LLVM  8.0.1
Public Member Functions | List of all members
llvm::GCNTTIImpl Class Referencefinal

#include "Target/AMDGPU/AMDGPUTargetTransformInfo.h"

Inheritance diagram for llvm::GCNTTIImpl:
Inheritance graph
[legend]
Collaboration diagram for llvm::GCNTTIImpl:
Collaboration graph
[legend]

Public Member Functions

 GCNTTIImpl (const AMDGPUTargetMachine *TM, const Function &F)
 
bool hasBranchDivergence ()
 
void getUnrollingPreferences (Loop *L, ScalarEvolution &SE, TTI::UnrollingPreferences &UP)
 
TTI::PopcntSupportKind getPopcntSupport (unsigned TyWidth)
 
unsigned getHardwareNumberOfRegisters (bool Vector) const
 
unsigned getNumberOfRegisters (bool Vector) const
 
unsigned getRegisterBitWidth (bool Vector) const
 
unsigned getMinVectorRegisterBitWidth () const
 
unsigned getLoadVectorFactor (unsigned VF, unsigned LoadSize, unsigned ChainSizeInBytes, VectorType *VecTy) const
 
unsigned getStoreVectorFactor (unsigned VF, unsigned StoreSize, unsigned ChainSizeInBytes, VectorType *VecTy) const
 
unsigned getLoadStoreVecRegBitWidth (unsigned AddrSpace) const
 
bool isLegalToVectorizeMemChain (unsigned ChainSizeInBytes, unsigned Alignment, unsigned AddrSpace) const
 
bool isLegalToVectorizeLoadChain (unsigned ChainSizeInBytes, unsigned Alignment, unsigned AddrSpace) const
 
bool isLegalToVectorizeStoreChain (unsigned ChainSizeInBytes, unsigned Alignment, unsigned AddrSpace) const
 
unsigned getMaxInterleaveFactor (unsigned VF)
 
bool getTgtMemIntrinsic (IntrinsicInst *Inst, MemIntrinsicInfo &Info) const
 
int getArithmeticInstrCost (unsigned Opcode, Type *Ty, 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 *>())
 
unsigned getCFInstrCost (unsigned Opcode)
 
int getVectorInstrCost (unsigned Opcode, Type *ValTy, unsigned Index)
 
bool isSourceOfDivergence (const Value *V) const
 
bool isAlwaysUniform (const Value *V) const
 
unsigned getFlatAddressSpace () const
 
unsigned getVectorSplitCost ()
 
unsigned getShuffleCost (TTI::ShuffleKind Kind, Type *Tp, int Index, Type *SubTp)
 
bool areInlineCompatible (const Function *Caller, const Function *Callee) const
 
unsigned getInliningThresholdMultiplier ()
 
int getArithmeticReductionCost (unsigned Opcode, Type *Ty, bool IsPairwise)
 
int getMinMaxReductionCost (Type *Ty, Type *CondTy, bool IsPairwiseForm, bool IsUnsigned)
 
- Public Member Functions inherited from llvm::BasicTTIImplBase< GCNTTIImpl >
bool allowsMisalignedMemoryAccesses (LLVMContext &Context, unsigned BitWidth, unsigned AddressSpace, unsigned Alignment, bool *Fast) const
 
bool hasBranchDivergence ()
 
bool isSourceOfDivergence (const Value *V)
 
bool isAlwaysUniform (const Value *V)
 
unsigned getFlatAddressSpace ()
 
bool isLegalAddImmediate (int64_t imm)
 
bool isLegalICmpImmediate (int64_t imm)
 
bool isLegalAddressingMode (Type *Ty, GlobalValue *BaseGV, int64_t BaseOffset, bool HasBaseReg, int64_t Scale, unsigned AddrSpace, Instruction *I=nullptr)
 
bool isIndexedLoadLegal (TTI::MemIndexedMode M, Type *Ty, const DataLayout &DL) const
 
bool isIndexedStoreLegal (TTI::MemIndexedMode M, Type *Ty, const DataLayout &DL) const
 
bool isLSRCostLess (TTI::LSRCost C1, TTI::LSRCost C2)
 
int getScalingFactorCost (Type *Ty, GlobalValue *BaseGV, int64_t BaseOffset, bool HasBaseReg, int64_t Scale, unsigned AddrSpace)
 
