maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[Target][CodeGen] Remove default CostKind arguments on inner/impl TTI overrides 

Based off a discussion on D110100, we should be avoiding default CostKinds whenever possible.

This initial patch removes them from the 'inner' target implementation callbacks - these should only be used by the main TTI calls, so this should guarantee that we don't cause changes in CostKind by missing it in an inner call. This exposed a few missing arguments in getGEPCost and reduction cost calls that I've cleaned up.

Differential Revision: https://reviews.llvm.org/D110242
This commit is contained in:
Simon Pilgrim 2021-09-22 14:40:29 +01:00
parent e5aaf03326
commit b1f38a27f0
14 changed files with 65 additions and 87 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/include/llvm/Analysis/TargetTransformInfo.h
View File

@ -1743,8 +1743,8 @@ public:
InstructionCost
getGEPCost(Type *PointeeType, const Value *Ptr,
ArrayRef<const Value *> Operands,
enum TargetTransformInfo::TargetCostKind CostKind) override {
return Impl.getGEPCost(PointeeType, Ptr, Operands);
TargetTransformInfo::TargetCostKind CostKind) override {
return Impl.getGEPCost(PointeeType, Ptr, Operands, CostKind);
}
unsigned getInliningThresholdMultiplier() override {
return Impl.getInliningThresholdMultiplier();

25
llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
View File

@ -47,10 +47,9 @@ public:
const DataLayout &getDataLayout() const { return DL; }
InstructionCost
getGEPCost(Type *PointeeType, const Value *Ptr,
ArrayRef<const Value *> Operands,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency) const {
InstructionCost getGEPCost(Type *PointeeType, const Value *Ptr,
ArrayRef<const Value *> Operands,
TTI::TargetCostKind CostKind) const {
// In the basic model, we just assume that all-constant GEPs will be folded
// into their uses via addressing modes.
for (unsigned Idx = 0, Size = Operands.size(); Idx != Size; ++Idx)
@ -638,9 +637,10 @@ public:
return 1;
}
InstructionCost getExtendedAddReductionCost(
bool IsMLA, bool IsUnsigned, Type *ResTy, VectorType *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput) const {
InstructionCost
getExtendedAddReductionCost(bool IsMLA, bool IsUnsigned, Type *ResTy,
VectorType *Ty,
TTI::TargetCostKind CostKind) const {
return 1;
}
@ -862,10 +862,9 @@ protected:
public:
using BaseT::getGEPCost;
InstructionCost
getGEPCost(Type *PointeeType, const Value *Ptr,
ArrayRef<const Value *> Operands,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency) {
InstructionCost getGEPCost(Type *PointeeType, const Value *Ptr,
ArrayRef<const Value *> Operands,
TTI::TargetCostKind CostKind) {
assert(PointeeType && Ptr && "can't get GEPCost of nullptr");
assert(cast<PointerType>(Ptr->getType()->getScalarType())
->isOpaqueOrPointeeTypeMatches(PointeeType) &&
@ -970,10 +969,10 @@ public:
return TTI::TCC_Free;
break;
case Instruction::GetElementPtr: {
const GEPOperator *GEP = cast<GEPOperator>(U);
const auto *GEP = cast<GEPOperator>(U);
return TargetTTI->getGEPCost(GEP->getSourceElementType(),
GEP->getPointerOperand(),
Operands.drop_front());
Operands.drop_front(), CostKind);
}
case Instruction::Add:
case Instruction::FAdd:

17
llvm/include/llvm/CodeGen/BasicTTIImpl.h
View File

@ -364,8 +364,9 @@ public:
}
InstructionCost getGEPCost(Type *PointeeType, const Value *Ptr,
ArrayRef<const Value *> Operands) {
return BaseT::getGEPCost(PointeeType, Ptr, Operands);
ArrayRef<const Value *> Operands,
TTI::TargetCostKind CostKind) {
return BaseT::getGEPCost(PointeeType, Ptr, Operands, CostKind);
}
unsigned getEstimatedNumberOfCaseClusters(const SwitchInst &SI,
@ -748,8 +749,7 @@ public:
unsigned getMaxInterleaveFactor(unsigned VF) { return 1; }
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,
@ -1992,9 +1992,9 @@ public:
/// \param RetTy Return value types.
/// \param Tys Argument types.
/// \returns The cost of Call instruction.
InstructionCost
getCallInstrCost(Function *F, Type *RetTy, ArrayRef<Type *> Tys,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency) {
InstructionCost getCallInstrCost(Function *F, Type *RetTy,
ArrayRef<Type *> Tys,
TTI::TargetCostKind CostKind) {
return 10;
}
@ -2078,7 +2078,8 @@ public:
// By default reductions need one shuffle per reduction level.
ShuffleCost += NumReduxLevels * thisT()->getShuffleCost(
TTI::SK_PermuteSingleSrc, Ty, None, 0, Ty);
ArithCost += NumReduxLevels * thisT()->getArithmeticInstrCost(Opcode, Ty);
ArithCost +=
NumReduxLevels * thisT()->getArithmeticInstrCost(Opcode, Ty, CostKind);
return ShuffleCost + ArithCost +
thisT()->getVectorInstrCost(Instruction::ExtractElement, Ty, 0);
}

