#include "Target/AMDGPU/R600Subtarget.h"

Inheritance diagram for llvm::R600Subtarget:

Public Member Functions
	R600Subtarget (const Triple &TT, StringRef CPU, StringRef FS, const TargetMachine &TM)

	~R600Subtarget () override

const R600InstrInfo *	getInstrInfo () const override

const R600FrameLowering *	getFrameLowering () const override

const R600TargetLowering *	getTargetLowering () const override

const R600RegisterInfo *	getRegisterInfo () const override

const InstrItineraryData *	getInstrItineraryData () const override

const SelectionDAGTargetInfo *	getSelectionDAGInfo () const override

void	ParseSubtargetFeatures (StringRef CPU, StringRef TuneCPU, StringRef FS)

Generation	getGeneration () const

Align	getStackAlignment () const

R600Subtarget &	initializeSubtargetDependencies (const Triple &TT, StringRef GPU, StringRef FS)

bool	hasBFE () const

bool	hasBFI () const

bool	hasBCNT (unsigned Size) const

bool	hasBORROW () const

bool	hasCARRY () const

bool	hasCaymanISA () const

bool	hasFFBL () const

bool	hasFFBH () const

bool	hasFMA () const

bool	hasCFAluBug () const

bool	hasVertexCache () const

short	getTexVTXClauseSize () const

bool	enableMachineScheduler () const override

bool	enableSubRegLiveness () const override

unsigned	getMaxWorkGroupsPerCU (unsigned FlatWorkGroupSize) const override

unsigned	getMinFlatWorkGroupSize () const override

unsigned	getMaxFlatWorkGroupSize () const override

unsigned	getWavesPerEUForWorkGroup (unsigned FlatWorkGroupSize) const override

unsigned	getMinWavesPerEU () const override

bool	requiresDisjointEarlyClobberAndUndef () const override

Public Member Functions inherited from llvm::AMDGPUSubtarget
	AMDGPUSubtarget (Triple TT)

std::pair< unsigned, unsigned >	getDefaultFlatWorkGroupSize (CallingConv::ID CC) const

std::pair< unsigned, unsigned >	getFlatWorkGroupSizes (const Function &F) const

std::pair< unsigned, unsigned >	getWavesPerEU (const Function &F) const

std::pair< unsigned, unsigned >	getWavesPerEU (const Function &F, std::pair< unsigned, unsigned > FlatWorkGroupSizes) const
	Overload which uses the specified values for the flat work group sizes, rather than querying the function itself.

std::pair< unsigned, unsigned >	getEffectiveWavesPerEU (std::pair< unsigned, unsigned > WavesPerEU, std::pair< unsigned, unsigned > FlatWorkGroupSizes) const

unsigned	getMaxLocalMemSizeWithWaveCount (unsigned WaveCount, const Function &) const
	Return the amount of LDS that can be used that will not restrict the occupancy lower than WaveCount.

unsigned	getOccupancyWithLocalMemSize (uint32_t Bytes, const Function &) const
	Inverse of getMaxLocalMemWithWaveCount.

unsigned	getOccupancyWithLocalMemSize (const MachineFunction &MF) const

bool	isAmdHsaOS () const

bool	isAmdPalOS () const

bool	isMesa3DOS () const

bool	isMesaKernel (const Function &F) const

bool	isAmdHsaOrMesa (const Function &F) const

bool	isGCN () const

bool	isGCN3Encoding () const

bool	has16BitInsts () const

bool	hasTrue16BitInsts () const
	Return true if the subtarget supports True16 instructions.

bool	useRealTrue16Insts () const
	Return true if real (non-fake) variants of True16 instructions using 16-bit registers should be code-generated.

