#include "Target/AMDGPU/GCNSubtarget.h"
Public Types | |
| enum class | TrapHandlerAbi { NONE = 0x00 , AMDHSA = 0x01 } |
| enum class | TrapID { LLVMAMDHSATrap = 0x02 , LLVMAMDHSADebugTrap = 0x03 } |
| 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 | |
| static bool | hasHalfRate64Ops (const TargetSubtargetInfo &STI) |
| 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) |
Detailed Description
Definition at line 33 of file GCNSubtarget.h.
Member Enumeration Documentation
◆ TrapHandlerAbi
|
strong |
| Enumerator | |
|---|---|
| NONE | |
| AMDHSA | |
Definition at line 40 of file GCNSubtarget.h.
◆ TrapID
|
strong |
| Enumerator | |
|---|---|
| LLVMAMDHSATrap | |
| LLVMAMDHSADebugTrap | |
Definition at line 45 of file GCNSubtarget.h.
Constructor & Destructor Documentation
◆ GCNSubtarget()
| GCNSubtarget::GCNSubtarget | ( | const Triple & | TT, |
| StringRef | GPU, | ||
| StringRef | FS, | ||
| const GCNTargetMachine & | TM ) |
Definition at line 167 of file GCNSubtarget.cpp.
References llvm::AMDGPUSubtarget::AMDGPUSubtarget(), llvm::AMDGPUSubtarget::EUsPerCU, llvm::AMDGPU::IsaInfo::getEUsPerCU(), llvm::AMDGPU::IsaInfo::getMaxWavesPerEU(), getStackAlignment(), getTargetLowering(), initializeSubtargetDependencies(), InstrItins, llvm::AMDGPUSubtarget::MaxWavesPerEU, and TargetID.
Referenced by initializeSubtargetDependencies(), and ~GCNSubtarget().
◆ ~GCNSubtarget()
|
overridedefault |
References F, and GCNSubtarget().
Member Function Documentation
◆ adjustSchedDependency()
|
override |
Definition at line 630 of file GCNSubtarget.cpp.
References llvm::SDep::Data, E(), getInstrItineraryData(), llvm::ilist_node_impl< OptionsT >::getIterator(), llvm::SDep::getKind(), llvm::SDep::getLatency(), llvm::MachineInstr::getParent(), llvm::SDep::getReg(), getRegisterInfo(), I, llvm::MachineBasicBlock::instr_end(), llvm::MachineInstr::isBundle(), llvm::SDep::setLatency(), and TRI.
◆ checkSubtargetFeatures()
Diagnose inconsistent subtarget features before attempting to codegen function F.
Definition at line 158 of file GCNSubtarget.cpp.
References F.
Referenced by llvm::AMDGPUDAGToDAGISel::runOnMachineFunction(), and llvm::AMDGPUInstructionSelector::setupMF().
◆ computeOccupancy()
| std::pair< unsigned, unsigned > GCNSubtarget::computeOccupancy | ( | const Function & | F, |
| unsigned | LDSSize = 0, | ||
| unsigned | NumSGPRs = 0, | ||
| unsigned | NumVGPRs = 0 ) const |
Subtarget's minimum/maximum occupancy, in number of waves per EU, that can be achieved when the only function running on a CU is F, each workgroup uses LDSSize bytes of LDS, and each wave uses NumSGPRs SGPRs and NumVGPRs VGPRs.
The flat workgroup sizes associated to the function are a range, so this returns a range as well.
Note that occupancy can be affected by the scratch allocation as well, but we do not have enough information to compute it.
Definition at line 440 of file GCNSubtarget.cpp.
References F, llvm::AMDGPU::getDynamicVGPRBlockSize(), getDynamicVGPRBlockSize(), getOccupancyWithNumSGPRs(), getOccupancyWithNumVGPRs(), llvm::AMDGPUSubtarget::getOccupancyWithWorkGroupSizes(), and isDynamicVGPREnabled().
◆ d16PreservesUnusedBits()
|
inline |
Definition at line 761 of file GCNSubtarget.h.
References hasD16LoadStore(), and TargetID.
◆ dumpCode()
|
inline |
Definition at line 572 of file GCNSubtarget.h.
References DumpCode.
Referenced by llvm::AMDGPUAsmPrinter::runOnMachineFunction().
◆ enableEarlyIfConversion()
|
inlineoverride |
Definition at line 1077 of file GCNSubtarget.h.
◆ enableFlatScratch()
|
inline |
Definition at line 717 of file GCNSubtarget.h.
References EnableFlatScratch, flatScratchIsArchitected(), and hasFlatScratchInsts().
Referenced by getMaxPrivateElementSize().
◆ enableMachineScheduler()
|
inlineoverride |
Definition at line 1060 of file GCNSubtarget.h.
◆ enableSIScheduler()
|
inline |
Definition at line 1233 of file GCNSubtarget.h.
References EnableSIScheduler.
Referenced by overrideSchedPolicy().
◆ enableSubRegLiveness()
|
inlineoverride |
Definition at line 1066 of file GCNSubtarget.h.
◆ flatScratchIsArchitected()
|
inline |
- Returns
- true if the flat_scratch register is initialized by the HW. In this case it is readonly.
Definition at line 1531 of file GCNSubtarget.h.
References HasArchitectedFlatScratch.
Referenced by enableFlatScratch().
◆ flatScratchIsPointer()
|
inline |
- Returns
- true if the flat_scratch register should be initialized with the pointer to the wave's scratch memory rather than a size and offset.
Definition at line 1525 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ getAddressableNumArchVGPRs()
|
inline |
- Returns
- Addressable number of architectural VGPRs supported by the subtarget.
Definition at line 1716 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getAddressableNumArchVGPRs().
◆ getAddressableNumSGPRs()
|
inline |
- Returns
- Addressable number of SGPRs supported by the subtarget.
Definition at line 1642 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getAddressableNumSGPRs().
◆ getAddressableNumVGPRs()
- Returns
- Addressable number of VGPRs supported by the subtarget.
Definition at line 1721 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getAddressableNumVGPRs().
◆ getBaseMaxNumSGPRs()
| unsigned GCNSubtarget::getBaseMaxNumSGPRs | ( | const Function & | F, |
| std::pair< unsigned, unsigned > | WavesPerEU, | ||
| unsigned | PreloadedSGPRs, | ||
| unsigned | ReservedNumSGPRs ) const |
- Returns
- max num SGPRs. This is the common utility function called by MachineFunction and Function variants of getMaxNumSGPRs.
Definition at line 457 of file GCNSubtarget.cpp.
References F, llvm::AMDGPU::IsaInfo::FIXED_NUM_SGPRS_FOR_INIT_BUG, getMaxNumSGPRs(), getMinNumSGPRs(), and hasSGPRInitBug().
Referenced by getMaxNumSGPRs(), and getMaxNumSGPRs().
◆ getBaseMaxNumVGPRs()
| unsigned GCNSubtarget::getBaseMaxNumVGPRs | ( | const Function & | F, |
| std::pair< unsigned, unsigned > | NumVGPRBounds ) const |
- Returns
- max num VGPRs. This is the common utility function called by MachineFunction and Function variants of getMaxNumVGPRs.
Definition at line 541 of file GCNSubtarget.cpp.
References F, and hasGFX90AInsts().
Referenced by getMaxNumVGPRs().
◆ getBaseReservedNumSGPRs()
- Returns
- Reserved number of SGPRs. This is common utility function called by MachineFunction and Function variants of getReservedNumSGPRs.
Definition at line 410 of file GCNSubtarget.cpp.
References getGeneration(), llvm::AMDGPUSubtarget::GFX10, HasArchitectedFlatScratch, isXNACKEnabled(), llvm::AMDGPUSubtarget::SEA_ISLANDS, and llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS.
Referenced by getReservedNumSGPRs(), and getReservedNumSGPRs().
◆ getBoolRC()
|
inline |
Definition at line 1795 of file GCNSubtarget.h.
References llvm::SIRegisterInfo::getBoolRC(), and getRegisterInfo().
◆ getCallLowering()
|
inlineoverride |
Definition at line 340 of file GCNSubtarget.h.
◆ getConstantBusLimit()
Definition at line 196 of file GCNSubtarget.cpp.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX10.
◆ getDynamicVGPRBlockSize()
|
inline |
Definition at line 1849 of file GCNSubtarget.h.
References DynamicVGPRBlockSize32.
Referenced by computeOccupancy(), and getMaxNumVGPRs().
◆ getFrameLowering()
|
inlineoverride |
Definition at line 326 of file GCNSubtarget.h.
◆ getGeneration()
|
inline |
Definition at line 370 of file GCNSubtarget.h.
References Gen.
Referenced by createOccupancy(), flatScratchIsPointer(), getBaseReservedNumSGPRs(), getConstantBusLimit(), getMaxWaveScratchSize(), getNSAThreshold(), getOccupancyWithNumSGPRs(), getSetRegWaitStates(), getSNopBits(), has12DWordStoreHazard(), hasAddr64(), hasD16Images(), hasD16LoadStore(), hasDenormModeInst(), hasDPPBroadcasts(), hasDPPWavefrontShifts(), hasDsAtomicAsyncBarrierArriveB64PipeBug(), hasDX10ClampMode(), hasExtendedWaitCounts(), hasFlat(), hasFlatLgkmVMemCountInOrder(), hasFlatScratchSVSSwizzleBug(), hasFmaakFmamkF32Insts(), hasFPAtomicToDenormModeHazard(), hasFractBug(), hasGWSAutoReplay(), hasIEEEMode(), hasInstPrefetch(), hasLdsDirect(), hasLdsWaitVMSRC(), hasLegacyGeometry(), hasMad64_32(), hasMed3_16(), hasMergedShaders(), hasMin3Max3_16(), hasMultiDwordFlatScratchAddressing(), hasNoF16PseudoScalarTransInlineConstants(), hasNonNSAEncoding(), hasOnlyRevVALUShifts(), hasPermLane64(), hasPermLaneX16(), hasPKF32InstsReplicatingLower32BitsOfScalarInput(), hasReadM0LdsDirectHazard(), hasReadM0LdsDmaHazard(), hasReadM0MovRelInterpHazard(), hasReadM0SendMsgHazard(), hasReadVCCZBug(), hasRFEHazards(), hasRrWGMode(), hasScalarAddSub64(), hasScalarCompareEq64(), hasScalarSMulU64(), hasScalarSubwordLoads(), hasSCmpK(), hasScratchBaseForwardingHazard(), hasSignedScratchOffsets(), hasSMRDReadVALUDefHazard(), hasSplitBarriers(), hasUsableDivScaleConditionOutput(), hasUsableDSOffset(), hasVALUMaskWriteHazard(), hasVALUPartialForwardingHazard(), hasVMEMReadSGPRVALUDefHazard(), hasVOP3DPP(), initializeSubtargetDependencies(), ldsRequiresM0Init(), partialVCCWritesUpdateVCCZ(), privateMemoryResourceIsRangeChecked(), llvm::AMDGPUAsmPrinter::runOnMachineFunction(), setRegModeNeedsVNOPs(), shouldClusterStores(), supportsGetDoorbellID(), supportsMinMaxDenormModes(), supportsWave32(), supportsWGP(), vmemWriteNeedsExpWaitcnt(), and zeroesHigh16BitsOfDest().