bool isTruncateFree (Type *Ty1, Type *Ty2)
 
bool isProfitableToHoist (Instruction *I)
 
bool useAA () const
 
bool isTypeLegal (Type *Ty)
 
int getGEPCost (Type *PointeeType, const Value *Ptr, ArrayRef< const Value * > Operands)
 
int getExtCost (const Instruction *I, const Value *Src)
 
unsigned getIntrinsicCost (Intrinsic::ID IID, Type *RetTy, ArrayRef< const Value * > Arguments)
 
unsigned getIntrinsicCost (Intrinsic::ID IID, Type *RetTy, ArrayRef< Type * > ParamTys)
 
unsigned getEstimatedNumberOfCaseClusters (const SwitchInst &SI, unsigned &JumpTableSize)
 
unsigned getJumpBufAlignment ()
 
unsigned getJumpBufSize ()
 
bool shouldBuildLookupTables ()
 
bool haveFastSqrt (Type *Ty)
 
bool isFCmpOrdCheaperThanFCmpZero (Type *Ty)
 
unsigned getFPOpCost (Type *Ty)
 
unsigned getOperationCost (unsigned Opcode, Type *Ty, Type *OpTy)
 
unsigned getInliningThresholdMultiplier ()
 
void getUnrollingPreferences (Loop *L, ScalarEvolution &SE, TTI::UnrollingPreferences &UP)
 
int getInstructionLatency (const Instruction *I)
 
unsigned getNumberOfRegisters (bool Vector)
 
unsigned getRegisterBitWidth (bool Vector) const
 
unsigned getScalarizationOverhead (Type *Ty, bool Insert, bool Extract)
 Estimate the overhead of scalarizing an instruction. More...
 
unsigned getScalarizationOverhead (Type *VecTy, ArrayRef< const Value * > Args)
 
unsigned getOperandsScalarizationOverhead (ArrayRef< const Value * > Args, unsigned VF)
 Estimate the overhead of scalarizing an instructions unique non-constant operands. More...
 
unsigned getMaxInterleaveFactor (unsigned VF)
 
unsigned getArithmeticInstrCost (unsigned Opcode, Type *Ty, 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 * >())
 
unsigned getShuffleCost (TTI::ShuffleKind Kind, Type *Tp, int Index, Type *SubTp)
 
unsigned getCastInstrCost (unsigned Opcode, Type *Dst, Type *Src, const Instruction *I=nullptr)
 
unsigned getExtractWithExtendCost (unsigned Opcode, Type *Dst, VectorType *VecTy, unsigned Index)
 
unsigned getCFInstrCost (unsigned Opcode)
 
unsigned getCmpSelInstrCost (unsigned Opcode, Type *ValTy, Type *CondTy, const Instruction *I)
 
unsigned getVectorInstrCost (unsigned Opcode, Type *Val, unsigned Index)
 
unsigned getMemoryOpCost (unsigned Opcode, Type *Src, unsigned Alignment, unsigned AddressSpace, const Instruction *I=nullptr)
 
unsigned getInterleavedMemoryOpCost (unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef< unsigned > Indices, unsigned Alignment, unsigned AddressSpace, bool UseMaskForCond=false, bool UseMaskForGaps=false)
 
unsigned getIntrinsicInstrCost (Intrinsic::ID IID, Type *RetTy, ArrayRef< Value * > Args, FastMathFlags FMF, unsigned VF=1)
 Get intrinsic cost based on arguments. More...
 
unsigned getIntrinsicInstrCost (Intrinsic::ID IID, Type *RetTy, ArrayRef< Type * > Tys, FastMathFlags FMF, unsigned ScalarizationCostPassed=std::numeric_limits< unsigned >::max())
 Get intrinsic cost based on argument types. More...
 
unsigned getCallInstrCost (Function *F, Type *RetTy, ArrayRef< Type * > Tys)
 Compute a cost of the given call instruction. More...
 
unsigned getNumberOfParts (Type *Tp)
 
unsigned getAddressComputationCost (Type *Ty, ScalarEvolution *, const SCEV *)
 
unsigned getArithmeticReductionCost (unsigned Opcode, Type *Ty, bool IsPairwise)
 Try to calculate arithmetic and shuffle op costs for reduction operations. More...
 