bool	hasBF16ConversionInsts () const

bool	hasMadMixInsts () const

bool	hasFP8ConversionScaleInsts () const

bool	hasBF8ConversionScaleInsts () const

bool	hasFP4ConversionScaleInsts () const

bool	hasFP6BF6ConversionScaleInsts () const

bool	hasF16BF16ToFP6BF6ConversionScaleInsts () const

bool	hasCvtPkF16F32Inst () const

bool	hasF32ToF16BF16ConversionSRInsts () const

bool	hasMadMacF32Insts () const

bool	hasDsSrc2Insts () const

bool	hasSDWA () const

bool	hasVOP3PInsts () const

bool	hasMulI24 () const

bool	hasMulU24 () const

bool	hasSMulHi () const

bool	hasInv2PiInlineImm () const

bool	hasFminFmaxLegacy () const

bool	hasTrigReducedRange () const

bool	hasFastFMAF32 () const

bool	isPromoteAllocaEnabled () const

unsigned	getWavefrontSize () const

unsigned	getWavefrontSizeLog2 () const

unsigned	getLocalMemorySize () const
	Return the maximum number of bytes of LDS available for all workgroups running on the same WGP or CU.

unsigned	getAddressableLocalMemorySize () const
	Return the maximum number of bytes of LDS that can be allocated to a single workgroup.

unsigned	getEUsPerCU () const
	Number of SIMDs/EUs (execution units) per "CU" ("compute unit"), where the "CU" is the unit onto which workgroups are mapped.

Align	getAlignmentForImplicitArgPtr () const

unsigned	getExplicitKernelArgOffset () const
	Returns the offset in bytes from the start of the input buffer of the first explicit kernel argument.

virtual unsigned	getMaxWorkGroupsPerCU (unsigned FlatWorkGroupSize) const =0

virtual unsigned	getMinFlatWorkGroupSize () const =0

virtual unsigned	getMaxFlatWorkGroupSize () const =0

virtual unsigned	getWavesPerEUForWorkGroup (unsigned FlatWorkGroupSize) const =0

virtual unsigned	getMinWavesPerEU () const =0

unsigned	getMaxWavesPerEU () const

unsigned	getMaxWorkitemID (const Function &Kernel, unsigned Dimension) const
	Return the maximum workitem ID value in the function, for the given (0, 1, 2) dimension.

SmallVector< unsigned >	getMaxNumWorkGroups (const Function &F) const
	Return the number of work groups for the function.

bool	isSingleLaneExecution (const Function &Kernel) const
	Return true if only a single workitem can be active in a wave.

bool	makeLIDRangeMetadata (Instruction *I) const
	Creates value range metadata on an workitemid.* intrinsic call or load.

unsigned	getImplicitArgNumBytes (const Function &F) const

uint64_t	getExplicitKernArgSize (const Function &F, Align &MaxAlign) const

unsigned	getKernArgSegmentSize (const Function &F, Align &MaxAlign) const

AMDGPUDwarfFlavour	getAMDGPUDwarfFlavour () const

virtual	~AMDGPUSubtarget ()=default

Additional Inherited Members
Public Types inherited from llvm::AMDGPUSubtarget
enum	Generation { INVALID = 0 , R600 = 1 , R700 = 2 , EVERGREEN = 3 , NORTHERN_ISLANDS = 4 , SOUTHERN_ISLANDS = 5 , SEA_ISLANDS = 6 , VOLCANIC_ISLANDS = 7 , GFX9 = 8 , GFX10 = 9 , GFX11 = 10 , GFX12 = 11 }

Static Public Member Functions inherited from llvm::AMDGPUSubtarget
static const AMDGPUSubtarget &	get (const MachineFunction &MF)

static const AMDGPUSubtarget &	get (const TargetMachine &TM, const Function &F)