◆ getInlineAsmLowering()
|
inlineoverride |
Definition at line 344 of file GCNSubtarget.h.
◆ getInstrInfo()
|
inlineoverride |
Definition at line 322 of file GCNSubtarget.h.
Referenced by llvm::SITargetLowering::AddMemOpInit(), llvm::SITargetLowering::AdjustInstrPostInstrSelection(), llvm::SITargetLowering::bundleInstWithWaitcnt(), llvm::SITargetLowering::emitGWSMemViolTestLoop(), llvm::AMDGPUAsmPrinter::emitInstruction(), llvm::SITargetLowering::EmitInstrWithCustomInserter(), llvm::SIProgramInfo::getFunctionCodeSize(), llvm::SITargetLowering::getPrefLoopAlignment(), llvm::SITargetLowering::PerformDAGCombine(), llvm::SITargetLowering::PostISelFolding(), llvm::SITargetLowering::splitKillBlock(), and llvm::SITargetLowering::wrapAddr64Rsrc().
◆ getInstrItineraryData()
|
inlineoverride |
Definition at line 364 of file GCNSubtarget.h.
References InstrItins.
Referenced by adjustSchedDependency().
◆ getInstructionSelector()
|
inlineoverride |
Definition at line 348 of file GCNSubtarget.h.
◆ getKnownHighZeroBitsForFrameIndex()
|
inline |
Return the number of high bits known to be zero for a frame index.
Definition at line 389 of file GCNSubtarget.h.
References llvm::countl_zero(), getMaxWaveScratchSize(), and llvm::AMDGPUSubtarget::getWavefrontSizeLog2().
Referenced by llvm::SITargetLowering::LowerFormalArguments().
◆ getLDSBankCount()
|
inline |
Definition at line 393 of file GCNSubtarget.h.
References LDSBankCount.
◆ getLegalizerInfo()
|
inlineoverride |
Definition at line 352 of file GCNSubtarget.h.
◆ getMaxFlatWorkGroupSize()
|
inlineoverridevirtual |
- Returns
- Maximum flat work group size supported by the subtarget.
Implements llvm::AMDGPUSubtarget.
Definition at line 1811 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getMaxFlatWorkGroupSize().
◆ getMaxLocalMemSizeWithWaveCount()
| unsigned llvm::GCNSubtarget::getMaxLocalMemSizeWithWaveCount | ( | unsigned | WaveCount, |
| const Function & | ) const |
Return the amount of LDS that can be used that will not restrict the occupancy lower than WaveCount.
◆ getMaxNumAGPRs()
Definition at line 1757 of file GCNSubtarget.h.
References F, and getMaxNumVGPRs().
◆ getMaxNumPreloadedSGPRs()
| unsigned GCNSubtarget::getMaxNumPreloadedSGPRs | ( | ) | const |
- Returns
- Maximum number of preloaded SGPRs for the subtarget.
Definition at line 511 of file GCNSubtarget.cpp.
Referenced by getMaxNumSGPRs().
◆ getMaxNumSGPRs() [1/3]
- Returns
- Maximum number of SGPRs that meets number of waves per execution unit requirement for function
F, or number of SGPRs explicitly requested using "amdgpu-num-sgpr" attribute attached to functionF. - Value that meets number of waves per execution unit requirement if explicitly requested value cannot be converted to integer, violates subtarget's specifications, or does not meet number of waves per execution unit requirement.
Definition at line 536 of file GCNSubtarget.cpp.
References F, getBaseMaxNumSGPRs(), getMaxNumPreloadedSGPRs(), getReservedNumSGPRs(), and llvm::AMDGPUSubtarget::getWavesPerEU().
◆ getMaxNumSGPRs() [2/3]
| unsigned GCNSubtarget::getMaxNumSGPRs | ( | const MachineFunction & | MF | ) | const |
- Returns
- Maximum number of SGPRs that meets number of waves per execution unit requirement for function
MF, or number of SGPRs explicitly requested using "amdgpu-num-sgpr" attribute attached to functionMF. - Value that meets number of waves per execution unit requirement if explicitly requested value cannot be converted to integer, violates subtarget's specifications, or does not meet number of waves per execution unit requirement.
Definition at line 504 of file GCNSubtarget.cpp.
References F, getBaseMaxNumSGPRs(), llvm::MachineFunction::getFunction(), llvm::MachineFunction::getInfo(), llvm::SIMachineFunctionInfo::getNumPreloadedSGPRs(), getReservedNumSGPRs(), and llvm::SIMachineFunctionInfo::getWavesPerEU().
◆ getMaxNumSGPRs() [3/3]
- Returns
- Maximum number of SGPRs that meets the given number of waves per execution unit requirement supported by the subtarget.
Definition at line 1654 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getMaxNumSGPRs().
Referenced by getBaseMaxNumSGPRs().
◆ getMaxNumUserSGPRs()
|
inline |
Definition at line 1089 of file GCNSubtarget.h.
References llvm::AMDGPU::getMaxNumUserSGPRs().
◆ getMaxNumVectorRegs()
Return a pair of maximum numbers of VGPRs and AGPRs that meet the number of waves per execution unit required for the function MF.
Definition at line 574 of file GCNSubtarget.cpp.
References llvm::alignTo(), assert(), F, llvm::AMDGPU::getIntegerPairAttribute(), getMaxNumVGPRs(), has1024AddressableVGPRs(), hasGFX90AInsts(), and hasMAIInsts().
◆ getMaxNumVGPRs() [1/3]
- Returns
- Maximum number of VGPRs that meets number of waves per execution unit requirement for function
F, or number of VGPRs explicitly requested using "amdgpu-num-vgpr" attribute attached to functionF. - Value that meets number of waves per execution unit requirement if explicitly requested value cannot be converted to integer, violates subtarget's specifications, or does not meet number of waves per execution unit requirement.
Definition at line 556 of file GCNSubtarget.cpp.
References F, getBaseMaxNumVGPRs(), llvm::AMDGPU::getDynamicVGPRBlockSize(), getDynamicVGPRBlockSize(), getMaxNumVGPRs(), getMinNumVGPRs(), llvm::AMDGPUSubtarget::getWavesPerEU(), and isDynamicVGPREnabled().
◆ getMaxNumVGPRs() [2/3]
| unsigned GCNSubtarget::getMaxNumVGPRs | ( | const MachineFunction & | MF | ) | const |
- Returns
- Maximum number of VGPRs that meets number of waves per execution unit requirement for function
MF, or number of VGPRs explicitly requested using "amdgpu-num-vgpr" attribute attached to functionMF. - Value that meets number of waves per execution unit requirement if explicitly requested value cannot be converted to integer, violates subtarget's specifications, or does not meet number of waves per execution unit requirement.
Definition at line 569 of file GCNSubtarget.cpp.
References llvm::MachineFunction::getFunction(), and getMaxNumVGPRs().
◆ getMaxNumVGPRs() [3/3]
|
inline |
- Returns
- the maximum number of VGPRs that can be used and still achieved at least the specified number of waves
WavesPerEU.
Definition at line 1735 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getMaxNumVGPRs().
Referenced by getMaxNumAGPRs(), getMaxNumVectorRegs(), getMaxNumVGPRs(), and getMaxNumVGPRs().
◆ getMaxPrivateElementSize()
Definition at line 397 of file GCNSubtarget.h.
References enableFlatScratch(), and MaxPrivateElementSize.
Referenced by llvm::SITargetLowering::canMergeStoresTo().
◆ getMaxWaveScratchSize()
|
inline |
Definition at line 374 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX11, and llvm::AMDGPUSubtarget::GFX12.
Referenced by getKnownHighZeroBitsForFrameIndex().
◆ getMaxWavesPerEU()
|
inline |
- Returns
- Maximum number of waves per execution unit supported by the subtarget without any kind of limitation.
Definition at line 382 of file AMDGPUSubtarget.h.
Referenced by getOccupancyWithNumSGPRs().
◆ getMaxWorkGroupsPerCU()
|
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 1801 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getMaxWorkGroupsPerCU().
◆ getMinFlatWorkGroupSize()
|
inlineoverridevirtual |
- Returns
- Minimum flat work group size supported by the subtarget.
Implements llvm::AMDGPUSubtarget.
Definition at line 1806 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getMinFlatWorkGroupSize().
◆ getMinNumSGPRs()
- Returns
- Minimum number of SGPRs that meets the given number of waves per execution unit requirement supported by the subtarget.
Definition at line 1648 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getMinNumSGPRs().
Referenced by getBaseMaxNumSGPRs().
◆ getMinNumVGPRs()
|
inline |
- Returns
- the minimum number of VGPRs that will prevent achieving more than the specified number of waves
WavesPerEU.
Definition at line 1727 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getMinNumVGPRs().
Referenced by getMaxNumVGPRs().
◆ getMinWavesPerEU()
|
inlineoverridevirtual |
- Returns
- Minimum number of waves per execution unit supported by the subtarget.
Implements llvm::AMDGPUSubtarget.
Definition at line 1824 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getMinWavesPerEU().
◆ getNSAMaxSize()
Definition at line 1203 of file GCNSubtarget.h.
References llvm::AMDGPU::getNSAMaxSize().
◆ getNSAThreshold()
| unsigned GCNSubtarget::getNSAThreshold | ( | const MachineFunction & | MF | ) | const |
Definition at line 675 of file GCNSubtarget.cpp.
References llvm::Function::getFnAttributeAsParsedInteger(), llvm::MachineFunction::getFunction(), getGeneration(), llvm::AMDGPUSubtarget::GFX12, and NSAThreshold.
◆ getOccupancyWithNumSGPRs()
Return the maximum number of waves per SIMD for kernels using SGPRs SGPRs.
Definition at line 397 of file GCNSubtarget.cpp.
References getGeneration(), getMaxWavesPerEU(), and llvm::AMDGPU::IsaInfo::getOccupancyWithNumSGPRs().
Referenced by computeOccupancy().
◆ getOccupancyWithNumVGPRs()
| unsigned GCNSubtarget::getOccupancyWithNumVGPRs | ( | unsigned | VGPRs, |
| unsigned | DynamicVGPRBlockSize ) const |
Return the maximum number of waves per SIMD for kernels using VGPRs VGPRs.
Definition at line 403 of file GCNSubtarget.cpp.
References llvm::AMDGPU::IsaInfo::getNumWavesPerEUWithNumVGPRs().
Referenced by computeOccupancy().
◆ getRegBankInfo()
|
inlineoverride |
Definition at line 356 of file GCNSubtarget.h.
◆ getRegisterInfo()
|
inlineoverride |
Definition at line 334 of file GCNSubtarget.h.
Referenced by adjustSchedDependency(), llvm::GCNTargetMachine::convertFuncInfoToYAML(), llvm::AMDGPUAsmPrinter::emitInstruction(), getBoolRC(), llvm::SITargetLowering::insertCopiesSplitCSR(), llvm::SITargetLowering::isEligibleForTailCallOptimization(), llvm::SITargetLowering::LowerFormalArguments(), llvm::AMDGPUCallLowering::lowerFormalArgumentsKernel(), llvm::SITargetLowering::LowerReturn(), and llvm::SITargetLowering::SITargetLowering().
