diff options
4 files changed, 193 insertions, 1 deletions
diff --git a/llvm/lib/Target/AArch64/AArch64GenRegisterBankInfo.def b/llvm/lib/Target/AArch64/AArch64GenRegisterBankInfo.def index 8f17ae4534c..39f50ade747 100644 --- a/llvm/lib/Target/AArch64/AArch64GenRegisterBankInfo.def +++ b/llvm/lib/Target/AArch64/AArch64GenRegisterBankInfo.def @@ -98,6 +98,18 @@ RegisterBankInfo::ValueMapping AArch64GenRegisterBankInfo::ValMappings[]{ // LastCrossRegCpyIdx. {&AArch64GenRegisterBankInfo::PartMappings[PMI_GPR64 - PMI_Min], 1}, {&AArch64GenRegisterBankInfo::PartMappings[PMI_FPR64 - PMI_Min], 1}, + // 41: FPExt: 16 to 32. <-- This must match FPExt16To32Idx. + {&AArch64GenRegisterBankInfo::PartMappings[PMI_FPR32 - PMI_Min], 1}, + {&AArch64GenRegisterBankInfo::PartMappings[PMI_FPR16 - PMI_Min], 1}, + // 43: FPExt: 16 to 32. <-- This must match FPExt16To64Idx. + {&AArch64GenRegisterBankInfo::PartMappings[PMI_FPR64 - PMI_Min], 1}, + {&AArch64GenRegisterBankInfo::PartMappings[PMI_FPR16 - PMI_Min], 1}, + // 45: FPExt: 32 to 64. <-- This must match FPExt32To64Idx. + {&AArch64GenRegisterBankInfo::PartMappings[PMI_FPR64 - PMI_Min], 1}, + {&AArch64GenRegisterBankInfo::PartMappings[PMI_FPR32 - PMI_Min], 1}, + // 47: FPExt vector: 64 to 128. <-- This must match FPExt64To128Idx. + {&AArch64GenRegisterBankInfo::PartMappings[PMI_FPR128 - PMI_Min], 1}, + {&AArch64GenRegisterBankInfo::PartMappings[PMI_FPR64 - PMI_Min], 1}, }; bool AArch64GenRegisterBankInfo::checkPartialMap(unsigned Idx, @@ -217,4 +229,35 @@ AArch64GenRegisterBankInfo::getCopyMapping(unsigned DstBankID, ValMappingIdx <= LastCrossRegCpyIdx && "Mapping out of bound"); return &ValMappings[ValMappingIdx]; } + +const RegisterBankInfo::ValueMapping * +AArch64GenRegisterBankInfo::getFPExtMapping(unsigned DstSize, + unsigned SrcSize) { + // We support: + // - For Scalar: + // - 16 to 32. + // - 16 to 64. + // - 32 to 64. + // => FPR 16 to FPR 32|64 + // => FPR 32 to FPR 64 + // - For vectors: + // - v4f16 to v4f32 + // - v2f32 to v2f64 + // => FPR 64 to FPR 128 + + // Check that we have been asked sensible sizes. + if (SrcSize == 16) { + assert((DstSize == 32 || DstSize == 64) && "Unexpected half extension"); + if (DstSize == 32) + return &ValMappings[FPExt16To32Idx]; + return &ValMappings[FPExt16To64Idx]; + } + + if (SrcSize == 32) { + assert(DstSize == 64 && "Unexpected float extension"); + return &ValMappings[FPExt32To64Idx]; + } + assert((SrcSize == 64 || DstSize == 128) && "Unexpected vector extension"); + return &ValMappings[FPExt64To128Idx]; +} } // End llvm namespace. diff --git a/llvm/lib/Target/AArch64/AArch64RegisterBankInfo.cpp b/llvm/lib/Target/AArch64/AArch64RegisterBankInfo.cpp index 6e246a798c5..83bf493c9f0 100644 --- a/llvm/lib/Target/AArch64/AArch64RegisterBankInfo.cpp +++ b/llvm/lib/Target/AArch64/AArch64RegisterBankInfo.cpp @@ -175,6 +175,30 @@ AArch64RegisterBankInfo::AArch64RegisterBankInfo(const TargetRegisterInfo &TRI) CHECK_VALUEMAP_CROSSREGCPY(FPR, FPR, 