Protected Attributes inherited from llvm::AMDGPUSubtarget
bool	GCN3Encoding = false

bool	Has16BitInsts = false

bool	HasTrue16BitInsts = false

bool	HasFP8ConversionScaleInsts = false

bool	HasBF8ConversionScaleInsts = false

bool	HasFP4ConversionScaleInsts = false

bool	HasFP6BF6ConversionScaleInsts = false

bool	HasF16BF16ToFP6BF6ConversionScaleInsts = false

bool	HasCvtPkF16F32Inst = false

bool	HasF32ToF16BF16ConversionSRInsts = false

bool	EnableRealTrue16Insts = false

bool	HasBF16ConversionInsts = false

bool	HasMadMixInsts = false

bool	HasMadMacF32Insts = false

bool	HasDsSrc2Insts = false

bool	HasSDWA = false

bool	HasVOP3PInsts = false

bool	HasMulI24 = true

bool	HasMulU24 = true

bool	HasSMulHi = false

bool	HasInv2PiInlineImm = false

bool	HasFminFmaxLegacy = true

bool	EnablePromoteAlloca = false

bool	HasTrigReducedRange = false

bool	FastFMAF32 = false

unsigned	EUsPerCU = 4

unsigned	MaxWavesPerEU = 10

unsigned	LocalMemorySize = 0

unsigned	AddressableLocalMemorySize = 0

char	WavefrontSizeLog2 = 0

Detailed Description

Definition at line 28 of file R600Subtarget.h.

Constructor & Destructor Documentation

◆ R600Subtarget()

R600Subtarget::R600Subtarget	(	const Triple &	TT,
		StringRef	CPU,
		StringRef	FS,
		const TargetMachine &	TM
	)

Definition at line 26 of file R600Subtarget.cpp.

References llvm::AMDGPUSubtarget::AddressableLocalMemorySize, and llvm::AMDGPUSubtarget::LocalMemorySize.

◆ ~R600Subtarget()

R600Subtarget::~R600Subtarget ( )

overridedefault

Member Function Documentation

◆ enableMachineScheduler()

bool llvm::R600Subtarget::enableMachineScheduler ( ) const

inlineoverride

Definition at line 125 of file R600Subtarget.h.

◆ enableSubRegLiveness()

bool llvm::R600Subtarget::enableSubRegLiveness ( ) const

inlineoverride

Definition at line 129 of file R600Subtarget.h.

◆ getFrameLowering()

const R600FrameLowering * llvm::R600Subtarget::getFrameLowering ( ) const

inlineoverride

Definition at line 53 of file R600Subtarget.h.

◆ getGeneration()

Generation llvm::R600Subtarget::getGeneration ( ) const

inline

Definition at line 73 of file R600Subtarget.h.

Referenced by llvm::R600InstrInfo::buildSlotOfVectorInstruction(), hasBCNT(), hasBFE(), hasBFI(), hasBORROW(), hasCARRY(), hasFFBH(), hasFFBL(), and initializeSubtargetDependencies().

◆ getInstrInfo()

const R600InstrInfo * llvm::R600Subtarget::getInstrInfo ( ) const

inlineoverride

Definition at line 51 of file R600Subtarget.h.

Referenced by llvm::R600AsmPrinter::emitInstruction(), and llvm::R600TargetLowering::EmitInstrWithCustomInserter().

◆ getInstrItineraryData()

const InstrItineraryData * llvm::R600Subtarget::getInstrItineraryData ( ) const

inlineoverride

Definition at line 65 of file R600Subtarget.h.

◆ getMaxFlatWorkGroupSize()

unsigned llvm::R600Subtarget::getMaxFlatWorkGroupSize ( ) const

inlineoverridevirtual

Returns: Maximum flat work group size supported by the subtarget.

Implements llvm::AMDGPUSubtarget.

Definition at line 145 of file R600Subtarget.h.

References llvm::AMDGPU::IsaInfo::getMaxFlatWorkGroupSize().

◆ getMaxWorkGroupsPerCU()

unsigned llvm::R600Subtarget::getMaxWorkGroupsPerCU ( unsigned FlatWorkGroupSize ) const

inlineoverridevirtual

Returns: Maximum number of work groups per compute unit supported by the subtarget and limited by given FlatWorkGroupSize.

Implements llvm::AMDGPUSubtarget.

Definition at line 135 of file R600Subtarget.h.