◆ getReservedNumSGPRs() [1/2]
- Returns
- Reserved number of SGPRs for given function
F.
Definition at line 431 of file GCNSubtarget.cpp.
References F, getBaseReservedNumSGPRs(), and hasFlatAddressSpace().
◆ getReservedNumSGPRs() [2/2]
| unsigned GCNSubtarget::getReservedNumSGPRs | ( | const MachineFunction & | MF | ) | const |
- Returns
- Reserved number of SGPRs for given machine function
MF.
Definition at line 426 of file GCNSubtarget.cpp.
References getBaseReservedNumSGPRs(), llvm::MachineFunction::getInfo(), llvm::SIMachineFunctionInfo::getUserSGPRInfo(), and llvm::GCNUserSGPRUsageInfo::hasFlatScratchInit().
Referenced by getMaxNumSGPRs(), and getMaxNumSGPRs().
◆ getScalarizeGlobalBehavior()
|
inline |
Definition at line 1071 of file GCNSubtarget.h.
References ScalarizeGlobal.
◆ getSelectionDAGInfo()
|
override |
Definition at line 192 of file GCNSubtarget.cpp.
◆ getSetRegWaitStates()
|
inline |
Number of hazard wait states for s_setreg_b32/s_setreg_imm32_b32.
Definition at line 568 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::SEA_ISLANDS.
◆ getSGPRAllocGranule()
|
inline |
- Returns
- SGPR allocation granularity supported by the subtarget.
Definition at line 1627 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getSGPRAllocGranule().
◆ getSGPREncodingGranule()
|
inline |
- Returns
- SGPR encoding granularity supported by the subtarget.
Definition at line 1632 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getSGPREncodingGranule().
◆ getSNopBits()
|
inline |
- Returns
- the number of significant bits in the immediate field of the S_NOP instruction.
Definition at line 1880 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX12, and llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS.
◆ getStackAlignment()
|
inline |
Definition at line 1058 of file GCNSubtarget.h.
Referenced by GCNSubtarget().
◆ getTargetID()
|
inline |
Definition at line 360 of file GCNSubtarget.h.
References TargetID.
Referenced by llvm::AMDGPUAsmPrinter::runOnMachineFunction().
◆ getTargetLowering()
|
inlineoverride |
Definition at line 330 of file GCNSubtarget.h.
Referenced by GCNSubtarget(), and getTM().
◆ getTotalNumSGPRs()
|
inline |
- Returns
- Total number of SGPRs supported by the subtarget.
Definition at line 1637 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getTotalNumSGPRs().
◆ getTotalNumVGPRs()
|
inline |
- Returns
- Total number of VGPRs supported by the subtarget.
Definition at line 1710 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getTotalNumVGPRs().
◆ getTrapHandlerAbi()
|
inline |
Definition at line 516 of file GCNSubtarget.h.
References AMDHSA, llvm::AMDGPUSubtarget::isAmdHsaOS(), and NONE.
◆ getVGPRAllocGranule()
- Returns
- VGPR allocation granularity supported by the subtarget.
Definition at line 1700 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getVGPRAllocGranule().
◆ getVGPREncodingGranule()
|
inline |
- Returns
- VGPR encoding granularity supported by the subtarget.
Definition at line 1705 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getVGPREncodingGranule().
◆ getWavesPerEUForWorkGroup()
|
inlineoverridevirtual |
- Returns
- Number of waves per execution unit required to support the given
FlatWorkGroupSize.
Implements llvm::AMDGPUSubtarget.
Definition at line 1818 of file GCNSubtarget.h.
References llvm::AMDGPU::IsaInfo::getWavesPerEUForWorkGroup().
◆ has1024AddressableVGPRs()
|
inline |
Definition at line 1473 of file GCNSubtarget.h.
References Has1024AddressableVGPRs.
Referenced by getMaxNumVectorRegs().
◆ has12DWordStoreHazard()
|
inline |
Definition at line 1259 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::SOUTHERN_ISLANDS.
◆ has1_5xVGPRs()
|
inline |
Definition at line 1393 of file GCNSubtarget.h.
References Has1_5xVGPRs.
◆ has45BitNumRecordsBufferResource()
|
inline |
- Returns
- true if the sub-target supports buffer resource (V#) with 45-bit num_records.
Definition at line 1890 of file GCNSubtarget.h.
References Has45BitNumRecordsBufferResource.
◆ has64BitLiterals()
|
inline |
Definition at line 1486 of file GCNSubtarget.h.
References Has64BitLiterals.
◆ hasA16()
|
inline |
Definition at line 1179 of file GCNSubtarget.h.
References HasA16.
◆ hasAddMinMaxInsts()
|
inline |
Definition at line 1601 of file GCNSubtarget.h.
References HasAddMinMaxInsts.
◆ hasAddNoCarry()
|
inline |
- Returns
- true if the target has integer add/sub instructions that do not produce a carry-out. This includes v_add_[iu]32, v_sub_[iu]32, v_add_[iu]16, and v_sub_[iu]16, all of which support the clamp modifier for saturation.
Definition at line 793 of file GCNSubtarget.h.
References AddNoCarryInsts.
◆ hasAddPC64Inst()
|
inline |
Definition at line 1471 of file GCNSubtarget.h.
References GFX1250Insts.
◆ hasAddr64()
|
inline |
Definition at line 438 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS.
Referenced by initializeSubtargetDependencies().
◆ hasAddSubU64Insts()
|
inline |
Definition at line 1585 of file GCNSubtarget.h.
References HasAddSubU64Insts.
◆ hasApertureRegs()
|
inline |
Definition at line 662 of file GCNSubtarget.h.
References HasApertureRegs.
◆ hasArchitectedSGPRs()
|
inline |
- Returns
- true if the architected SGPRs are enabled.
Definition at line 1534 of file GCNSubtarget.h.
References HasArchitectedSGPRs.
◆ hasAshrPkInsts()
|
inline |
Definition at line 1455 of file GCNSubtarget.h.
References HasAshrPkInsts.
◆ hasAtomicBufferGlobalPkAddF16Insts()
|
inline |
Definition at line 954 of file GCNSubtarget.h.
References HasAtomicBufferGlobalPkAddF16Insts.
◆ hasAtomicBufferGlobalPkAddF16NoRtnInsts()
|
inline |
Definition at line 950 of file GCNSubtarget.h.
References HasAtomicBufferGlobalPkAddF16NoRtnInsts.
◆ hasAtomicBufferPkAddBF16Inst()
|
inline |
Definition at line 962 of file GCNSubtarget.h.
References HasAtomicBufferPkAddBF16Inst.
◆ hasAtomicCSub()
|
inline |
Definition at line 726 of file GCNSubtarget.h.
References GFX10_BEncoding.
◆ hasAtomicCSubNoRtnInsts()
|
inline |
Definition at line 1562 of file GCNSubtarget.h.
References HasAtomicCSubNoRtnInsts.
◆ hasAtomicDsPkAdd16Insts()
|
inline |
Definition at line 938 of file GCNSubtarget.h.
References HasAtomicDsPkAdd16Insts.
◆ hasAtomicFaddInsts()
|
inline |
Definition at line 942 of file GCNSubtarget.h.
References HasAtomicFaddNoRtnInsts, and HasAtomicFaddRtnInsts.
◆ hasAtomicFaddNoRtnInsts()
|
inline |
Definition at line 948 of file GCNSubtarget.h.
References HasAtomicFaddNoRtnInsts.
◆ hasAtomicFaddRtnInsts()
|
inline |
Definition at line 946 of file GCNSubtarget.h.
References HasAtomicFaddRtnInsts.
◆ hasAtomicFlatPkAdd16Insts()
|
inline |
Definition at line 940 of file GCNSubtarget.h.
References HasAtomicFlatPkAdd16Insts.
◆ hasAtomicFMinFMaxF32FlatInsts()
|
inline |
Definition at line 930 of file GCNSubtarget.h.
References HasAtomicFMinFMaxF32FlatInsts.
◆ hasAtomicFMinFMaxF32GlobalInsts()
|
inline |
Definition at line 922 of file GCNSubtarget.h.
References HasAtomicFMinFMaxF32GlobalInsts.
◆ hasAtomicFMinFMaxF64FlatInsts()
|
inline |
Definition at line 934 of file GCNSubtarget.h.
References HasAtomicFMinFMaxF64FlatInsts.
◆ hasAtomicFMinFMaxF64GlobalInsts()
|
inline |
Definition at line 926 of file GCNSubtarget.h.
References HasAtomicFMinFMaxF64GlobalInsts.
◆ hasAtomicGlobalPkAddBF16Inst()
|
inline |
Definition at line 958 of file GCNSubtarget.h.
References HasAtomicGlobalPkAddBF16Inst.
◆ hasAutoWaitcntBeforeBarrier()
|
inline |
Definition at line 622 of file GCNSubtarget.h.
References AutoWaitcntBeforeBarrier.
◆ hasBCNT()
Definition at line 468 of file GCNSubtarget.h.
References Size.
◆ hasBFE()
|
inline |
Definition at line 456 of file GCNSubtarget.h.
Referenced by hasBFM().
◆ hasBFI()
|
inline |
Definition at line 460 of file GCNSubtarget.h.
◆ hasBFM()
|
inline |
Definition at line 464 of file GCNSubtarget.h.
References hasBFE().
◆ hasBitOp3Insts()
|
inline |
Definition at line 1451 of file GCNSubtarget.h.
References HasBitOp3Insts.
◆ hasBVHDualAndBVH8Insts()
|
inline |
Definition at line 1499 of file GCNSubtarget.h.
References HasBVHDualAndBVH8Insts.
◆ hasCARRY()
|
inline |
Definition at line 494 of file GCNSubtarget.h.
◆ hasClusters()
|
inline |
- Returns
- true if the subtarget supports clusters of workgroups.
Definition at line 1872 of file GCNSubtarget.h.
References HasClusters.
◆ hasCompressedExport()
|
inline |
Return true if the target's EXP instruction has the COMPR flag, which affects the meaning of the EN (enable) bits.
Definition at line 1387 of file GCNSubtarget.h.
References GFX11Insts.
◆ hasCubeInsts()
|
inline |
Definition at line 902 of file GCNSubtarget.h.
References HasCubeInsts.
◆ hasCvtFP8VOP1Bug()
|
inline |
Definition at line 1558 of file GCNSubtarget.h.
References HasCvtFP8Vop1Bug.
◆ hasCvtNormInsts()
|
inline |
Definition at line 910 of file GCNSubtarget.h.
References HasCvtNormInsts.
◆ hasCvtPkNormVOP2Insts()
|
inline |
Definition at line 912 of file GCNSubtarget.h.
References HasCvtPkNormVOP2Insts.
◆ hasCvtPkNormVOP3Insts()
|
inline |
Definition at line 914 of file GCNSubtarget.h.