unsigned getMinMaxReductionCost (Type *Ty, Type *CondTy, bool IsPairwise, bool)
 Try to calculate op costs for min/max reduction operations. More...
 
unsigned getVectorSplitCost ()
 
- Public Member Functions inherited from llvm::TargetTransformInfoImplCRTPBase< GCNTTIImpl >
unsigned getCallCost (const Function *F, int NumArgs)
 
unsigned getCallCost (const Function *F, ArrayRef< const Value * > Arguments)
 
int getGEPCost (Type *PointeeType, const Value *Ptr, ArrayRef< const Value * > Operands)
 
unsigned getIntrinsicCost (Intrinsic::ID IID, Type *RetTy, ArrayRef< const Value * > Arguments)
 
unsigned getUserCost (const User *U, ArrayRef< const Value * > Operands)
 
int getInstructionLatency (const Instruction *I)
 
- Public Member Functions inherited from llvm::TargetTransformInfoImplBase
 TargetTransformInfoImplBase (const TargetTransformInfoImplBase &Arg)
 
 TargetTransformInfoImplBase (TargetTransformInfoImplBase &&Arg)
 
const DataLayout & getDataLayout () const
 
unsigned getOperationCost (unsigned Opcode, Type *Ty, Type *OpTy)
 
int getGEPCost (Type *PointeeType, const Value *Ptr, ArrayRef< const Value *> Operands)
 
unsigned getEstimatedNumberOfCaseClusters (const SwitchInst &SI, unsigned &JTSize)
 
int getExtCost (const Instruction *I, const Value *Src)
 
unsigned getCallCost (FunctionType *FTy, int NumArgs)
 
unsigned getInliningThresholdMultiplier ()
 
unsigned getIntrinsicCost (Intrinsic::ID IID, Type *RetTy, ArrayRef< Type *> ParamTys)
 
bool hasBranchDivergence ()
 
bool isSourceOfDivergence (const Value *V)
 
bool isAlwaysUniform (const Value *V)
 
unsigned getFlatAddressSpace ()
 
bool isLoweredToCall (const Function *F)
 
void getUnrollingPreferences (Loop *, ScalarEvolution &, TTI::UnrollingPreferences &)
 
bool isLegalAddImmediate (int64_t Imm)
 
bool isLegalICmpImmediate (int64_t Imm)
 
bool isLegalAddressingMode (Type *Ty, GlobalValue *BaseGV, int64_t BaseOffset, bool HasBaseReg, int64_t Scale, unsigned AddrSpace, Instruction *I=nullptr)
 
bool isLSRCostLess (TTI::LSRCost &C1, TTI::LSRCost &C2)
 
bool canMacroFuseCmp ()
 
bool shouldFavorPostInc () const
 
bool isLegalMaskedStore (Type *DataType)
 
bool isLegalMaskedLoad (Type *DataType)
 
bool isLegalMaskedScatter (Type *DataType)
 
bool isLegalMaskedGather (Type *DataType)
 
bool hasDivRemOp (Type *DataType, bool IsSigned)
 
bool hasVolatileVariant (Instruction *I, unsigned AddrSpace)
 
bool prefersVectorizedAddressing ()
 
int getScalingFactorCost (Type *Ty, GlobalValue *BaseGV, int64_t BaseOffset, bool HasBaseReg, int64_t Scale, unsigned AddrSpace)
 
bool LSRWithInstrQueries ()
 
bool isTruncateFree (Type *Ty1, Type *Ty2)
 
bool isProfitableToHoist (Instruction *I)
 
bool useAA ()
 
bool isTypeLegal (Type *Ty)
 
unsigned getJumpBufAlignment ()
 
unsigned getJumpBufSize ()
 
bool shouldBuildLookupTables ()
 
bool shouldBuildLookupTablesForConstant (Constant *C)
 
bool useColdCCForColdCall (Function &F)
 
unsigned getScalarizationOverhead (Type *Ty, bool Insert, bool Extract)
 
unsigned getOperandsScalarizationOverhead (ArrayRef< const Value *> Args, unsigned VF)
 
bool supportsEfficientVectorElementLoadStore ()
 
bool enableAggressiveInterleaving (bool LoopHasReductions)
 
const TTI::MemCmpExpansionOptions * enableMemCmpExpansion (bool IsZeroCmp) const
 
bool enableInterleavedAccessVectorization ()
 
bool enableMaskedInterleavedAccessVectorization ()
 
bool isFPVectorizationPotentiallyUnsafe ()
 
bool allowsMisalignedMemoryAccesses (LLVMContext &Context, unsigned BitWidth, unsigned AddressSpace, unsigned Alignment, bool *Fast)
 