12
llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h
View File

@ -183,8 +183,7 @@ public:
InstructionCost getSpliceCost(VectorType *Tp, int Index);
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,
@ -303,8 +302,7 @@ public:
InstructionCost getInterleavedMemoryOpCost(
unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef<unsigned> Indices,
Align Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency,
Align Alignment, unsigned AddressSpace, TTI::TargetCostKind CostKind,
bool UseMaskForCond = false, bool UseMaskForGaps = false);
bool
@ -322,9 +320,9 @@ public:
bool isLegalToVectorizeReduction(const RecurrenceDescriptor &RdxDesc,
ElementCount VF) const;
InstructionCost getArithmeticReductionCost(
unsigned Opcode, VectorType *Ty, Optional<FastMathFlags> FMF,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput);
InstructionCost getArithmeticReductionCost(unsigned Opcode, VectorType *Ty,
Optional<FastMathFlags> FMF,
TTI::TargetCostKind CostKind);
InstructionCost getShuffleCost(TTI::ShuffleKind Kind, VectorType *Tp,
ArrayRef<int> Mask, int Index,

16
llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h
View File

@ -79,24 +79,21 @@ class GCNTTIImpl final : public BasicTTIImplBase<GCNTTIImpl> {
return TargetTransformInfo::TCC_Basic;
}
static inline int getHalfRateInstrCost(
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput) {
static inline int getHalfRateInstrCost(TTI::TargetCostKind CostKind) {
return CostKind == TTI::TCK_CodeSize ? 2
: 2 * TargetTransformInfo::TCC_Basic;
}
// TODO: The size is usually 8 bytes, but takes 4x as many cycles. Maybe
// should be 2 or 4.
static inline int getQuarterRateInstrCost(
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput) {
static inline int getQuarterRateInstrCost(TTI::TargetCostKind CostKind) {
return CostKind == TTI::TCK_CodeSize ? 2
: 4 * TargetTransformInfo::TCC_Basic;
}
// On some parts, normal fp64 operations are half rate, and others
// quarter. This also applies to some integer operations.
int get64BitInstrCost(
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput) const;
int get64BitInstrCost(TTI::TargetCostKind CostKind) const;
public:
explicit GCNTTIImpl(const AMDGPUTargetMachine *TM, const Function &F);
@ -152,8 +149,7 @@ public:
bool getTgtMemIntrinsic(IntrinsicInst *Inst, MemIntrinsicInfo &Info) const;
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,
@ -217,13 +213,13 @@ public:
InstructionCost getArithmeticReductionCost(
unsigned Opcode, VectorType *Ty, Optional<FastMathFlags> FMF,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput);
TTI::TargetCostKind CostKind);
InstructionCost getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
TTI::TargetCostKind CostKind);
InstructionCost getMinMaxReductionCost(
VectorType *Ty, VectorType *CondTy, bool IsUnsigned,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput);
TTI::TargetCostKind CostKind);
};
} // end namespace llvm

6
llvm/lib/Target/ARM/ARMTargetTransformInfo.h
View File

@ -231,8 +231,7 @@ public:
const SCEV *Ptr);
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Op1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Op2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,
@ -251,8 +250,7 @@ public:
InstructionCost getInterleavedMemoryOpCost(
unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef<unsigned> Indices,
Align Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency,
Align Alignment, unsigned AddressSpace, TTI::TargetCostKind CostKind,
bool UseMaskForCond = false, bool UseMaskForGaps = false);
InstructionCost getGatherScatterOpCost(unsigned Opcode, Type *DataTy,

11
llvm/lib/Target/BPF/BPFTargetTransformInfo.h
View File

@ -56,12 +56,11 @@ public:
}
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,
TTI::OperandValueProperties Opd2PropInfo = TTI::OP_None,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,
TTI::OperandValueProperties Opd2PropInfo = TTI::OP_None,
ArrayRef<const Value *> Args = ArrayRef<const Value *>(),
const Instruction *CxtI = nullptr) {
int ISD = TLI->InstructionOpcodeToISD(Opcode);

13
llvm/lib/Target/Hexagon/HexagonTargetTransformInfo.h
View File

@ -121,10 +121,9 @@ public:
MaybeAlign Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind,
const Instruction *I = nullptr);
InstructionCost
getMaskedMemoryOpCost(unsigned Opcode, Type *Src, Align Alignment,
unsigned AddressSpace,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency);
InstructionCost getMaskedMemoryOpCost(unsigned Opcode, Type *Src,
Align Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind);
InstructionCost getShuffleCost(TTI::ShuffleKind Kind, Type *Tp,
ArrayRef<int> Mask, int Index, Type *SubTp);
InstructionCost getGatherScatterOpCost(unsigned Opcode, Type *DataTy,
@ -134,16 +133,14 @@ public:
const Instruction *I);
InstructionCost getInterleavedMemoryOpCost(
unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef<unsigned> Indices,
Align Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency,
Align Alignment, unsigned AddressSpace, TTI::TargetCostKind CostKind,
bool UseMaskForCond = false, bool UseMaskForGaps = false);
InstructionCost getCmpSelInstrCost(unsigned Opcode, Type *ValTy, Type *CondTy,
CmpInst::Predicate VecPred,
TTI::TargetCostKind CostKind,
const Instruction *I = nullptr);
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,