References HasCvtPkNormVOP3Insts.
◆ hasCvtScaleForwardingHazard()
|
inline |
Definition at line 1361 of file GCNSubtarget.h.
References GFX950Insts.
◆ hasD16Images()
|
inline |
Definition at line 765 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS.
◆ hasD16LoadStore()
|
inline |
Definition at line 757 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
Referenced by d16PreservesUnusedBits().
◆ hasDefaultComponentBroadcast()
|
inline |
Definition at line 995 of file GCNSubtarget.h.
References HasDefaultComponentBroadcast.
◆ hasDefaultComponentZero()
|
inline |
Definition at line 993 of file GCNSubtarget.h.
References HasDefaultComponentZero.
◆ hasDelayAlu()
|
inline |
Return true if the target has the S_DELAY_ALU instruction.
Definition at line 1400 of file GCNSubtarget.h.
References GFX11Insts.
◆ hasDenormModeInst()
|
inline |
- Returns
- If target supports S_DENORM_MODE.
Definition at line 586 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX10.
◆ hasDLInsts()
|
inline |
Definition at line 834 of file GCNSubtarget.h.
References HasDLInsts.
◆ hasDot10Insts()
|
inline |
Definition at line 876 of file GCNSubtarget.h.
References HasDot10Insts.
◆ hasDot11Insts()
|
inline |
Definition at line 880 of file GCNSubtarget.h.
References HasDot11Insts.
◆ hasDot12Insts()
|
inline |
Definition at line 884 of file GCNSubtarget.h.
References HasDot12Insts.
◆ hasDot13Insts()
|
inline |
Definition at line 888 of file GCNSubtarget.h.
References HasDot13Insts.
◆ hasDot1Insts()
|
inline |
Definition at line 840 of file GCNSubtarget.h.
References HasDot1Insts.
◆ hasDot2Insts()
|
inline |
Definition at line 844 of file GCNSubtarget.h.
References HasDot2Insts.
◆ hasDot3Insts()
|
inline |
Definition at line 848 of file GCNSubtarget.h.
References HasDot3Insts.
◆ hasDot4Insts()
|
inline |
Definition at line 852 of file GCNSubtarget.h.
References HasDot4Insts.
◆ hasDot5Insts()
|
inline |
Definition at line 856 of file GCNSubtarget.h.
References HasDot5Insts.
◆ hasDot6Insts()
|
inline |
Definition at line 860 of file GCNSubtarget.h.
References HasDot6Insts.
◆ hasDot7Insts()
|
inline |
Definition at line 864 of file GCNSubtarget.h.
References HasDot7Insts.
◆ hasDot8Insts()
|
inline |
Definition at line 868 of file GCNSubtarget.h.
References HasDot8Insts.
◆ hasDot9Insts()
|
inline |
Definition at line 872 of file GCNSubtarget.h.
References HasDot9Insts.
◆ hasDOTOpSelHazard()
|
inline |
Definition at line 1328 of file GCNSubtarget.h.
References GFX11Insts, and GFX940Insts.
◆ hasDPALU_DPP()
|
inline |
Definition at line 1146 of file GCNSubtarget.h.
References HasDPALU_DPP.
◆ hasDPP()
|
inline |
Definition at line 1130 of file GCNSubtarget.h.
References HasDPP.
◆ hasDPP8()
|
inline |
Definition at line 1142 of file GCNSubtarget.h.
References HasDPP8.
◆ hasDPPBroadcasts()
|
inline |
Definition at line 1134 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX10, and HasDPP.
◆ hasDPPSrc1SGPR()
|
inline |
Definition at line 1150 of file GCNSubtarget.h.
References HasDPPSrc1SGPR.
◆ hasDPPWavefrontShifts()
|
inline |
Definition at line 1138 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX10, and HasDPP.
◆ hasDS96AndDS128()
|
inline |
- Returns
- If target supports ds_read/write_b96/128.
Definition at line 601 of file GCNSubtarget.h.
References CIInsts.
◆ hasDsAtomicAsyncBarrierArriveB64PipeBug()
|
inline |
Definition at line 1861 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasDstSelForwardingHazard()
|
inline |
Definition at line 1325 of file GCNSubtarget.h.
References GFX940Insts.
◆ hasDwordx3LoadStores()
|
inline |
Definition at line 1264 of file GCNSubtarget.h.
References CIInsts.
◆ hasDX10ClampMode()
|
inline |
Definition at line 1565 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasEmulatedSystemScopeAtomics()
|
inline |
- Returns
- true is HW emulates system scope atomics unsupported by the PCI-e via CAS loop.
Definition at line 989 of file GCNSubtarget.h.
References HasEmulatedSystemScopeAtomics.
Referenced by globalMemoryFPAtomicIsLegal().
◆ hasExportInsts()
|
inline |
Definition at line 734 of file GCNSubtarget.h.
References hasGFX1250Insts(), and hasGFX940Insts().
◆ hasExtendedImageInsts()
|
inline |
Definition at line 1171 of file GCNSubtarget.h.
References HasExtendedImageInsts.
◆ hasExtendedWaitCounts()
|
inline |
- Returns
- true if the target uses LOADcnt/SAMPLEcnt/BVHcnt, DScnt/KMcnt and STOREcnt rather than VMcnt, LGKMcnt and VScnt respectively.
Definition at line 1433 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasFFBH()
|
inline |
Definition at line 476 of file GCNSubtarget.h.
◆ hasFFBL()
|
inline |
Definition at line 472 of file GCNSubtarget.h.
◆ hasFlat()
|
inline |
Definition at line 442 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::SOUTHERN_ISLANDS.
Referenced by initializeSubtargetDependencies().
◆ hasFlatAddressSpace()
|
inline |
Definition at line 684 of file GCNSubtarget.h.
References FlatAddressSpace.
Referenced by getReservedNumSGPRs(), and hasFlatScrRegister().
◆ hasFlatAtomicFaddF32Inst()
|
inline |
Definition at line 966 of file GCNSubtarget.h.
References HasFlatAtomicFaddF32Inst.
◆ hasFlatBufferGlobalAtomicFaddF64Inst()
|
inline |
- Returns
- true if the target has flat, global, and buffer atomic fadd for double.
Definition at line 970 of file GCNSubtarget.h.
References HasFlatBufferGlobalAtomicFaddF64Inst.
◆ hasFlatGlobalInsts()
|
inline |
Definition at line 696 of file GCNSubtarget.h.
References FlatGlobalInsts.
◆ hasFlatGVSMode()
|
inline |
Definition at line 1228 of file GCNSubtarget.h.
References FlatGVSMode.
◆ hasFlatInstOffsets()
|
inline |
Definition at line 692 of file GCNSubtarget.h.
References FlatInstOffsets.
◆ hasFlatLgkmVMemCountInOrder()
|
inline |
Definition at line 753 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ hasFlatScratchInsts()
|
inline |
Definition at line 700 of file GCNSubtarget.h.
References FlatScratchInsts.
Referenced by enableFlatScratch(), and hasFlatScratchSTMode().
◆ hasFlatScratchSTMode()
|
inline |
Definition at line 707 of file GCNSubtarget.h.
References hasFlatScratchInsts(), hasGFX10_3Insts(), and hasGFX940Insts().
◆ hasFlatScratchSVSMode()
|
inline |
Definition at line 711 of file GCNSubtarget.h.
References GFX11Insts, and GFX940Insts.
◆ hasFlatScratchSVSSwizzleBug()
|
inline |
Definition at line 1397 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX11.
◆ hasFlatScrRegister()
|
inline |
Definition at line 688 of file GCNSubtarget.h.
References hasFlatAddressSpace().
◆ hasFlatSegmentOffsetBug()
|
inline |
Definition at line 749 of file GCNSubtarget.h.
References HasFlatSegmentOffsetBug.
◆ hasFMA()
|
inline |
Definition at line 498 of file GCNSubtarget.h.
References FMA.
◆ hasFmaakFmamkF32Insts()
|
inline |
Definition at line 1161 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX10, and hasGFX940Insts().
◆ hasFmaakFmamkF64Insts()
|
inline |
Definition at line 1165 of file GCNSubtarget.h.
References hasGFX1250Insts().
◆ hasFmacF64Inst()
|
inline |
Definition at line 838 of file GCNSubtarget.h.
References HasFmacF64Inst.
◆ hasFmaMixBF16Insts()
|
inline |
Definition at line 492 of file GCNSubtarget.h.
References HasFmaMixBF16Insts.
◆ hasFmaMixInsts()
|
inline |
Definition at line 488 of file GCNSubtarget.h.
References HasFmaMixInsts.
◆ hasFormattedMUBUFInsts()
|
inline |
Definition at line 732 of file GCNSubtarget.h.
References hasGFX1250Insts().
◆ hasFP64()
|
inline |
Definition at line 418 of file GCNSubtarget.h.
References FP64.
Referenced by initializeSubtargetDependencies().
◆ hasFP8ConversionInsts()
|
inline |
Definition at line 900 of file GCNSubtarget.h.
References HasFP8ConversionInsts.
◆ hasFP8E5M3Insts()
|
inline |
Definition at line 916 of file GCNSubtarget.h.
References HasFP8E5M3Insts.
◆ hasFP8Insts()
|
inline |
Definition at line 896 of file GCNSubtarget.h.
References HasFP8Insts.
◆ hasFPAtomicToDenormModeHazard()
|
inline |
Definition at line 1345 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX10.
◆ hasFractBug()
|
inline |
Definition at line 452 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::SOUTHERN_ISLANDS.
◆ hasFullRate64Ops()
|
inline |
Definition at line 434 of file GCNSubtarget.h.
References FullRate64Ops.
◆ hasG16()
|
inline |
Definition at line 1181 of file GCNSubtarget.h.
References HasG16.
◆ hasGDS()
|
inline |
- Returns
- true if Global Data Share is supported.
Definition at line 1537 of file GCNSubtarget.h.
References HasGDS.
◆ hasGetPCZeroExtension()
|
inline |
Definition at line 1618 of file GCNSubtarget.h.
References GFX1250Insts, and GFX12Insts.
◆ hasGetWaveIdInst()
|
inline |
Definition at line 1007 of file GCNSubtarget.h.
References HasGetWaveIdInst.
◆ hasGFX10_3Insts()
|
inline |
Definition at line 1215 of file GCNSubtarget.h.
References GFX10_3Insts.
Referenced by hasFlatScratchSTMode().
◆ hasGFX10_AEncoding()
|
inline |
Definition at line 1207 of file GCNSubtarget.h.
References GFX10_AEncoding.
◆ hasGFX10_BEncoding()
|
inline |
Definition at line 1211 of file GCNSubtarget.h.
References GFX10_BEncoding.
◆ hasGFX1250Insts()
|
inline |
Definition at line 1580 of file GCNSubtarget.h.
References GFX1250Insts.
Referenced by hasExportInsts(), hasFmaakFmamkF64Insts(), hasFormattedMUBUFInsts(), hasLdsAtomicAddF64(), hasMTBUFInsts(), hasVINTERPEncoding(), and supportsWave64().