TTI::PopcntSupportKind getPopcntSupport (unsigned IntTyWidthInBit)
 
bool haveFastSqrt (Type *Ty)
 
bool isFCmpOrdCheaperThanFCmpZero (Type *Ty)
 
unsigned getFPOpCost (Type *Ty)
 
int getIntImmCodeSizeCost (unsigned Opcode, unsigned Idx, const APInt &Imm, Type *Ty)
 
unsigned getIntImmCost (const APInt &Imm, Type *Ty)
 
unsigned getIntImmCost (unsigned Opcode, unsigned Idx, const APInt &Imm, Type *Ty)
 
unsigned getIntImmCost (Intrinsic::ID IID, unsigned Idx, const APInt &Imm, Type *Ty)
 
unsigned getNumberOfRegisters (bool Vector)
 
unsigned getRegisterBitWidth (bool Vector) const
 
unsigned getMinVectorRegisterBitWidth ()
 
bool shouldMaximizeVectorBandwidth (bool OptSize) const
 
unsigned getMinimumVF (unsigned ElemWidth) const
 
bool shouldConsiderAddressTypePromotion (const Instruction &I, bool &AllowPromotionWithoutCommonHeader)
 
unsigned getCacheLineSize ()
 
llvm::Optional< unsigned > getCacheSize (TargetTransformInfo::CacheLevel Level)
 
llvm::Optional< unsigned > getCacheAssociativity (TargetTransformInfo::CacheLevel Level)
 
unsigned getPrefetchDistance ()
 
unsigned getMinPrefetchStride ()
 
unsigned getMaxPrefetchIterationsAhead ()
 
unsigned getMaxInterleaveFactor (unsigned VF)
 
unsigned getArithmeticInstrCost (unsigned Opcode, Type *Ty, TTI::OperandValueKind Opd1Info, TTI::OperandValueKind Opd2Info, TTI::OperandValueProperties Opd1PropInfo, TTI::OperandValueProperties Opd2PropInfo, ArrayRef< const Value *> Args)
 
unsigned getShuffleCost (TTI::ShuffleKind Kind, Type *Ty, int Index, Type *SubTp)
 
unsigned getCastInstrCost (unsigned Opcode, Type *Dst, Type *Src, const Instruction *I)
 
unsigned getExtractWithExtendCost (unsigned Opcode, Type *Dst, VectorType *VecTy, unsigned Index)
 
unsigned getCFInstrCost (unsigned Opcode)
 
unsigned getCmpSelInstrCost (unsigned Opcode, Type *ValTy, Type *CondTy, const Instruction *I)
 
unsigned getVectorInstrCost (unsigned Opcode, Type *Val, unsigned Index)
 
unsigned getMemoryOpCost (unsigned Opcode, Type *Src, unsigned Alignment, unsigned AddressSpace, const Instruction *I)
 
unsigned getMaskedMemoryOpCost (unsigned Opcode, Type *Src, unsigned Alignment, unsigned AddressSpace)
 
unsigned getGatherScatterOpCost (unsigned Opcode, Type *DataTy, Value *Ptr, bool VariableMask, unsigned Alignment)
 
unsigned getInterleavedMemoryOpCost (unsigned Opcode, Type *VecTy, unsigned Factor, ArrayRef< unsigned > Indices, unsigned Alignment, unsigned AddressSpace, bool UseMaskForCond=false, bool UseMaskForGaps=false)
 
unsigned getIntrinsicInstrCost (Intrinsic::ID ID, Type *RetTy, ArrayRef< Type *> Tys, FastMathFlags FMF, unsigned ScalarizationCostPassed)
 
unsigned getIntrinsicInstrCost (Intrinsic::ID ID, Type *RetTy, ArrayRef< Value *> Args, FastMathFlags FMF, unsigned VF)
 
unsigned getCallInstrCost (Function *F, Type *RetTy, ArrayRef< Type *> Tys)
 