64); CHECK_VALUEMAP_CROSSREGCPY(FPR, GPR, 64); +#define CHECK_VALUEMAP_FPEXT(DstSize, SrcSize) \ + do { \ + unsigned PartialMapDstIdx = PMI_FPR##DstSize - PMI_Min; \ + unsigned PartialMapSrcIdx = PMI_FPR##SrcSize - PMI_Min; \ + (void)PartialMapDstIdx; \ + (void)PartialMapSrcIdx; \ + const ValueMapping *Map = getFPExtMapping(DstSize, SrcSize); \ + (void)Map; \ + assert(Map[0].BreakDown == \ + &AArch64GenRegisterBankInfo::PartMappings[PartialMapDstIdx] && \ + Map[0].NumBreakDowns == 1 && "FPR" #DstSize \ + " Dst is incorrectly initialized"); \ + assert(Map[1].BreakDown == \ + &AArch64GenRegisterBankInfo::PartMappings[PartialMapSrcIdx] && \ + Map[1].NumBreakDowns == 1 && "FPR" #SrcSize \ + " Src is incorrectly initialized"); \ + \ + } while (false) + + CHECK_VALUEMAP_FPEXT(32, 16); + CHECK_VALUEMAP_FPEXT(64, 16); + CHECK_VALUEMAP_FPEXT(64, 32); + CHECK_VALUEMAP_FPEXT(128, 64); + assert(verify(TRI) && "Invalid register bank information"); } @@ -455,6 +479,14 @@ AArch64RegisterBankInfo::getInstrMapping(const MachineInstr &MI) const { case TargetOpcode::G_FMUL: case TargetOpcode::G_FDIV: return getSameKindOfOperandsMapping(MI); + case TargetOpcode::G_FPEXT: { + LLT DstTy = MRI.getType(MI.getOperand(0).getReg()); + LLT SrcTy = MRI.getType(MI.getOperand(1).getReg()); + return getInstructionMapping( + DefaultMappingID, /*Cost*/ 1, + getFPExtMapping(DstTy.getSizeInBits(), SrcTy.getSizeInBits()), + /*NumOperands*/ 2); + } case TargetOpcode::COPY: { unsigned DstReg = MI.getOperand(0).getReg(); unsigned SrcReg = MI.getOperand(1).getReg(); diff --git a/llvm/lib/Target/AArch64/AArch64RegisterBankInfo.h b/llvm/lib/Target/AArch64/AArch64RegisterBankInfo.h index 384b9772927..008221dbef5 100644 --- a/llvm/lib/Target/AArch64/AArch64RegisterBankInfo.h +++ b/llvm/lib/Target/AArch64/AArch64RegisterBankInfo.h @@ -53,7 +53,11 @@ protected: DistanceBetweenRegBanks = 3, FirstCrossRegCpyIdx = 25, LastCrossRegCpyIdx = 39, - DistanceBetweenCrossRegCpy = 2 + DistanceBetweenCrossRegCpy = 2, + FPExt16To32Idx = 41, + FPExt16To64Idx = 43, + FPExt32To64Idx = 45, + FPExt64To128Idx = 47, }; static bool checkPartialMap(unsigned Idx, unsigned ValStartIdx, @@ -82,6 +86,15 @@ protected: static const RegisterBankInfo::ValueMapping * getCopyMapping(unsigned DstBankID, unsigned SrcBankID, unsigned Size); + /// Get the instruction mapping for G_FPEXT. + /// + /// \pre (DstSize, SrcSize) pair is one of the following: + /// (32, 16), (64, 16), (64, 32), (128, 64) + /// + /// \return An InstructionMapping with statically allocated OperandsMapping. + static const RegisterBankInfo::ValueMapping * + getFPExtMapping(unsigned DstSize, unsigned SrcSize); + #define GET_TARGET_REGBANK_CLASS #include "AArch64GenRegisterBank.inc" }; diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/arm64-regbankselect.mir b/llvm/test/CodeGen/AArch64/GlobalISel/arm64-regbankselect.mir index 4042047dfc2..cc158a29c3e 100644 --- a/llvm/test/CodeGen/AArch64/GlobalISel/arm64-regbankselect.mir +++ b/llvm/test/CodeGen/AArch64/GlobalISel/arm64-regbankselect.mir @@ -92,6 +92,10 @@ store double %vres, double* %addr ret void } + + define void @fp16Ext32() { ret void } + define void @fp16Ext64() { ret void } + define void @fp32Ext64() { ret void } ... --- @@ -742,3 +746,103 @@ body: | RET_ReallyLR ... + +--- +# Make sure we map FPEXT on FPR register bank. +# CHECK-LABEL: name: fp16Ext32 +name: fp16Ext32 +alignment: 2 +legalized: true +# CHECK: registers: +# CHECK-NEXT: - { id: 0, class: gpr, preferred-register: '' } +# CHECK-NEXT: - { id: 1, class: gpr, preferred-register: '' } +# CHECK-NEXT: - { id: 2, class: fpr, preferred-register: '' } +# CHECK-NEXT: - { id: 3, class: fpr, preferred-register: '' } +registers: + - { id: 0, class: _ } + - { id: 1, class: _ } + - { id: 2, class: _ } +# CHECK: %1:gpr(s32) = COPY %w0 +# CHECK-NEXT: %0:gpr(s16) = G_TRUNC %1 +# %0 has been mapped to GPR, we need to repair to match FPR. +# CHECK-NEXT: %3:fpr(s16) = COPY %0 +# CHECK-NEXT: %2:fpr(s32) = G_FPEXT %3 +# CHECK-NEXT: %s0 = COPY %2 +# CHECK-NEXT: RET_ReallyLR + +body: | + bb.1: + liveins: %w0 + + %1(s32) = COPY %w0 + %0(s16) = G_TRUNC %1(s32) + %2(s32) = G_FPEXT %0(s16) + %s0 = COPY %2(s32) + RET_ReallyLR implicit %s0 + +... + +--- +# Make sure we map FPEXT on FPR register bank. +# CHECK-LABEL: name: fp16Ext64 +name: fp16Ext64 +alignment: 2 +legalized: true +# CHECK: registers: +# CHECK-NEXT: - { id: 0, class: gpr, preferred-register: '' } +# CHECK-NEXT: - { id: 1, class: gpr, preferred-register: '' } +# CHECK-NEXT: - { id: 2, class: fpr, preferred-register: '' } +# CHECK-NEXT: - { id: 3, class: fpr, preferred-register: '' } +registers: + - { id: 0, class: _ } + - { id: 1, class: _ } + - { id: 2, class: _ } +# CHECK: %1:gpr(s32) = COPY %w0 +# CHECK-NEXT: %0:gpr(s16) = G_TRUNC %1 +# %0 has been mapped to GPR, we need to repair to match FPR. +# CHECK-NEXT: %3:fpr(s16) = COPY %0 +# CHECK-NEXT: %2:fpr(s64) = G_FPEXT %3 +# CHECK-NEXT: %d0 = COPY %2 +# CHECK-NEXT: RET_ReallyLR + +body: | + bb.1: + liveins: %w0 + + %1(s32) = COPY %w0 + %0(s16) = G_TRUNC %1(s32) + %2(s64) = G_FPEXT %0(s16) + %d0 = COPY %2(s64) + RET_ReallyLR implicit %d0 + +... + +--- +# Make sure we map FPEXT on FPR register bank. +# CHECK-LABEL: name: fp32Ext64 +name: fp32Ext64 +alignment: 2 +legalized: true +# CHECK: registers: +# CHECK-NEXT: - { id: 0, class: gpr, preferred-register: '' } +# CHECK-NEXT: - { id: 1, class: fpr, preferred-register: '' } +# CHECK-NEXT: - { id: 2, class: fpr, preferred-register: '' } +registers: + - { id: 0, class: _ } + - { id: 1, class: _ } +# CHECK: %0:gpr(s32) = COPY %w0 +# %0 has been mapped to GPR, we need to repair to match FPR. +# CHECK-NEXT: %2:fpr(s32) = COPY %0 +# CHECK-NEXT: %1:fpr(s64) = G_FPEXT %2 +# CHECK-NEXT: %d0 = COPY %1 +# CHECK-NEXT: RET_ReallyLR +body: | + bb.1: + liveins: %w0 + + %0(s32) = COPY %w0 + %1(s64) = G_FPEXT %0(s32) + %d0 = COPY %1(s64) + RET_ReallyLR implicit %d0 + +... |