◆ hasGFX90AInsts()
|
inline |
Definition at line 1343 of file GCNSubtarget.h.
References GFX90AInsts.
Referenced by getBaseMaxNumVGPRs(), getMaxNumVectorRegs(), hasLdsAtomicAddF64(), and llvm::AMDGPUAsmPrinter::runOnMachineFunction().
◆ hasGFX940Insts()
|
inline |
Definition at line 1406 of file GCNSubtarget.h.
References GFX940Insts.
Referenced by hasExportInsts(), hasFlatScratchSTMode(), and hasFmaakFmamkF32Insts().
◆ hasGFX950Insts()
|
inline |
Definition at line 1410 of file GCNSubtarget.h.
References GFX950Insts.
Referenced by hasLDSLoadB96_B128().
◆ hasGlobalAddTidInsts()
|
inline |
Definition at line 722 of file GCNSubtarget.h.
References GFX10_BEncoding.
◆ hasGloballyAddressableScratch()
|
inline |
Definition at line 1367 of file GCNSubtarget.h.
References HasGloballyAddressableScratch.
Referenced by getPrivateAtomicExpansionKind().
◆ hasGWS()
|
inline |
- Returns
- true if Global Wave Sync is supported.
Definition at line 1540 of file GCNSubtarget.h.
References HasGWS.
◆ hasGWSAutoReplay()
|
inline |
Definition at line 780 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ hasGWSSemaReleaseAll()
|
inline |
- Returns
- if target has ds_gws_sema_release_all instruction.
Definition at line 785 of file GCNSubtarget.h.
References CIInsts.
◆ hasHalfRate64Ops() [1/2]
|
inline |
Definition at line 430 of file GCNSubtarget.h.
References HalfRate64Ops.
◆ hasHalfRate64Ops() [2/2]
|
static |
◆ hasHardClauses()
|
inline |
Definition at line 1341 of file GCNSubtarget.h.
References MaxHardClauseLength.
◆ hasHWFP64()
|
inline |
Definition at line 426 of file GCNSubtarget.h.
References FP64.
◆ hasIEEEMinimumMaximumInsts()
|
inline |
Definition at line 1571 of file GCNSubtarget.h.
References HasIEEEMinimumMaximumInsts.
◆ hasIEEEMode()
|
inline |
Definition at line 1568 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasImageGather4D16Bug()
|
inline |
Definition at line 1189 of file GCNSubtarget.h.
References HasImageGather4D16Bug.
◆ hasImageInsts()
|
inline |
Definition at line 1167 of file GCNSubtarget.h.
References HasImageInsts.
◆ hasImageStoreD16Bug()
|
inline |
Definition at line 1187 of file GCNSubtarget.h.
References HasImageStoreD16Bug.
◆ hasInstFwdPrefetchBug()
|
inline |
Definition at line 1301 of file GCNSubtarget.h.
References HasInstFwdPrefetchBug.
◆ hasInstPrefetch()
|
inline |
Definition at line 1035 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX10, and llvm::AMDGPUSubtarget::GFX11.
◆ hasIntClamp()
|
inline |
Definition at line 414 of file GCNSubtarget.h.
References HasIntClamp.
◆ hasIntMinMax64()
|
inline |
Definition at line 1598 of file GCNSubtarget.h.
References GFX1250Insts.
◆ hasKernargPreload()
|
inline |
Definition at line 1552 of file GCNSubtarget.h.
References KernargPreload.
Referenced by needsKernArgPreloadProlog().
◆ hasLdsAtomicAddF64()
|
inline |
Definition at line 741 of file GCNSubtarget.h.
References hasGFX1250Insts(), and hasGFX90AInsts().
◆ hasLdsBarrierArriveAtomic()
|
inline |
Definition at line 1490 of file GCNSubtarget.h.
References HasLdsBarrierArriveAtomic.
◆ hasLdsBranchVmemWARHazard()
|
inline |
Definition at line 1309 of file GCNSubtarget.h.
References HasLdsBranchVmemWARHazard.
◆ hasLdsDirect()
|
inline |
Definition at line 1351 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX11.
◆ hasLDSFPAtomicAddF32()
|
inline |
Definition at line 1121 of file GCNSubtarget.h.
References GFX8Insts.
◆ hasLDSFPAtomicAddF64()
|
inline |
Definition at line 1122 of file GCNSubtarget.h.
References GFX1250Insts, and GFX90AInsts.
◆ hasLDSLoadB96_B128()
|
inline |
Returns true if the target supports global_load_lds_dwordx3/global_load_lds_dwordx4 or buffer_load_dwordx3/buffer_load_dwordx4 with the lds bit.
Definition at line 1415 of file GCNSubtarget.h.
References hasGFX950Insts().
◆ hasLDSMisalignedBug()
|
inline |
Definition at line 1297 of file GCNSubtarget.h.
References EnableCuMode, and LDSMisalignedBug.
◆ hasLdsWaitVMSRC()
|
inline |
Definition at line 1353 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasLegacyGeometry()
|
inline |
Definition at line 1549 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX11.
◆ hasLerpInst()
|
inline |
Definition at line 904 of file GCNSubtarget.h.
References HasLerpInst.
◆ hasLshlAddU64Inst()
|
inline |
Definition at line 1223 of file GCNSubtarget.h.
References HasLshlAddU64Inst.
◆ hasMad64_32()
|
inline |
Definition at line 810 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::SEA_ISLANDS.
◆ hasMadF16()
| bool GCNSubtarget::hasMadF16 | ( | ) | const |
Definition at line 387 of file GCNSubtarget.cpp.
Referenced by llvm::SITargetLowering::SITargetLowering().
◆ hasMADIntraFwdBug()
|
inline |
Definition at line 1191 of file GCNSubtarget.h.
References HasMADIntraFwdBug.
◆ hasMadU32Inst()
|
inline |
Definition at line 1588 of file GCNSubtarget.h.
References HasMadU32Inst.
◆ hasMadU64U32NoCarry()
|
inline |
Definition at line 1595 of file GCNSubtarget.h.
References GFX1250Insts.
◆ hasMAIInsts()
|
inline |
Definition at line 892 of file GCNSubtarget.h.
References HasMAIInsts.
Referenced by llvm::AMDGPU::HSAMD::MetadataStreamerMsgPackV4::getHSAKernelProps(), getMaxNumVectorRegs(), and llvm::AMDGPUAsmPrinter::runOnMachineFunction().
◆ hasMed3_16()
|
inline |
Definition at line 480 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ hasMemoryAtomicFaddF32DenormalSupport()
|
inline |
- Returns
- true if the target's flat, global, and buffer atomic fadd for float supports denormal handling.
Definition at line 976 of file GCNSubtarget.h.
References HasMemoryAtomicFaddF32DenormalSupport.
◆ hasMergedShaders()
|
inline |
- Returns
- true if the machine has merged shaders in which s0-s7 are reserved by the hardware and user SGPRs start at s8
Definition at line 1544 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ hasMFMAInlineLiteralBug()
|
inline |
Definition at line 1255 of file GCNSubtarget.h.
References HasMFMAInlineLiteralBug.
◆ hasMIMG_R128()
|
inline |
Definition at line 422 of file GCNSubtarget.h.
References MIMG_R128.
◆ hasMin3Max3_16()
|
inline |
Definition at line 484 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
Referenced by supportsMin3Max3().
◆ hasMin3Max3PKF16()
|
inline |
Definition at line 1465 of file GCNSubtarget.h.
References HasMin3Max3PKF16.
Referenced by supportsMin3Max3().
◆ hasMinimum3Maximum3F16()
|
inline |
Definition at line 1461 of file GCNSubtarget.h.
References HasMinimum3Maximum3F16.
Referenced by supportsMin3Max3().
◆ hasMinimum3Maximum3F32()
|
inline |
Definition at line 1457 of file GCNSubtarget.h.
References HasMinimum3Maximum3F32.
Referenced by supportsMin3Max3().
◆ hasMinimum3Maximum3PKF16()
|
inline |
Definition at line 1475 of file GCNSubtarget.h.
References HasMinimum3Maximum3PKF16.
Referenced by supportsMin3Max3().
◆ hasMovB64()
|
inline |
Definition at line 1221 of file GCNSubtarget.h.
References GFX1250Insts, and GFX940Insts.
◆ hasMovrel()
|
inline |
Definition at line 1097 of file GCNSubtarget.h.
References HasMovrel.
Referenced by useVGPRIndexMode().
◆ hasMSAALoadDstSelBug()
|
inline |
Definition at line 1193 of file GCNSubtarget.h.
References HasMSAALoadDstSelBug.
◆ hasMTBUFInsts()
|
inline |
Definition at line 730 of file GCNSubtarget.h.
References hasGFX1250Insts().
◆ hasMultiDwordFlatScratchAddressing()
|
inline |
Definition at line 745 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ hasNegativeScratchOffsetBug()
|
inline |
Definition at line 1249 of file GCNSubtarget.h.
References NegativeScratchOffsetBug.
◆ hasNegativeUnalignedScratchOffsetBug()
|
inline |
Definition at line 1251 of file GCNSubtarget.h.
References NegativeUnalignedScratchOffsetBug.
◆ hasNoDataDepHazard()
|
inline |
Definition at line 1027 of file GCNSubtarget.h.
References HasNoDataDepHazard.
◆ hasNoF16PseudoScalarTransInlineConstants()
|
inline |
- Returns
- true if inline constants are not supported for F16 pseudo scalar transcendentals.
Definition at line 1437 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasNonNSAEncoding()
|
inline |
Definition at line 1199 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasNoSdstCMPX()
|
inline |
Definition at line 999 of file GCNSubtarget.h.
References HasNoSdstCMPX.
◆ hasNSAClauseBug()
|
inline |
Definition at line 1339 of file GCNSubtarget.h.
References HasNSAClauseBug.
◆ hasNSAEncoding()
|
inline |
Definition at line 1197 of file GCNSubtarget.h.
References HasNSAEncoding.
◆ hasNSAtoVMEMBug()
|
inline |
Definition at line 1335 of file GCNSubtarget.h.
References HasNSAtoVMEMBug.
◆ hasNullExportTarget()
|
inline |
Return true if the target's EXP instruction supports the NULL export target.
Definition at line 1391 of file GCNSubtarget.h.
References GFX11Insts.
◆ hasOffset3fBug()
|
inline |
Definition at line 1183 of file GCNSubtarget.h.
References HasOffset3fBug.
Referenced by llvm::AMDGPUAsmPrinter::emitInstruction().
◆ hasOnlyRevVALUShifts()
|
inline |
Definition at line 448 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS.
◆ hasPackedFP32Ops()
|
inline |
Definition at line 1152 of file GCNSubtarget.h.
References HasPackedFP32Ops.
◆ hasPackedTID()
|
inline |
Definition at line 1402 of file GCNSubtarget.h.
References HasPackedTID.
◆ hasPartialNSAEncoding()
|
inline |
Definition at line 1201 of file GCNSubtarget.h.
References HasPartialNSAEncoding.
◆ hasPermlane16Swap()
|
inline |
Definition at line 1453 of file GCNSubtarget.h.
References HasPermlane16Swap.