unsigned getNumberOfParts (Type *Tp)
 
unsigned getAddressComputationCost (Type *Tp, ScalarEvolution *, const SCEV *)
 
unsigned getArithmeticReductionCost (unsigned, Type *, bool)
 
unsigned getMinMaxReductionCost (Type *, Type *, bool, bool)
 
unsigned getCostOfKeepingLiveOverCall (ArrayRef< Type *> Tys)
 
bool getTgtMemIntrinsic (IntrinsicInst *Inst, MemIntrinsicInfo &Info)
 
unsigned getAtomicMemIntrinsicMaxElementSize () const
 
Value * getOrCreateResultFromMemIntrinsic (IntrinsicInst *Inst, Type *ExpectedType)
 
Type * getMemcpyLoopLoweringType (LLVMContext &Context, Value *Length, unsigned SrcAlign, unsigned DestAlign) const
 
void getMemcpyLoopResidualLoweringType (SmallVectorImpl< Type *> &OpsOut, LLVMContext &Context, unsigned RemainingBytes, unsigned SrcAlign, unsigned DestAlign) const
 
bool areInlineCompatible (const Function *Caller, const Function *Callee) const
 
bool areFunctionArgsABICompatible (const Function *Caller, const Function *Callee, SmallPtrSetImpl< Argument *> &Args) const
 
bool isIndexedLoadLegal (TTI::MemIndexedMode Mode, Type *Ty, const DataLayout &DL) const
 
bool isIndexedStoreLegal (TTI::MemIndexedMode Mode, Type *Ty, const DataLayout &DL) const
 
unsigned getLoadStoreVecRegBitWidth (unsigned AddrSpace) const
 
bool isLegalToVectorizeLoad (LoadInst *LI) const
 
bool isLegalToVectorizeStore (StoreInst *SI) const
 
bool isLegalToVectorizeLoadChain (unsigned ChainSizeInBytes, unsigned Alignment, unsigned AddrSpace) const
 
bool isLegalToVectorizeStoreChain (unsigned ChainSizeInBytes, unsigned Alignment, unsigned AddrSpace) const
 
unsigned getLoadVectorFactor (unsigned VF, unsigned LoadSize, unsigned ChainSizeInBytes, VectorType *VecTy) const
 
unsigned getStoreVectorFactor (unsigned VF, unsigned StoreSize, unsigned ChainSizeInBytes, VectorType *VecTy) const
 
bool useReductionIntrinsic (unsigned Opcode, Type *Ty, TTI::ReductionFlags Flags) const
 
bool shouldExpandReduction (const IntrinsicInst *II) const
 

Additional Inherited Members

- Protected Types inherited from llvm::TargetTransformInfoImplBase
typedef TargetTransformInfo TTI
 
- Protected Member Functions inherited from llvm::BasicTTIImplBase< GCNTTIImpl >
 BasicTTIImplBase (const TargetMachine *TM, const DataLayout &DL)
 
- Protected Member Functions inherited from llvm::TargetTransformInfoImplCRTPBase< GCNTTIImpl >
 TargetTransformInfoImplCRTPBase (const DataLayout &DL)
 
- Protected Member Functions inherited from llvm::TargetTransformInfoImplBase
 TargetTransformInfoImplBase (const DataLayout &DL)
 
unsigned minRequiredElementSize (const Value *Val, bool &isSigned)
 
bool isStridedAccess (const SCEV *Ptr)
 
const SCEVConstant * getConstantStrideStep (ScalarEvolution *SE, const SCEV *Ptr)
 
bool isConstantStridedAccessLessThan (ScalarEvolution *SE, const SCEV *Ptr, int64_t MergeDistance)
 
- Protected Attributes inherited from llvm::TargetTransformInfoImplBase
const DataLayout & DL
 

Detailed Description

Definition at line 59 of file AMDGPUTargetTransformInfo.h.

Constructor & Destructor Documentation

◆ GCNTTIImpl()

llvm::GCNTTIImpl::GCNTTIImpl ( const AMDGPUTargetMachine *  TM,
const Function &  F 
)
inlineexplicit

Definition at line 119 of file AMDGPUTargetTransformInfo.h.