References llvm::AMDGPU::IsaInfo::getMaxWorkGroupsPerCU().

◆ getMinFlatWorkGroupSize()

unsigned llvm::R600Subtarget::getMinFlatWorkGroupSize ( ) const

inlineoverridevirtual

Returns: Minimum flat work group size supported by the subtarget.

Implements llvm::AMDGPUSubtarget.

Definition at line 140 of file R600Subtarget.h.

References llvm::AMDGPU::IsaInfo::getMinFlatWorkGroupSize().

◆ getMinWavesPerEU()

unsigned llvm::R600Subtarget::getMinWavesPerEU ( ) const

inlineoverridevirtual

Returns: Minimum number of waves per execution unit supported by the subtarget.

Implements llvm::AMDGPUSubtarget.

Definition at line 158 of file R600Subtarget.h.

References llvm::AMDGPU::IsaInfo::getMinWavesPerEU().

◆ getRegisterInfo()

const R600RegisterInfo * llvm::R600Subtarget::getRegisterInfo ( ) const

inlineoverride

Definition at line 61 of file R600Subtarget.h.

References llvm::R600InstrInfo::getRegisterInfo().

Referenced by llvm::R600TargetLowering::R600TargetLowering().

◆ getSelectionDAGInfo()

const SelectionDAGTargetInfo * R600Subtarget::getSelectionDAGInfo ( ) const

override

Definition at line 39 of file R600Subtarget.cpp.

◆ getStackAlignment()

Align llvm::R600Subtarget::getStackAlignment ( ) const

inline

Definition at line 77 of file R600Subtarget.h.

◆ getTargetLowering()

const R600TargetLowering * llvm::R600Subtarget::getTargetLowering ( ) const

inlineoverride

Definition at line 57 of file R600Subtarget.h.

◆ getTexVTXClauseSize()

short llvm::R600Subtarget::getTexVTXClauseSize ( ) const

inline

Definition at line 123 of file R600Subtarget.h.

◆ getWavesPerEUForWorkGroup()

unsigned llvm::R600Subtarget::getWavesPerEUForWorkGroup ( unsigned FlatWorkGroupSize ) const

inlineoverridevirtual

Returns: Number of waves per execution unit required to support the given FlatWorkGroupSize.

Implements llvm::AMDGPUSubtarget.

Definition at line 152 of file R600Subtarget.h.

References llvm::AMDGPU::IsaInfo::getWavesPerEUForWorkGroup().

◆ hasBCNT()

bool llvm::R600Subtarget::hasBCNT ( unsigned Size ) const

inline

Definition at line 90 of file R600Subtarget.h.

References llvm::AMDGPUSubtarget::EVERGREEN, getGeneration(), and Size.

Referenced by llvm::R600TargetLowering::R600TargetLowering().

◆ hasBFE()

bool llvm::R600Subtarget::hasBFE ( ) const

inline

Definition at line 82 of file R600Subtarget.h.

References llvm::AMDGPUSubtarget::EVERGREEN, and getGeneration().

Referenced by llvm::R600TargetLowering::R600TargetLowering().

◆ hasBFI()

bool llvm::R600Subtarget::hasBFI ( ) const

inline

Definition at line 86 of file R600Subtarget.h.

References llvm::AMDGPUSubtarget::EVERGREEN, and getGeneration().

Referenced by llvm::R600TargetLowering::R600TargetLowering().

◆ hasBORROW()

bool llvm::R600Subtarget::hasBORROW ( ) const

inline

Definition at line 97 of file R600Subtarget.h.

References llvm::AMDGPUSubtarget::EVERGREEN, and getGeneration().

Referenced by llvm::R600TargetLowering::R600TargetLowering().

◆ hasCARRY()

bool llvm::R600Subtarget::hasCARRY ( ) const

inline

Definition at line 101 of file R600Subtarget.h.

References llvm::AMDGPUSubtarget::EVERGREEN, and getGeneration().