◆ hasPermlane32Swap()
|
inline |
Definition at line 1454 of file GCNSubtarget.h.
References HasPermlane32Swap.
◆ hasPermLane64()
|
inline |
- Returns
- true if the subtarget has the v_permlane64_b32 instruction.
Definition at line 1128 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX11.
◆ hasPermLaneX16()
|
inline |
- Returns
- true if the subtarget has the v_permlanex16_b32 instruction.
Definition at line 1125 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX10.
◆ hasPkAddMinMaxInsts()
|
inline |
Definition at line 1604 of file GCNSubtarget.h.
References HasPkAddMinMaxInsts.
◆ hasPKF32InstsReplicatingLower32BitsOfScalarInput()
|
inline |
- Returns
- true if the target has packed f32 instructions that only read 32 bits from a scalar operand (SGPR or literal) and replicates the bits to both channels.
Definition at line 1447 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX12, and GFX1250Insts.
◆ hasPkFmacF16Inst()
|
inline |
Definition at line 918 of file GCNSubtarget.h.
References HasPkFmacF16Inst.
◆ hasPkMinMax3Insts()
|
inline |
Definition at line 1607 of file GCNSubtarget.h.
References GFX1250Insts.
◆ hasPkMovB32()
|
inline |
Definition at line 1157 of file GCNSubtarget.h.
References GFX90AInsts.
◆ hasPointSampleAccel()
|
inline |
Definition at line 1488 of file GCNSubtarget.h.
References HasPointSampleAccel.
◆ hasPrefetch()
|
inline |
Definition at line 1039 of file GCNSubtarget.h.
References GFX12Insts.
◆ hasPrivEnabledTrap2NopBug()
|
inline |
Definition at line 1195 of file GCNSubtarget.h.
References HasPrivEnabledTrap2NopBug.
◆ hasPrngInst()
|
inline |
Definition at line 1497 of file GCNSubtarget.h.
References HasPrngInst.
◆ hasPseudoScalarTrans()
|
inline |
Definition at line 1423 of file GCNSubtarget.h.
References HasPseudoScalarTrans.
◆ hasQsadInsts()
|
inline |
Definition at line 908 of file GCNSubtarget.h.
References HasQsadInsts.
◆ hasR128A16()
|
inline |
Definition at line 1175 of file GCNSubtarget.h.
References HasR128A16.
◆ hasReadM0LdsDirectHazard()
|
inline |
Definition at line 1281 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ hasReadM0LdsDmaHazard()
|
inline |
Definition at line 1277 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ hasReadM0MovRelInterpHazard()
|
inline |
Definition at line 1268 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ hasReadM0SendMsgHazard()
|
inline |
Definition at line 1272 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX9, and llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS.
◆ hasReadVCCZBug()
|
inline |
Extra wait hazard is needed in some cases before s_cbranch_vccnz/s_cbranch_vccz.
Definition at line 542 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::SEA_ISLANDS.
◆ hasRelaxedBufferOOBMode()
|
inline |
Definition at line 660 of file GCNSubtarget.h.
References RelaxedBufferOOBMode.
◆ hasRequiredExportPriority()
|
inline |
Definition at line 1427 of file GCNSubtarget.h.
References HasRequiredExportPriority.
◆ hasRestrictedSOffset()
|
inline |
Definition at line 1425 of file GCNSubtarget.h.
References HasRestrictedSOffset.
Referenced by selectSOffset().
◆ hasRFEHazards()
|
inline |
Definition at line 563 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS.
◆ hasRrWGMode()
|
inline |
Definition at line 1574 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasSadInsts()
|
inline |
Definition at line 906 of file GCNSubtarget.h.
References HasSadInsts.
◆ hasSafeCUPrefetch()
|
inline |
Definition at line 1045 of file GCNSubtarget.h.
References HasSafeCUPrefetch.
◆ hasSafeSmemPrefetch()
|
inline |
Definition at line 1043 of file GCNSubtarget.h.
References HasSafeSmemPrefetch.
◆ hasSALUFloatInsts()
|
inline |
Definition at line 1421 of file GCNSubtarget.h.
References HasSALUFloatInsts.
◆ hasScalarAddSub64()
|
inline |
Definition at line 797 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasScalarAtomics()
|
inline |
Definition at line 1117 of file GCNSubtarget.h.
References HasScalarAtomics.
◆ hasScalarCompareEq64()
|
inline |
Definition at line 1107 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS.
◆ hasScalarDwordx3Loads()
|
inline |
Definition at line 1111 of file GCNSubtarget.h.
References HasScalarDwordx3Loads.
◆ hasScalarFlatScratchInsts()
|
inline |
Definition at line 713 of file GCNSubtarget.h.
References ScalarFlatScratchInsts.
◆ hasScalarMulHiInsts()
|
inline |
Definition at line 510 of file GCNSubtarget.h.
References GFX9Insts.
◆ hasScalarPackInsts()
|
inline |
Definition at line 506 of file GCNSubtarget.h.
References GFX9Insts.
◆ hasScalarSMulU64()
|
inline |
Definition at line 799 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasScalarStores()
|
inline |
Definition at line 1113 of file GCNSubtarget.h.
References HasScalarStores.
◆ hasScalarSubwordLoads()
|
inline |
Definition at line 514 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasScaleOffset()
|
inline |
Definition at line 1226 of file GCNSubtarget.h.
References GFX1250Insts.
◆ hasSCmpK()
|
inline |
Definition at line 1048 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasScratchBaseForwardingHazard()
|
inline |
Definition at line 1867 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX12, and GFX1250Insts.
◆ hasSDWAMac()
|
inline |
Definition at line 826 of file GCNSubtarget.h.
References HasSDWAMac.
◆ hasSDWAOmod()
|
inline |
Definition at line 814 of file GCNSubtarget.h.
References HasSDWAOmod.
◆ hasSDWAOutModsVOPC()
|
inline |
Definition at line 830 of file GCNSubtarget.h.
References HasSDWAOutModsVOPC.
◆ hasSDWAScalar()
|
inline |
Definition at line 818 of file GCNSubtarget.h.
References HasSDWAScalar.
◆ hasSDWASdst()
|
inline |
Definition at line 822 of file GCNSubtarget.h.
References HasSDWASdst.
◆ hasSetPrioIncWgInst()
|
inline |
Definition at line 1613 of file GCNSubtarget.h.
References HasSetPrioIncWgInst.
◆ hasSGetShaderCyclesInst()
|
inline |
Definition at line 1610 of file GCNSubtarget.h.
References GFX1250Insts.
◆ hasSGPRInitBug()
|
inline |
Definition at line 1241 of file GCNSubtarget.h.
References SGPRInitBug.
Referenced by getBaseMaxNumSGPRs().
◆ hasShaderCyclesHiLoRegisters()
|
inline |
Definition at line 1019 of file GCNSubtarget.h.
References HasShaderCyclesHiLoRegisters.
Referenced by llvm::SITargetLowering::EmitInstrWithCustomInserter().
◆ hasShaderCyclesRegister()
|
inline |
Definition at line 1015 of file GCNSubtarget.h.
References HasShaderCyclesRegister.
◆ hasShift64HighRegBug()
|
inline |
Definition at line 1315 of file GCNSubtarget.h.
References GFX90AInsts, and GFX940Insts.
◆ hasSignedGVSOffset()
|
inline |
Definition at line 1231 of file GCNSubtarget.h.
References GFX1250Insts.
◆ hasSignedScratchOffsets()
|
inline |
- Returns
- true if VADDR and SADDR fields in VSCRATCH can use negative values.
Definition at line 1578 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasSMemRealTime()
|
inline |
Definition at line 1093 of file GCNSubtarget.h.
References HasSMemRealTime.
◆ hasSMemTimeInst()
|
inline |
Definition at line 1011 of file GCNSubtarget.h.
References HasSMemTimeInst.
◆ hasSMEMtoVectorWriteHazard()
|
inline |
Definition at line 1293 of file GCNSubtarget.h.
References HasSMEMtoVectorWriteHazard.
◆ hasSMRDReadVALUDefHazard()
|
inline |
A read of an SGPR by SMRD instruction requires 4 wait states when the SGPR was written by a VALU instruction.
Definition at line 553 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::SOUTHERN_ISLANDS.
◆ hasSPackHL()
|
inline |
Return true if the target has the S_PACK_HL_B32_B16 instruction.
Definition at line 1383 of file GCNSubtarget.h.
References GFX11Insts.
◆ hasSplitBarriers()
|
inline |
Definition at line 1555 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX12.
◆ hasSwap()
|
inline |
Definition at line 502 of file GCNSubtarget.h.
References GFX9Insts.
◆ hasTanhInsts()
|
inline |
Definition at line 1467 of file GCNSubtarget.h.
References HasTanhInsts.
◆ hasTensorCvtLutInsts()
|
inline |
Definition at line 1469 of file GCNSubtarget.h.
References HasTensorCvtLutInsts.
◆ hasTransForwardingHazard()
|
inline |
Definition at line 1321 of file GCNSubtarget.h.
References GFX940Insts.
◆ hasTransposeLoadF4F6Insts()
|
inline |
Definition at line 1479 of file GCNSubtarget.h.
References HasTransposeLoadF4F6Insts.
◆ hasUnalignedAccessMode()
|
inline |
Definition at line 656 of file GCNSubtarget.h.
References UnalignedAccessMode.
◆ hasUnalignedBufferAccess()
|
inline |
Definition at line 632 of file GCNSubtarget.h.
References UnalignedBufferAccess.
◆ hasUnalignedBufferAccessEnabled()
|
inline |
Definition at line 636 of file GCNSubtarget.h.
References UnalignedAccessMode, and UnalignedBufferAccess.
◆ hasUnalignedDSAccess()
|
inline |
Definition at line 640 of file GCNSubtarget.h.
References UnalignedDSAccess.
◆ hasUnalignedDSAccessEnabled()
|
inline |
Definition at line 644 of file GCNSubtarget.h.
References UnalignedAccessMode, and UnalignedDSAccess.
◆ hasUnalignedScratchAccess()
|
inline |
Definition at line 648 of file GCNSubtarget.h.
References UnalignedScratchAccess.
◆ hasUnalignedScratchAccessEnabled()
|
inline |
Definition at line 652 of file GCNSubtarget.h.
References UnalignedAccessMode, and UnalignedScratchAccess.
◆ hasUnpackedD16VMem()
|
inline |
Definition at line 801 of file GCNSubtarget.h.
References HasUnpackedD16VMem.
◆ hasUsableDivScaleConditionOutput()
|
inline |
Condition output from div_scale is usable.
Definition at line 536 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::SOUTHERN_ISLANDS.
◆ hasUsableDSOffset()
|
inline |
True if the offset field of DS instructions works as expected.
On SI, the offset uses a 16-bit adder and does not always wrap properly.
Definition at line 527 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::SEA_ISLANDS.
◆ hasUserSGPRInit16Bug()
|
inline |
Definition at line 1245 of file GCNSubtarget.h.
References isWave32(), and UserSGPRInit16Bug.
◆ hasVALUMaskWriteHazard()
|
inline |
Definition at line 1371 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX11.