Member Function Documentation

◆ areInlineCompatible()

bool GCNTTIImpl::areInlineCompatible ( const Function *  Caller,
const Function *  Callee 
) const

Definition at line 611 of file AMDGPUTargetTransformInfo.cpp.

References llvm::MCSubtargetInfo::getFeatureBits(), llvm::TargetMachine::getSubtargetImpl(), and llvm::SystemZISD::TM.

◆ getArithmeticInstrCost()

int GCNTTIImpl::getArithmeticInstrCost ( unsigned  Opcode,
Type *  Ty,
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 *>() 
)

Definition at line 337 of file AMDGPUTargetTransformInfo.cpp.

References llvm::ISD::ADD, llvm::ISD::AND, llvm::TargetTransformInfoImplBase::DL, llvm::ArrayRef< T >::empty(), llvm::MVT::f16, llvm::MVT::f32, llvm::MVT::f64, llvm::ISD::FADD, llvm::ISD::FDIV, llvm::ISD::FMUL, llvm::ISD::FREM, llvm::ISD::FSUB, llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getArithmeticInstrCost(), llvm::MVT::i64, llvm::EVT::isSimple(), llvm::AArch64CC::LT, llvm::PatternMatch::m_FPOne(), llvm::PatternMatch::match(), llvm::ISD::MUL, llvm::ISD::OR, llvm::ISD::SHL, llvm::AMDGPUSubtarget::SOUTHERN_ISLANDS, llvm::ISD::SRA, llvm::ISD::SRL, llvm::ARM_MB::ST, llvm::ISD::SUB, and llvm::ISD::XOR.

◆ getArithmeticReductionCost()

int GCNTTIImpl::getArithmeticReductionCost ( unsigned  Opcode,
Type *  Ty,
bool  IsPairwise 
)

Definition at line 458 of file AMDGPUTargetTransformInfo.cpp.

References llvm::TargetTransformInfoImplBase::DL, llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getArithmeticReductionCost(), llvm::EVT::getScalarSizeInBits(), llvm::AArch64CC::LT, and llvm::ARM_MB::ST.

◆ getCFInstrCost()

unsigned GCNTTIImpl::getCFInstrCost ( unsigned  Opcode)

Definition at line 447 of file AMDGPUTargetTransformInfo.cpp.

References llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getCFInstrCost(), and llvm::MipsISD::Ret.

◆ getFlatAddressSpace()

unsigned llvm::GCNTTIImpl::getFlatAddressSpace ( ) const
inline

Definition at line 176 of file AMDGPUTargetTransformInfo.h.

References AMDGPUAS::FLAT_ADDRESS, llvm::GCNSubtarget::hasFlatAddressSpace(), and AMDGPUAS::UNKNOWN_ADDRESS_SPACE.

◆ getHardwareNumberOfRegisters()

unsigned GCNTTIImpl::getHardwareNumberOfRegisters ( bool  Vector) const

Definition at line 213 of file AMDGPUTargetTransformInfo.cpp.

◆ getInliningThresholdMultiplier()

unsigned llvm::GCNTTIImpl::getInliningThresholdMultiplier ( )
inline

Definition at line 193 of file AMDGPUTargetTransformInfo.h.

References llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getArithmeticReductionCost(), and llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getMinMaxReductionCost().

◆ getLoadStoreVecRegBitWidth()

unsigned GCNTTIImpl::getLoadStoreVecRegBitWidth ( unsigned  AddrSpace) const

Definition at line 254 of file AMDGPUTargetTransformInfo.cpp.

References AMDGPUAS::CONSTANT_ADDRESS, AMDGPUAS::CONSTANT_ADDRESS_32BIT, AMDGPUAS::FLAT_ADDRESS, AMDGPUAS::GLOBAL_ADDRESS, llvm_unreachable, AMDGPUAS::LOCAL_ADDRESS, AMDGPUAS::PRIVATE_ADDRESS, AMDGPUAS::REGION_ADDRESS, and llvm::ARM_MB::ST.

◆ getLoadVectorFactor()

unsigned GCNTTIImpl::getLoadVectorFactor ( unsigned  VF,
unsigned  LoadSize,
unsigned  ChainSizeInBytes,
VectorType *  VecTy 
) const

Definition at line 233 of file AMDGPUTargetTransformInfo.cpp.

References llvm::Type::getScalarSizeInBits().

