8>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<4 x s8>), [[DEF2]](<4 x s8>), [[DEF2]](<4 x s8>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s8>), [[UV9:%[0-9]+]]:_(<3 x s8>), [[UV10:%[0-9]+]]:_(<3 x s8>), [[UV11:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>) + ; CHECK: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV8]](<3 x s8>) ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT8]](<3 x s32>) %0:_(<3 x s8>) = G_IMPLICIT_DEF %1:_(<3 x s8>) = G_IMPLICIT_DEF Index: llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.s.buffer.load.ll =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.s.buffer.load.ll +++ llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.s.buffer.load.ll @@ -172,19 +172,24 @@ ; GFX6: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5 ; GFX6: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr6 ; GFX6: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 - ; GFX6: [[S_BUFFER_LOAD_DWORDX4_SGPR:%[0-9]+]]:sgpr_128_with_sub0_sub1_sub2 = S_BUFFER_LOAD_DWORDX4_SGPR [[REG_SEQUENCE]], [[COPY4]], 0, 0 :: (dereferenceable invariant load 12, align 4) - ; GFX6: [[COPY5:%[0-9]+]]:sgpr_96 = COPY [[S_BUFFER_LOAD_DWORDX4_SGPR]].sub0_sub1_sub2 - ; GFX6: [[COPY6:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub0 - ; GFX6: [[COPY7:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub1 - ; GFX6: [[COPY8:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub2 - ; GFX6: [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[COPY6]] - ; GFX6: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY9]], implicit $exec + ; GFX6: [[S_BUFFER_LOAD_DWORDX4_SGPR:%[0-9]+]]:sgpr_128 = S_BUFFER_LOAD_DWORDX4_SGPR [[REG_SEQUENCE]], [[COPY4]], 0, 0 :: (dereferenceable invariant load 12, align 4) + ; GFX6: [[DEF:%[0-9]+]]:sgpr_128 = IMPLICIT_DEF + ; GFX6: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_512 = REG_SEQUENCE [[S_BUFFER_LOAD_DWORDX4_SGPR]], %subreg.sub0_sub1_sub2_sub3, [[DEF]], %subreg.sub4_sub5_sub6_sub7, [[DEF]], %subreg.sub8_sub9_sub10_sub11 + ; GFX6: [[COPY5:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub0_sub1_sub2 + ; GFX6: [[COPY6:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub3_sub4_sub5 + ; GFX6: [[COPY7:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub6_sub7_sub8 + ; GFX6: [[COPY8:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub9_sub10_sub11 + ; GFX6: [[COPY9:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub0 + ; GFX6: [[COPY10:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub1 + ; GFX6: [[COPY11:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub2 + ; GFX6: [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[COPY9]] + ; GFX6: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY12]], implicit $exec ; GFX6: $sgpr0 = COPY [[V_READFIRSTLANE_B32_]] - ; GFX6: [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[COPY7]] - ; GFX6: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY10]], implicit $exec + ; GFX6: [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[COPY10]] + ; GFX6: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY13]], implicit $exec ; GFX6: $sgpr1 = COPY [[V_READFIRSTLANE_B32_1]] - ; GFX6: [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[COPY8]] - ; GFX6: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY11]], implicit $exec + ; GFX6: [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[COPY11]] + ; GFX6: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY14]], implicit $exec ; GFX6: $sgpr2 = COPY [[V_READFIRSTLANE_B32_2]] ; GFX6: SI_RETURN_TO_EPILOG implicit $sgpr0, implicit $sgpr1, implicit $sgpr2 ; GFX7-LABEL: name: s_buffer_load_v3i32 @@ -196,19 +201,24 @@ ; GFX7: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5 ; GFX7: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr6 ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 - ; GFX7: [[S_BUFFER_LOAD_DWORDX4_SGPR:%[0-9]+]]:sgpr_128_with_sub0_sub1_sub2 = S_BUFFER_LOAD_DWORDX4_SGPR [[REG_SEQUENCE]], [[COPY4]], 0, 0 :: (dereferenceable invariant load 12, align 4) - ; GFX7: [[COPY5:%[0-9]+]]:sgpr_96 = COPY [[S_BUFFER_LOAD_DWORDX4_SGPR]].sub0_sub1_sub2 - ; GFX7: [[COPY6:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub0 - ; GFX7: [[COPY7:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub1 - ; GFX7: [[COPY8:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub2 - ; GFX7: [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[COPY6]] - ; GFX7: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY9]], implicit $exec + ; GFX7: [[S_BUFFER_LOAD_DWORDX4_SGPR:%[0-9]+]]:sgpr_128 = S_BUFFER_LOAD_DWORDX4_SGPR [[REG_SEQUENCE]], [[COPY4]], 0, 0 :: (dereferenceable invariant load 12, align 4) + ; GFX7: [[DEF:%[0-9]+]]:sgpr_128 = IMPLICIT_DEF + ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_512 = REG_SEQUENCE [[S_BUFFER_LOAD_DWORDX4_SGPR]], %subreg.sub0_sub1_sub2_sub3, [[DEF]], %subreg.sub4_sub5_sub6_sub7, [[DEF]], %subreg.sub8_sub9_sub10_sub11 + ; GFX7: [[COPY5:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub0_sub1_sub2 + ; GFX7: [[COPY6:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub3_sub4_sub5 + ; GFX7: [[COPY7:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub6_sub7_sub8 + ; GFX7: [[COPY8:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub9_sub10_sub11 + ; GFX7: [[COPY9:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub0 + ; GFX7: [[COPY10:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub1 + ; GFX7: [[COPY11:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub2 + ; GFX7: [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[COPY9]] + ; GFX7: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY12]], implicit $exec ; GFX7: $sgpr0 = COPY [[V_READFIRSTLANE_B32_]] - ; GFX7: [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[COPY7]] - ; GFX7: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY10]], implicit $exec + ; GFX7: [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[COPY10]] + ; GFX7: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY13]], implicit $exec ; GFX7: $sgpr1 = COPY [[V_READFIRSTLANE_B32_1]] - ; GFX7: [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[COPY8]] - ; GFX7: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY11]], implicit $exec + ; GFX7: [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[COPY11]] + ; GFX7: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY14]], implicit $exec ; GFX7: $sgpr2 = COPY [[V_READFIRSTLANE_B32_2]] ; GFX7: SI_RETURN_TO_EPILOG implicit $sgpr0, implicit $sgpr1, implicit $sgpr2 ; GFX8-LABEL: name: s_buffer_load_v3i32 @@ -220,19 +230,24 @@ ; GFX8: [[COPY3:%[0-9]+]]:sreg_32 = COPY $sgpr5 ; GFX8: [[COPY4:%[0-9]+]]:sreg_32 = COPY $sgpr6 ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 - ; GFX8: [[S_BUFFER_LOAD_DWORDX4_SGPR:%[0-9]+]]:sgpr_128_with_sub0_sub1_sub2 = S_BUFFER_LOAD_DWORDX4_SGPR [[REG_SEQUENCE]], [[COPY4]], 0, 0 :: (dereferenceable invariant load 12, align 4) - ; GFX8: [[COPY5:%[0-9]+]]:sgpr_96 = COPY [[S_BUFFER_LOAD_DWORDX4_SGPR]].sub0_sub1_sub2 - ; GFX8: [[COPY6:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub0 - ; GFX8: [[COPY7:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub1 - ; GFX8: [[COPY8:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub2 - ; GFX8: [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[COPY6]] - ; GFX8: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY9]], implicit $exec + ; GFX8: [[S_BUFFER_LOAD_DWORDX4_SGPR:%[0-9]+]]:sgpr_128 = S_BUFFER_LOAD_DWORDX4_SGPR [[REG_SEQUENCE]], [[COPY4]], 0, 0 :: (dereferenceable invariant load 12, align 4) + ; GFX8: [[DEF:%[0-9]+]]:sgpr_128 = IMPLICIT_DEF + ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_512 = REG_SEQUENCE [[S_BUFFER_LOAD_DWORDX4_SGPR]], %subreg.sub0_sub1_sub2_sub3, [[DEF]], %subreg.sub4_sub5_sub6_sub7, [[DEF]], %subreg.sub8_sub9_sub10_sub11 + ; GFX8: [[COPY5:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub0_sub1_sub2 + ; GFX8: [[COPY6:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub3_sub4_sub5 + ; GFX8: [[COPY7:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub6_sub7_sub8 + ; GFX8: [[COPY8:%[0-9]+]]:sgpr_96 = COPY [[REG_SEQUENCE1]].sub9_sub10_sub11 + ; GFX8: [[COPY9:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub0 + ; GFX8: [[COPY10:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub1 + ; GFX8: [[COPY11:%[0-9]+]]:sreg_32 = COPY [[COPY5]].sub2 + ; GFX8: [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[COPY9]] + ; GFX8: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY12]], implicit $exec ; GFX8: $sgpr0 = COPY [[V_READFIRSTLANE_B32_]] - ; GFX8: [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[COPY7]] - ; GFX8: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY10]], implicit $exec + ; GFX8: [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[COPY10]] + ; GFX8: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY13]], implicit $exec ; GFX8: $sgpr1 = COPY [[V_READFIRSTLANE_B32_1]] - ; GFX8: [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[COPY8]] - ; GFX8: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY11]], implicit $exec + ; GFX8: [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[COPY11]] + ; GFX8: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32 = V_READFIRSTLANE_B32 [[COPY14]], implicit $exec ; GFX8: $sgpr2 = COPY [[V_READFIRSTLANE_B32_2]] ; GFX8: SI_RETURN_TO_EPILOG implicit $sgpr0, implicit $sgpr1, implicit $sgpr2 %val = call <3 x i32> @llvm.amdgcn.s.buffer.load.v3i32(<4 x i32> %rsrc, i32 %soffset, i32 0) @@ -1580,13 +1595,20 @@ ; GFX6: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 ; GFX6: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 ; GFX6: [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16, align 4) - ; GFX6: [[COPY5:%[0-9]+]]:vreg_96 = COPY [[BUFFER_LOAD_DWORDX4_OFFEN]].sub0_sub1_sub2 - ; GFX6: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub0 - ; GFX6: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub1 - ; GFX6: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub2 - ; GFX6: $vgpr0 = COPY [[COPY6]] - ; GFX6: $vgpr1 = COPY [[COPY7]] - ; GFX6: $vgpr2 = COPY [[COPY8]] + ; GFX6: [[DEF:%[0-9]+]]:sgpr_128 = IMPLICIT_DEF + ; GFX6: [[COPY5:%[0-9]+]]:vreg_128 = COPY [[DEF]] + ; GFX6: [[COPY6:%[0-9]+]]:vreg_128 = COPY [[DEF]] + ; GFX6: [[REG_SEQUENCE1:%[0-9]+]]:vreg_512 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[COPY5]], %subreg.sub4_sub5_sub6_sub7, [[COPY6]], %subreg.sub8_sub9_sub10_sub11 + ; GFX6: [[COPY7:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub0_sub1_sub2 + ; GFX6: [[COPY8:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub3_sub4_sub5 + ; GFX6: [[COPY9:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub6_sub7_sub8 + ; GFX6: [[COPY10:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub9_sub10_sub11 + ; GFX6: [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[COPY7]].sub0 + ; GFX6: [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[COPY7]].sub1 + ; GFX6: [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[COPY7]].sub2 + ; GFX6: $vgpr0 = COPY [[COPY11]] + ; GFX6: $vgpr1 = COPY [[COPY12]] + ; GFX6: $vgpr2 = COPY [[COPY13]] ; GFX6: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2 ; GFX7-LABEL: name: s_buffer_load_v3f32_vgpr_offset ; GFX7: bb.1 (%ir-block.0): @@ -1599,13 +1621,20 @@ ; GFX7: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 ; GFX7: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 ; GFX7: [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16, align 4) - ; GFX7: [[COPY5:%[0-9]+]]:vreg_96 = COPY [[BUFFER_LOAD_DWORDX4_OFFEN]].sub0_sub1_sub2 - ; GFX7: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub0 - ; GFX7: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub1 - ; GFX7: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub2 - ; GFX7: $vgpr0 = COPY [[COPY6]] - ; GFX7: $vgpr1 = COPY [[COPY7]] - ; GFX7: $vgpr2 = COPY [[COPY8]] + ; GFX7: [[DEF:%[0-9]+]]:sgpr_128 = IMPLICIT_DEF + ; GFX7: [[COPY5:%[0-9]+]]:vreg_128 = COPY [[DEF]] + ; GFX7: [[COPY6:%[0-9]+]]:vreg_128 = COPY [[DEF]] + ; GFX7: [[REG_SEQUENCE1:%[0-9]+]]:vreg_512 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[COPY5]], %subreg.sub4_sub5_sub6_sub7, [[COPY6]], %subreg.sub8_sub9_sub10_sub11 + ; GFX7: [[COPY7:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub0_sub1_sub2 + ; GFX7: [[COPY8:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub3_sub4_sub5 + ; GFX7: [[COPY9:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub6_sub7_sub8 + ; GFX7: [[COPY10:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub9_sub10_sub11 + ; GFX7: [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[COPY7]].sub0 + ; GFX7: [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[COPY7]].sub1 + ; GFX7: [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[COPY7]].sub2 + ; GFX7: $vgpr0 = COPY [[COPY11]] + ; GFX7: $vgpr1 = COPY [[COPY12]] + ; GFX7: $vgpr2 = COPY [[COPY13]] ; GFX7: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2 ; GFX8-LABEL: name: s_buffer_load_v3f32_vgpr_offset ; GFX8: bb.1 (%ir-block.0): @@ -1618,13 +1647,20 @@ ; GFX8: [[REG_SEQUENCE:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3 ; GFX8: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 ; GFX8: [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE]], [[S_MOV_B32_]], 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16, align 4) - ; GFX8: [[COPY5:%[0-9]+]]:vreg_96 = COPY [[BUFFER_LOAD_DWORDX4_OFFEN]].sub0_sub1_sub2 - ; GFX8: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub0 - ; GFX8: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub1 - ; GFX8: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY5]].sub2 - ; GFX8: $vgpr0 = COPY [[COPY6]] - ; GFX8: $vgpr1 = COPY [[COPY7]] - ; GFX8: $vgpr2 = COPY [[COPY8]] + ; GFX8: [[DEF:%[0-9]+]]:sgpr_128 = IMPLICIT_DEF + ; GFX8: [[COPY5:%[0-9]+]]:vreg_128 = COPY [[DEF]] + ; GFX8: [[COPY6:%[0-9]+]]:vreg_128 = COPY [[DEF]] + ; GFX8: [[REG_SEQUENCE1:%[0-9]+]]:vreg_512 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[COPY5]], %subreg.sub4_sub5_sub6_sub7, [[COPY6]], %subreg.sub8_sub9_sub10_sub11 + ; GFX8: [[COPY7:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub0_sub1_sub2 + ; GFX8: [[COPY8:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub3_sub4_sub5 + ; GFX8: [[COPY9:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub6_sub7_sub8 + ; GFX8: [[COPY10:%[0-9]+]]:vreg_96 = COPY [[REG_SEQUENCE1]].sub9_sub10_sub11 + ; GFX8: [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[COPY7]].sub0 + ; GFX8: [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[COPY7]].sub1 + ; GFX8: [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[COPY7]].sub2 + ; GFX8: $vgpr0 = COPY [[COPY11]] + ; GFX8: $vgpr1 = COPY [[COPY12]] + ; GFX8: $vgpr2 = COPY [[COPY13]] ; GFX8: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2 %val = call <3 x float> @llvm.amdgcn.s.buffer.load.v3f32(<4 x i32> %rsrc, i32 %soffset, i32 0) ret <3 x float> %val Index: llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-amdgcn.s.buffer.load.ll =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-amdgcn.s.buffer.load.ll +++ llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-amdgcn.s.buffer.load.ll @@ -56,15 +56,17 @@ ; CHECK: [[COPY4:%[0-9]+]]:sgpr(s32) = COPY $sgpr6 ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32) ; CHECK: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:sgpr(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[COPY4]](s32), 0 :: (dereferenceable invariant load 12, align 4) - ; CHECK: [[EXTRACT:%[0-9]+]]:sgpr(<3 x s32>) = G_EXTRACT [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), 0 - ; CHECK: [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32), [[UV2:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s32>) - ; CHECK: [[COPY5:%[0-9]+]]:vgpr(s32) = COPY [[UV]](s32) + ; CHECK: [[DEF:%[0-9]+]]:sgpr(<4 x s32>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:sgpr(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>) + ; CHECK: [[UV:%[0-9]+]]:sgpr(<3 x s32>), [[UV1:%[0-9]+]]:sgpr(<3 x s32>), [[UV2:%[0-9]+]]:sgpr(<3 x s32>), [[UV3:%[0-9]+]]:sgpr(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) + ; CHECK: [[UV4:%[0-9]+]]:sgpr(s32), [[UV5:%[0-9]+]]:sgpr(s32), [[UV6:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[UV]](<3 x s32>) + ; CHECK: [[COPY5:%[0-9]+]]:vgpr(s32) = COPY [[UV4]](s32) ; CHECK: [[INT:%[0-9]+]]:sgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.readfirstlane), [[COPY5]](s32) ; CHECK: $sgpr0 = COPY [[INT]](s32) - ; CHECK: [[COPY6:%[0-9]+]]:vgpr(s32) = COPY [[UV1]](s32) + ; CHECK: [[COPY6:%[0-9]+]]:vgpr(s32) = COPY [[UV5]](s32) ; CHECK: [[INT1:%[0-9]+]]:sgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.readfirstlane), [[COPY6]](s32) ; CHECK: $sgpr1 = COPY [[INT1]](s32) - ; CHECK: [[COPY7:%[0-9]+]]:vgpr(s32) = COPY [[UV2]](s32) + ; CHECK: [[COPY7:%[0-9]+]]:vgpr(s32) = COPY [[UV6]](s32) ; CHECK: [[INT2:%[0-9]+]]:sgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.readfirstlane), [[COPY7]](s32) ; CHECK: $sgpr2 = COPY [[INT2]](s32) ; CHECK: SI_RETURN_TO_EPILOG implicit $sgpr0, implicit $sgpr1, implicit $sgpr2 @@ -232,11 +234,15 @@ ; CHECK: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 ; CHECK: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 ; CHECK: [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 0, 0, 0 :: (dereferenceable invariant load 16, align 4) - ; CHECK: [[EXTRACT:%[0-9]+]]:vgpr(<3 x s32>) = G_EXTRACT [[AMDGPU_BUFFER_LOAD]](<4 x s32>), 0 - ; CHECK: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32), [[UV2:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s32>) - ; CHECK: $vgpr0 = COPY [[UV]](s32) - ; CHECK: $vgpr1 = COPY [[UV1]](s32) - ; CHECK: $vgpr2 = COPY [[UV2]](s32) + ; CHECK: [[DEF:%[0-9]+]]:sgpr(<4 x s32>) = G_IMPLICIT_DEF + ; CHECK: [[COPY5:%[0-9]+]]:vgpr(<4 x s32>) = COPY [[DEF]](<4 x s32>) + ; CHECK: [[COPY6:%[0-9]+]]:vgpr(<4 x s32>) = COPY [[DEF]](<4 x s32>) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[COPY5]](<4 x s32>), [[COPY6]](<4 x s32>) + ; CHECK: [[UV:%[0-9]+]]:vgpr(<3 x s32>), [[UV1:%[0-9]+]]:vgpr(<3 x s32>), [[UV2:%[0-9]+]]:vgpr(<3 x s32>), [[UV3:%[0-9]+]]:vgpr(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>) + ; CHECK: [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[UV]](<3 x s32>) + ; CHECK: $vgpr0 = COPY [[UV4]](s32) + ; CHECK: $vgpr1 = COPY [[UV5]](s32) + ; CHECK: $vgpr2 = COPY [[UV6]](s32) ; CHECK: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2 %val = call <3 x float> @llvm.amdgcn.s.buffer.load.v3f32(<4 x i32> %rsrc, i32 %soffset, i32 0) ret <3 x float> %val Index: llvm/unittests/CodeGen/GlobalISel/LegalizerHelperTest.cpp =================================================================== --- llvm/unittests/CodeGen/GlobalISel/LegalizerHelperTest.cpp +++ llvm/unittests/CodeGen/GlobalISel/LegalizerHelperTest.cpp @@ -3041,11 +3041,10 @@ const auto *CheckStr = R"( CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY CHECK: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY]] - CHECK: [[UNDEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF CHECK: [[CV:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[BITCAST]]:_(<2 x s32>), [[UNDEF]] CHECK: [[FREEZE:%[0-9]+]]:_(<4 x s32>) = G_FREEZE [[CV]] - CHECK: [[EXTR:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[FREEZE]]:_(<4 x s32>), 0 + CHECK: [[EXTR0:%[0-9]+]]:_(<2 x s32>), [[EXTR1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[FREEZE]]:_(<4 x s32>) )"; // Check