Referenced by llvm::R600TargetLowering::R600TargetLowering().

◆ hasCaymanISA()

bool llvm::R600Subtarget::hasCaymanISA ( ) const

inline

Definition at line 105 of file R600Subtarget.h.

Referenced by initializeSubtargetDependencies(), and llvm::R600InstrInfo::isTransOnly().

◆ hasCFAluBug()

bool llvm::R600Subtarget::hasCFAluBug ( ) const

inline

Definition at line 119 of file R600Subtarget.h.

◆ hasFFBH()

bool llvm::R600Subtarget::hasFFBH ( ) const

inline

Definition at line 113 of file R600Subtarget.h.

References llvm::AMDGPUSubtarget::EVERGREEN, and getGeneration().

Referenced by llvm::R600TargetLowering::R600TargetLowering().

◆ hasFFBL()

bool llvm::R600Subtarget::hasFFBL ( ) const

inline

Definition at line 109 of file R600Subtarget.h.

References llvm::AMDGPUSubtarget::EVERGREEN, and getGeneration().

Referenced by llvm::R600TargetLowering::R600TargetLowering().

◆ hasFMA()

bool llvm::R600Subtarget::hasFMA ( ) const

inline

Definition at line 117 of file R600Subtarget.h.

Referenced by llvm::R600TargetLowering::R600TargetLowering().

◆ hasVertexCache()

bool llvm::R600Subtarget::hasVertexCache ( ) const

inline

Definition at line 121 of file R600Subtarget.h.

Referenced by llvm::R600InstrInfo::usesTextureCache(), and llvm::R600InstrInfo::usesVertexCache().

◆ initializeSubtargetDependencies()

R600Subtarget & R600Subtarget::initializeSubtargetDependencies	(	const Triple &	TT,
		StringRef	GPU,
		StringRef	FS
	)

Definition at line 43 of file R600Subtarget.cpp.

References llvm::AMDGPUSubtarget::EVERGREEN, getGeneration(), hasCaymanISA(), llvm::AMDGPUSubtarget::HasMulI24, llvm::AMDGPUSubtarget::HasMulU24, and ParseSubtargetFeatures().

◆ ParseSubtargetFeatures()

void llvm::R600Subtarget::ParseSubtargetFeatures	(	StringRef	CPU,
		StringRef	TuneCPU,
		StringRef	FS
	)

Referenced by initializeSubtargetDependencies().

◆ requiresDisjointEarlyClobberAndUndef()

bool llvm::R600Subtarget::requiresDisjointEarlyClobberAndUndef ( ) const

inlineoverride

Definition at line 162 of file R600Subtarget.h.

The documentation for this class was generated from the following files:

lib/Target/AMDGPU/R600Subtarget.h
lib/Target/AMDGPU/R600Subtarget.cpp

Public Member Functions

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ R600Subtarget()

◆ ~R600Subtarget()

Member Function Documentation

◆ enableMachineScheduler()

◆ enableSubRegLiveness()

◆ getFrameLowering()

◆ getGeneration()

◆ getInstrInfo()

◆ getInstrItineraryData()

◆ getMaxFlatWorkGroupSize()

◆ getMaxWorkGroupsPerCU()

◆ getMinFlatWorkGroupSize()

◆ getMinWavesPerEU()

◆ getRegisterInfo()

◆ getSelectionDAGInfo()

◆ getStackAlignment()

◆ getTargetLowering()

◆ getTexVTXClauseSize()

◆ getWavesPerEUForWorkGroup()

◆ hasBCNT()

◆ hasBFE()

◆ hasBFI()

◆ hasBORROW()

◆ hasCARRY()

◆ hasCaymanISA()

◆ hasCFAluBug()

◆ hasFFBH()

◆ hasFFBL()

◆ hasFMA()

◆ hasVertexCache()

◆ initializeSubtargetDependencies()

◆ ParseSubtargetFeatures()

◆ requiresDisjointEarlyClobberAndUndef()