◆ hasVALUPartialForwardingHazard()
|
inline |
Definition at line 1355 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX11.
◆ hasVALUReadSGPRHazard()
|
inline |
Definition at line 1373 of file GCNSubtarget.h.
References GFX1250Insts, and GFX12Insts.
◆ hasVALUTransUseHazard()
|
inline |
Definition at line 1359 of file GCNSubtarget.h.
References HasVALUTransUseHazard.
◆ hasVcmpxExecWARHazard()
|
inline |
Definition at line 1305 of file GCNSubtarget.h.
References HasVcmpxExecWARHazard.
◆ hasVcmpxPermlaneHazard()
|
inline |
Definition at line 1285 of file GCNSubtarget.h.
References HasVcmpxPermlaneHazard.
◆ hasVDecCoExecHazard()
|
inline |
Definition at line 1331 of file GCNSubtarget.h.
References GFX940Insts.
◆ hasVectorMulU64()
|
inline |
Definition at line 1591 of file GCNSubtarget.h.
References GFX1250Insts.
◆ hasVGPRIndexMode()
|
inline |
Definition at line 1101 of file GCNSubtarget.h.
References HasVGPRIndexMode.
Referenced by useVGPRIndexMode().
◆ hasVINTERPEncoding()
|
inline |
Definition at line 738 of file GCNSubtarget.h.
References GFX11Insts, and hasGFX1250Insts().
◆ hasVmemPrefInsts()
|
inline |
Definition at line 1041 of file GCNSubtarget.h.
References HasVmemPrefInsts.
◆ hasVMEMReadSGPRVALUDefHazard()
|
inline |
A read of an SGPR by a VMEM instruction requires 5 wait states when the SGPR was written by a VALU Instruction.
Definition at line 559 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS.
◆ hasVMemToLDSLoad()
|
inline |
Definition at line 1419 of file GCNSubtarget.h.
References HasVMemToLDSLoad.
◆ hasVMEMtoScalarWriteHazard()
|
inline |
Definition at line 1289 of file GCNSubtarget.h.
References HasVMEMtoScalarWriteHazard.
◆ hasVmemWriteVgprInOrder()
|
inline |
Definition at line 1429 of file GCNSubtarget.h.
References HasVmemWriteVgprInOrder.
◆ hasVOP3DPP()
|
inline |
Definition at line 1349 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX11.
◆ hasVOP3Literal()
|
inline |
Definition at line 1023 of file GCNSubtarget.h.
References HasVOP3Literal.
◆ hasVOPD3()
|
inline |
Definition at line 1582 of file GCNSubtarget.h.
References GFX1250Insts.
◆ hasVOPDInsts()
|
inline |
Definition at line 1395 of file GCNSubtarget.h.
References HasVOPDInsts.
◆ hasVscnt()
|
inline |
Definition at line 1003 of file GCNSubtarget.h.
References HasVscnt.
◆ hasWaitXCnt()
|
inline |
- Returns
- true if the target has s_wait_xcnt insertion. Supported for GFX1250.
Definition at line 1483 of file GCNSubtarget.h.
References HasWaitXcnt.
◆ hasXF32Insts()
|
inline |
- Returns
- true if the target has instructions with xf32 format support.
Definition at line 1442 of file GCNSubtarget.h.
References HasXF32Insts.
◆ haveRoundOpsF64()
|
inline |
Have v_trunc_f64, v_ceil_f64, v_rndne_f64.
Definition at line 606 of file GCNSubtarget.h.
References CIInsts.
◆ initializeSubtargetDependencies()
| GCNSubtarget & GCNSubtarget::initializeSubtargetDependencies | ( | const Triple & | TT, |
| StringRef | GPU, | ||
| StringRef | FS ) |
Definition at line 57 of file GCNSubtarget.cpp.
References llvm::AMDGPUSubtarget::AddressableLocalMemorySize, llvm::Triple::AMDHSA, assert(), llvm::dbgs(), FlatForGlobal, GCNSubtarget(), Gen, getGeneration(), llvm::AMDGPU::IsaInfo::getLocalMemorySize(), llvm::AMDGPUSubtarget::getWavefrontSizeLog2(), llvm::AMDGPUSubtarget::GFX10, llvm::AMDGPUSubtarget::GFX9, hasAddr64(), hasFlat(), llvm::AMDGPUSubtarget::HasFminFmaxLegacy, hasFP64(), llvm::AMDGPUSubtarget::HasSMulHi, llvm::AMDGPUSubtarget::INVALID, llvm::AMDGPUSubtarget::isAmdHsaOS(), LDSBankCount, LLVM_DEBUG, llvm::AMDGPUSubtarget::LocalMemorySize, MaxPrivateElementSize, ParseSubtargetFeatures(), llvm::AMDGPUSubtarget::SEA_ISLANDS, llvm::AMDGPUSubtarget::SOUTHERN_ISLANDS, TargetID, llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS, and llvm::AMDGPUSubtarget::WavefrontSizeLog2.
Referenced by GCNSubtarget().
◆ isCuModeEnabled()
|
inline |
Definition at line 678 of file GCNSubtarget.h.
References EnableCuMode.
◆ isDynamicVGPREnabled()
|
inline |
Definition at line 1848 of file GCNSubtarget.h.
References DynamicVGPR.
Referenced by computeOccupancy(), and getMaxNumVGPRs().
◆ isMesaGfxShader()
Definition at line 806 of file GCNSubtarget.h.
References F, llvm::AMDGPUSubtarget::isMesa3DOS(), and llvm::AMDGPU::isShader().
◆ isPreciseMemoryEnabled()
|
inline |
Definition at line 682 of file GCNSubtarget.h.
References EnablePreciseMemory.
◆ isTgSplitEnabled()
|
inline |
Definition at line 674 of file GCNSubtarget.h.
References EnableTgSplit.
◆ isTrapHandlerEnabled()
|
inline |
Definition at line 666 of file GCNSubtarget.h.
References TrapHandler.
◆ isWave32()
|
inline |
Definition at line 1779 of file GCNSubtarget.h.
References llvm::AMDGPUSubtarget::getWavefrontSize().
Referenced by hasUserSGPRInit16Bug(), and mirFileLoaded().
◆ isWave64()
|
inline |
Definition at line 1783 of file GCNSubtarget.h.
References llvm::AMDGPUSubtarget::getWavefrontSize().
◆ isWaveSizeKnown()
|
inline |
Returns if the wavesize of this subtarget is known reliable.
This is false only for the a default target-cpu that does not have an explicit +wavefrontsize target feature.
Definition at line 1790 of file GCNSubtarget.h.
◆ isXNACKEnabled()
|
inline |
Definition at line 670 of file GCNSubtarget.h.
References TargetID.
Referenced by getBaseReservedNumSGPRs(), and needsConstrainedOpcode().
◆ ldsRequiresM0Init()
|
inline |
Return if most LDS instructions have an m0 use that require m0 to be initialized.
Definition at line 771 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ loadStoreOptEnabled()
|
inline |
Definition at line 1237 of file GCNSubtarget.h.
References EnableLoadStoreOpt.
◆ maxHardClauseLength()
|
inline |
- Returns
- The maximum number of instructions that can be enclosed in an S_CLAUSE on the given subtarget, or 0 for targets that do not support that instruction.
Definition at line 1495 of file GCNSubtarget.h.
References MaxHardClauseLength.
◆ mirFileLoaded()
|
override |
Definition at line 376 of file GCNSubtarget.cpp.
References isWave32(), MBB, and MI.
◆ needsAlignedVGPRs()
|
inline |
Return if operations acting on VGPR tuples require even alignment.
Definition at line 1380 of file GCNSubtarget.h.
References RequiresAlignVGPR.
◆ needsKernArgPreloadProlog()
|
inline |
Definition at line 1622 of file GCNSubtarget.h.
References GFX1250Insts, and hasKernargPreload().
◆ overridePostRASchedPolicy()
|
override |
Definition at line 339 of file GCNSubtarget.cpp.
References llvm::dbgs(), F, llvm::Attribute::getValueAsString(), llvm::Attribute::isValid(), LLVM_DEBUG, llvm::MachineSchedPolicy::OnlyBottomUp, and llvm::MachineSchedPolicy::OnlyTopDown.
◆ overrideSchedPolicy()
|
override |
Definition at line 322 of file GCNSubtarget.cpp.
References enableSIScheduler(), llvm::MachineSchedPolicy::OnlyBottomUp, llvm::MachineSchedPolicy::OnlyTopDown, llvm::MachineSchedPolicy::ShouldTrackLaneMasks, and llvm::MachineSchedPolicy::ShouldTrackPressure.
◆ ParseSubtargetFeatures()
Referenced by initializeSubtargetDependencies().
◆ partialVCCWritesUpdateVCCZ()
|
inline |
Writes to VCC_LO/VCC_HI update the VCCZ flag.
Definition at line 547 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX10.
◆ privateMemoryResourceIsRangeChecked()
|
inline |
- Returns
- If MUBUF instructions always perform range checking, even for buffer resources used for private memory access.
Definition at line 612 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ requiresCodeObjectV6()
|
inline |
Definition at line 1363 of file GCNSubtarget.h.
References RequiresCOV6.
◆ requiresDisjointEarlyClobberAndUndef()
|
inlineoverride |
Definition at line 1853 of file GCNSubtarget.h.
◆ requiresNopBeforeDeallocVGPRs()
|
inline |
Definition at line 1842 of file GCNSubtarget.h.
References GFX1250Insts.
◆ requiresWaitIdleBeforeGetReg()
|
inline |
Definition at line 1846 of file GCNSubtarget.h.
References GFX1250Insts.
◆ requiresWaitsBeforeSystemScopeStores()
|
inline |
Definition at line 1894 of file GCNSubtarget.h.
References RequiresWaitsBeforeSystemScopeStores.
◆ requiresWaitXCntBeforeAtomicStores()
|
inline |
- Returns
- true if the subtarget requires a wait for xcnt before atomic flat/global stores & rmw.
Definition at line 1876 of file GCNSubtarget.h.
References GFX1250Insts.
◆ setRegModeNeedsVNOPs()
|
inline |
Definition at line 1375 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX12, and GFX1250Insts.
◆ setScalarizeGlobalBehavior()
|
inline |
Definition at line 1070 of file GCNSubtarget.h.
References ScalarizeGlobal.
◆ shouldClusterStores()
|
inline |
Definition at line 1834 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX11.
◆ supportsAgentScopeFineGrainedRemoteMemoryAtomics()
|
inline |
- Returns
- true if atomic operations targeting fine-grained memory work correctly at device scope, in allocations in host or peer PCIe device memory.
Definition at line 983 of file GCNSubtarget.h.
References HasAgentScopeFineGrainedRemoteMemoryAtomics.
Referenced by globalMemoryFPAtomicIsLegal().
◆ supportsBackOffBarrier()
|
inline |
- Returns
- true if the target supports backing off of s_barrier instructions when an exception is raised.
Definition at line 628 of file GCNSubtarget.h.
References BackOffBarrier.