◆ getMaxInterleaveFactor()

unsigned GCNTTIImpl::getMaxInterleaveFactor ( unsigned  VF)

Definition at line 297 of file AMDGPUTargetTransformInfo.cpp.

◆ getMinMaxReductionCost()

int GCNTTIImpl::getMinMaxReductionCost ( Type *  Ty,
Type *  CondTy,
bool  IsPairwiseForm,
bool  IsUnsigned 
)

Definition at line 473 of file AMDGPUTargetTransformInfo.cpp.

References llvm::TargetTransformInfoImplBase::DL, llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getMinMaxReductionCost(), llvm::EVT::getScalarSizeInBits(), llvm::AArch64CC::LT, and llvm::ARM_MB::ST.

◆ getMinVectorRegisterBitWidth()

unsigned GCNTTIImpl::getMinVectorRegisterBitWidth ( ) const

Definition at line 229 of file AMDGPUTargetTransformInfo.cpp.

◆ getNumberOfRegisters()

unsigned GCNTTIImpl::getNumberOfRegisters ( bool  Vector) const

Definition at line 219 of file AMDGPUTargetTransformInfo.cpp.

◆ getPopcntSupport()

TTI::PopcntSupportKind llvm::GCNTTIImpl::getPopcntSupport ( unsigned  TyWidth)
inline

Definition at line 131 of file AMDGPUTargetTransformInfo.h.

References llvm::AMDGPU::HSAMD::Kernel::Key::Args, assert(), llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getArithmeticInstrCost(), llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getCFInstrCost(), llvm::TargetTransformInfoImplBase::getLoadStoreVecRegBitWidth(), llvm::TargetTransformInfoImplBase::getLoadVectorFactor(), llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getMaxInterleaveFactor(), llvm::TargetTransformInfoImplBase::getMinVectorRegisterBitWidth(), llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getNumberOfRegisters(), llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getRegisterBitWidth(), llvm::TargetTransformInfoImplBase::getStoreVectorFactor(), llvm::TargetTransformInfoImplBase::getTgtMemIntrinsic(), llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getVectorInstrCost(), Info, llvm::BasicTTIImplBase< AMDGPUTTIImpl >::isAlwaysUniform(), llvm::TargetTransformInfoImplBase::isLegalToVectorizeLoadChain(), llvm::TargetTransformInfoImplBase::isLegalToVectorizeStoreChain(), llvm::isPowerOf2_32(), llvm::BasicTTIImplBase< AMDGPUTTIImpl >::isSourceOfDivergence(), llvm::TargetTransformInfo::OK_AnyValue, llvm::TargetTransformInfo::OP_None, and llvm::TargetTransformInfo::PSK_FastHardware.

◆ getRegisterBitWidth()

unsigned GCNTTIImpl::getRegisterBitWidth ( bool  Vector) const

Definition at line 225 of file AMDGPUTargetTransformInfo.cpp.

◆ getShuffleCost()

unsigned GCNTTIImpl::getShuffleCost ( TTI::ShuffleKind  Kind,
Type *  Tp,
int  Index,
Type *  SubTp 
)

Definition at line 588 of file AMDGPUTargetTransformInfo.cpp.

References llvm::TargetTransformInfoImplBase::DL, llvm::SequentialType::getElementType(), llvm::SequentialType::getNumElements(), llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getShuffleCost(), llvm::DataLayout::getTypeSizeInBits(), llvm::TargetTransformInfo::SK_Broadcast, llvm::TargetTransformInfo::SK_PermuteSingleSrc, llvm::TargetTransformInfo::SK_Reverse, and llvm::ARM_MB::ST.

◆ getStoreVectorFactor()

unsigned GCNTTIImpl::getStoreVectorFactor ( unsigned  VF,
unsigned  StoreSize,
unsigned  ChainSizeInBytes,
VectorType *  VecTy 
) const

Definition at line 244 of file AMDGPUTargetTransformInfo.cpp.

◆ getTgtMemIntrinsic()

bool GCNTTIImpl::getTgtMemIntrinsic ( IntrinsicInst *  Inst,
MemIntrinsicInfo &  Info 
) const

Definition at line 306 of file AMDGPUTargetTransformInfo.cpp.