3
llvm/lib/Target/Lanai/LanaiTargetTransformInfo.h
View File

@ -91,8 +91,7 @@ public:
}
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,

3
llvm/lib/Target/NVPTX/NVPTXTargetTransformInfo.h
View File

@ -94,8 +94,7 @@ public:
unsigned getInliningThresholdMultiplier() { return 5; }
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,

6
llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h
View File

@ -103,8 +103,7 @@ public:
InstructionCost vectorCostAdjustment(InstructionCost Cost, unsigned Opcode,
Type *Ty1, Type *Ty2);
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,
@ -131,8 +130,7 @@ public:
const Instruction *I = nullptr);
InstructionCost getInterleavedMemoryOpCost(
unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef<unsigned> Indices,
Align Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency,
Align Alignment, unsigned AddressSpace, TTI::TargetCostKind CostKind,
bool UseMaskForCond = false, bool UseMaskForGaps = false);
InstructionCost getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
TTI::TargetCostKind CostKind);

6
llvm/lib/Target/SystemZ/SystemZTargetTransformInfo.h
View File

@ -83,8 +83,7 @@ public:
bool enableInterleavedAccessVectorization() { return true; }
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,
@ -116,8 +115,7 @@ public:
InstructionCost getInterleavedMemoryOpCost(
unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef<unsigned> Indices,
Align Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency,
Align Alignment, unsigned AddressSpace, TTI::TargetCostKind CostKind,
bool UseMaskForCond = false, bool UseMaskForGaps = false);
InstructionCost getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,

3
llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.h
View File

@ -60,8 +60,7 @@ public:
unsigned getNumberOfRegisters(unsigned ClassID) const;
TypeSize getRegisterBitWidth(TargetTransformInfo::RegisterKind K) const;
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,

27
llvm/lib/Target/X86/X86TargetTransformInfo.h
View File

@ -122,8 +122,7 @@ public:
unsigned getLoadStoreVecRegBitWidth(unsigned AS) const;
unsigned getMaxInterleaveFactor(unsigned VF);
InstructionCost getArithmeticInstrCost(
unsigned Opcode, Type *Ty,
TTI::TargetCostKind CostKind = TTI::TCK_RecipThroughput,
unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
TTI::OperandValueKind Opd1Info = TTI::OK_AnyValue,
TTI::OperandValueKind Opd2Info = TTI::OK_AnyValue,
TTI::OperandValueProperties Opd1PropInfo = TTI::OP_None,
@ -150,10 +149,9 @@ public:
MaybeAlign Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind,
const Instruction *I = nullptr);
InstructionCost
getMaskedMemoryOpCost(unsigned Opcode, Type *Src, Align Alignment,
unsigned AddressSpace,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency);
InstructionCost getMaskedMemoryOpCost(unsigned Opcode, Type *Src,
Align Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind);
InstructionCost getGatherScatterOpCost(unsigned Opcode, Type *DataTy,
const Value *Ptr, bool VariableMask,
Align Alignment,
@ -182,9 +180,9 @@ public:
InstructionCost getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
TTI::TargetCostKind CostKind);
InstructionCost getArithmeticReductionCost(
unsigned Opcode, VectorType *Ty, Optional<FastMathFlags> FMF,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency);
InstructionCost getArithmeticReductionCost(unsigned Opcode, VectorType *Ty,
Optional<FastMathFlags> FMF,
TTI::TargetCostKind CostKind);
InstructionCost getMinMaxCost(Type *Ty, Type *CondTy, bool IsUnsigned);
@ -194,19 +192,18 @@ public:
InstructionCost getInterleavedMemoryOpCost(
unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef<unsigned> Indices,
Align Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency,
Align Alignment, unsigned AddressSpace, TTI::TargetCostKind CostKind,
bool UseMaskForCond = false, bool UseMaskForGaps = false);
InstructionCost getInterleavedMemoryOpCostAVX512(
unsigned Opcode, FixedVectorType *VecTy, unsigned Factor,
ArrayRef<unsigned> Indices, Align Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency,
bool UseMaskForCond = false, bool UseMaskForGaps = false);
TTI::TargetCostKind CostKind, bool UseMaskForCond = false,
bool UseMaskForGaps = false);
InstructionCost getInterleavedMemoryOpCostAVX2(
unsigned Opcode, FixedVectorType *VecTy, unsigned Factor,
ArrayRef<unsigned> Indices, Align Alignment, unsigned AddressSpace,
TTI::TargetCostKind CostKind = TTI::TCK_SizeAndLatency,
bool UseMaskForCond = false, bool UseMaskForGaps = false);
TTI::TargetCostKind CostKind, bool UseMaskForCond = false,
bool UseMaskForGaps = false);
InstructionCost getIntImmCost(int64_t);