References llvm::Intrinsic::amdgcn_atomic_dec, llvm::Intrinsic::amdgcn_atomic_inc, llvm::Intrinsic::amdgcn_ds_fadd, llvm::Intrinsic::amdgcn_ds_fmax, llvm::Intrinsic::amdgcn_ds_fmin, llvm::Intrinsic::amdgcn_ds_ordered_add, llvm::Intrinsic::amdgcn_ds_ordered_swap, llvm::dyn_cast(), llvm::CallBase::getArgOperand(), llvm::IntrinsicInst::getIntrinsicID(), llvm::ConstantInt::getZExtValue(), llvm::MemIntrinsicInfo::IsVolatile, llvm::MemIntrinsicInfo::Ordering, llvm::MemIntrinsicInfo::PtrVal, llvm::MemIntrinsicInfo::ReadMem, llvm::SequentiallyConsistent, and llvm::MemIntrinsicInfo::WriteMem.

◆ getUnrollingPreferences()

void GCNTTIImpl::getUnrollingPreferences ( Loop *  L,
ScalarEvolution &  SE,
TTI::UnrollingPreferences &  UP 
)

Definition at line 624 of file AMDGPUTargetTransformInfo.cpp.

◆ getVectorInstrCost()

int GCNTTIImpl::getVectorInstrCost ( unsigned  Opcode,
Type *  ValTy,
unsigned  Index 
)

Definition at line 489 of file AMDGPUTargetTransformInfo.cpp.

References llvm::TargetTransformInfoImplBase::DL, llvm::DataLayout::getTypeSizeInBits(), llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getVectorInstrCost(), and llvm::ARM_MB::ST.

◆ getVectorSplitCost()

unsigned llvm::GCNTTIImpl::getVectorSplitCost ( )
inline

Definition at line 185 of file AMDGPUTargetTransformInfo.h.

References llvm::TargetTransformInfoImplBase::areInlineCompatible(), Callee, llvm::BasicTTIImplBase< AMDGPUTTIImpl >::getShuffleCost(), and Kind.

◆ hasBranchDivergence()

bool llvm::GCNTTIImpl::hasBranchDivergence ( )
inline

Definition at line 126 of file AMDGPUTargetTransformInfo.h.

References llvm::AMDGPUTTIImpl::getUnrollingPreferences().

◆ isAlwaysUniform()

bool GCNTTIImpl::isAlwaysUniform ( const Value *  V) const

Definition at line 575 of file AMDGPUTargetTransformInfo.cpp.

References llvm::Intrinsic::amdgcn_readfirstlane, and llvm::Intrinsic::amdgcn_readlane.

◆ isLegalToVectorizeLoadChain()

bool GCNTTIImpl::isLegalToVectorizeLoadChain ( unsigned  ChainSizeInBytes,
unsigned  Alignment,
unsigned  AddrSpace 
) const

Definition at line 285 of file AMDGPUTargetTransformInfo.cpp.

◆ isLegalToVectorizeMemChain()

bool GCNTTIImpl::isLegalToVectorizeMemChain ( unsigned  ChainSizeInBytes,
unsigned  Alignment,
unsigned  AddrSpace 
) const

Definition at line 272 of file AMDGPUTargetTransformInfo.cpp.

References AMDGPUAS::PRIVATE_ADDRESS, and llvm::ARM_MB::ST.

◆ isLegalToVectorizeStoreChain()

bool GCNTTIImpl::isLegalToVectorizeStoreChain ( unsigned  ChainSizeInBytes,
unsigned  Alignment,
unsigned  AddrSpace 
) const

Definition at line 291 of file AMDGPUTargetTransformInfo.cpp.

◆ isSourceOfDivergence()

bool GCNTTIImpl::isSourceOfDivergence ( const Value *  V) const
Returns
true if the result of the value could potentially be different across workitems in a wavefront.

Definition at line 544 of file AMDGPUTargetTransformInfo.cpp.

References AMDGPUAS::FLAT_ADDRESS, isArgPassedInSGPR(), llvm::AMDGPU::isIntrinsicSourceOfDivergence(), llvm::SPII::Load, and AMDGPUAS::PRIVATE_ADDRESS.

The documentation for this class was generated from the following files:
Generated on Sun Dec 20 2020 14:02:46 for LLVM by   doxygen 1.8.13