◆ supportsGetDoorbellID()
|
inline |
Definition at line 520 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ supportsMinMaxDenormModes()
|
inline |
Definition at line 581 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX9.
◆ supportsWave32()
|
inline |
Definition at line 1775 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::GFX10.
◆ supportsWave64()
|
inline |
Definition at line 1777 of file GCNSubtarget.h.
References hasGFX1250Insts().
◆ supportsWGP()
|
inline |
Definition at line 408 of file GCNSubtarget.h.
References getGeneration(), llvm::AMDGPUSubtarget::GFX10, and GFX1250Insts.
Referenced by llvm::AMDGPU::HSAMD::MetadataStreamerMsgPackV4::getHSAKernelProps().
◆ unsafeDSOffsetFoldingEnabled()
|
inline |
Definition at line 531 of file GCNSubtarget.h.
References EnableUnsafeDSOffsetFolding.
◆ useAA()
|
override |
Definition at line 395 of file GCNSubtarget.cpp.
References UseAA.
◆ useDS128()
|
inline |
- Returns
- If target supports ds_read/write_b128 and user enables generation of ds_read/write_b128.
Definition at line 596 of file GCNSubtarget.h.
References CIInsts, and EnableDS128.
◆ useFlatForGlobal()
|
inline |
Definition at line 590 of file GCNSubtarget.h.
References FlatForGlobal.
◆ usePRTStrictNull()
|
inline |
- Returns
- If target requires PRT Struct NULL support (zero result registers for sparse texture support).
Definition at line 618 of file GCNSubtarget.h.
References EnablePRTStrictNull.
◆ useVGPRBlockOpsForCSR()
|
inline |
Definition at line 1365 of file GCNSubtarget.h.
References UseBlockVGPROpsForCSR.
◆ useVGPRIndexMode()
| bool GCNSubtarget::useVGPRIndexMode | ( | ) | const |
Definition at line 391 of file GCNSubtarget.cpp.
References EnableVGPRIndexMode, hasMovrel(), and hasVGPRIndexMode().
◆ vmemWriteNeedsExpWaitcnt()
|
inline |
Definition at line 1031 of file GCNSubtarget.h.
References getGeneration(), and llvm::AMDGPUSubtarget::SEA_ISLANDS.
◆ zeroesHigh16BitsOfDest()
Returns if the result of this instruction with a 16-bit result returned in a 32-bit register implicitly zeroes the high 16-bits, rather than preserve the original value.
This list was mostly derived from experimentation.
Definition at line 224 of file GCNSubtarget.cpp.
References getGeneration(), llvm::AMDGPUSubtarget::GFX9, and llvm::AMDGPUSubtarget::VOLCANIC_ISLANDS.
Member Data Documentation
◆ AddNoCarryInsts
|
protected |
Definition at line 234 of file GCNSubtarget.h.
Referenced by hasAddNoCarry().
◆ AssemblerPermissiveWavesize
|
protected |
Definition at line 101 of file GCNSubtarget.h.
◆ AutoWaitcntBeforeBarrier
|
protected |
Definition at line 76 of file GCNSubtarget.h.
Referenced by hasAutoWaitcntBeforeBarrier().
◆ BackOffBarrier
|
protected |
Definition at line 77 of file GCNSubtarget.h.
Referenced by supportsBackOffBarrier().
◆ CIInsts
|
protected |
Definition at line 107 of file GCNSubtarget.h.
Referenced by hasDS96AndDS128(), hasDwordx3LoadStores(), hasGWSSemaReleaseAll(), haveRoundOpsF64(), and useDS128().
◆ DumpCode
|
protected |
Definition at line 100 of file GCNSubtarget.h.
Referenced by dumpCode().
◆ DynamicVGPR
|
protected |
Definition at line 205 of file GCNSubtarget.h.
Referenced by isDynamicVGPREnabled().
◆ DynamicVGPRBlockSize32
|
protected |
Definition at line 206 of file GCNSubtarget.h.
Referenced by getDynamicVGPRBlockSize().
◆ EnableCuMode
|
protected |
Definition at line 90 of file GCNSubtarget.h.
Referenced by hasLDSMisalignedBug(), and isCuModeEnabled().
◆ EnableDS128
|
protected |
Definition at line 98 of file GCNSubtarget.h.
Referenced by useDS128().
◆ EnableFlatScratch
|
protected |
Definition at line 230 of file GCNSubtarget.h.
Referenced by enableFlatScratch().
◆ EnableLoadStoreOpt
|
protected |
Definition at line 95 of file GCNSubtarget.h.
Referenced by loadStoreOptEnabled().
◆ EnablePreciseMemory
|
protected |
Definition at line 92 of file GCNSubtarget.h.
Referenced by isPreciseMemoryEnabled().
◆ EnablePRTStrictNull
|
protected |
Definition at line 99 of file GCNSubtarget.h.
Referenced by usePRTStrictNull().
◆ EnableSIScheduler
|
protected |
Definition at line 97 of file GCNSubtarget.h.
Referenced by enableSIScheduler().
◆ EnableSRAMECC
|
protected |
Definition at line 212 of file GCNSubtarget.h.
◆ EnableTgSplit
|
protected |
Definition at line 89 of file GCNSubtarget.h.
Referenced by isTgSplitEnabled().
◆ EnableUnsafeDSOffsetFolding
|
protected |
Definition at line 96 of file GCNSubtarget.h.
Referenced by unsafeDSOffsetFoldingEnabled().
◆ EnableXNACK
|
protected |
Definition at line 87 of file GCNSubtarget.h.
◆ FastDenormalF32
|
protected |
Definition at line 70 of file GCNSubtarget.h.
◆ FeatureDisable
|
protected |
Definition at line 303 of file GCNSubtarget.h.
◆ FlatAddressSpace
|
protected |
Definition at line 223 of file GCNSubtarget.h.
Referenced by hasFlatAddressSpace().
◆ FlatForGlobal
|
protected |
Definition at line 75 of file GCNSubtarget.h.
Referenced by initializeSubtargetDependencies(), and useFlatForGlobal().
◆ FlatGlobalInsts
|
protected |
Definition at line 225 of file GCNSubtarget.h.
Referenced by hasFlatGlobalInsts().
◆ FlatGVSMode
|
protected |
Definition at line 227 of file GCNSubtarget.h.
Referenced by hasFlatGVSMode().
◆ FlatInstOffsets
|
protected |
Definition at line 224 of file GCNSubtarget.h.
Referenced by hasFlatInstOffsets().
◆ FlatScratchInsts
|
protected |
Definition at line 226 of file GCNSubtarget.h.
Referenced by hasFlatScratchInsts().
◆ FMA
|
protected |
Definition at line 105 of file GCNSubtarget.h.
Referenced by hasFMA().
◆ FP64
|
protected |
Definition at line 104 of file GCNSubtarget.h.
Referenced by hasFP64(), and hasHWFP64().
◆ FullRate64Ops
|
protected |
Definition at line 72 of file GCNSubtarget.h.
Referenced by hasFullRate64Ops().
◆ Gen
Definition at line 64 of file GCNSubtarget.h.
Referenced by getGeneration(), and initializeSubtargetDependencies().
◆ GFX10_3Insts
|
protected |
Definition at line 117 of file GCNSubtarget.h.
Referenced by hasGFX10_3Insts().
◆ GFX10_AEncoding
|
protected |
Definition at line 149 of file GCNSubtarget.h.
Referenced by hasGFX10_AEncoding().
◆ GFX10_BEncoding
|
protected |
Definition at line 150 of file GCNSubtarget.h.
Referenced by hasAtomicCSub(), hasGFX10_BEncoding(), and hasGlobalAddTidInsts().
◆ GFX10Insts
|
protected |
Definition at line 113 of file GCNSubtarget.h.
◆ GFX11Insts
|
protected |
Definition at line 114 of file GCNSubtarget.h.
Referenced by hasCompressedExport(), hasDelayAlu(), hasDOTOpSelHazard(), hasFlatScratchSVSMode(), hasNullExportTarget(), hasSPackHL(), and hasVINTERPEncoding().
◆ GFX1250Insts
|
protected |
Definition at line 116 of file GCNSubtarget.h.
Referenced by hasAddPC64Inst(), hasGetPCZeroExtension(), hasGFX1250Insts(), hasIntMinMax64(), hasLDSFPAtomicAddF64(), hasMadU64U32NoCarry(), hasMovB64(), hasPKF32InstsReplicatingLower32BitsOfScalarInput(), hasPkMinMax3Insts(), hasScaleOffset(), hasScratchBaseForwardingHazard(), hasSGetShaderCyclesInst(), hasSignedGVSOffset(), hasVALUReadSGPRHazard(), hasVectorMulU64(), hasVOPD3(), needsKernArgPreloadProlog(), requiresNopBeforeDeallocVGPRs(), requiresWaitIdleBeforeGetReg(), requiresWaitXCntBeforeAtomicStores(), setRegModeNeedsVNOPs(), and supportsWGP().
◆ GFX12Insts
|
protected |
Definition at line 115 of file GCNSubtarget.h.
Referenced by hasGetPCZeroExtension(), hasPrefetch(), and hasVALUReadSGPRHazard().
◆ GFX7GFX8GFX9Insts
|
protected |
Definition at line 118 of file GCNSubtarget.h.
◆ GFX8Insts
|
protected |
Definition at line 108 of file GCNSubtarget.h.
Referenced by hasLDSFPAtomicAddF32().
◆ GFX90AInsts
|
protected |
Definition at line 110 of file GCNSubtarget.h.
Referenced by hasGFX90AInsts(), hasLDSFPAtomicAddF64(), hasPkMovB32(), and hasShift64HighRegBug().
◆ GFX940Insts
|
protected |
Definition at line 111 of file GCNSubtarget.h.
Referenced by hasDOTOpSelHazard(), hasDstSelForwardingHazard(), hasFlatScratchSVSMode(), hasGFX940Insts(), hasMovB64(), hasShift64HighRegBug(), hasTransForwardingHazard(), and hasVDecCoExecHazard().
◆ GFX950Insts
|
protected |
Definition at line 112 of file GCNSubtarget.h.
Referenced by hasCvtScaleForwardingHazard(), and hasGFX950Insts().
◆ GFX9Insts
|
protected |
Definition at line 109 of file GCNSubtarget.h.
Referenced by hasScalarMulHiInsts(), hasScalarPackInsts(), and hasSwap().
◆ HalfRate64Ops
|
protected |
Definition at line 71 of file GCNSubtarget.h.
Referenced by hasHalfRate64Ops().
◆ Has1024AddressableVGPRs
|
protected |
Definition at line 246 of file GCNSubtarget.h.
Referenced by has1024AddressableVGPRs().
◆ Has1_5xVGPRs
|
protected |
Definition at line 272 of file GCNSubtarget.h.
Referenced by has1_5xVGPRs().
◆ Has45BitNumRecordsBufferResource
|
protected |
Definition at line 297 of file GCNSubtarget.h.
Referenced by has45BitNumRecordsBufferResource().
◆ Has64BitLiterals
|
protected |
Definition at line 245 of file GCNSubtarget.h.
Referenced by has64BitLiterals().
◆ HasA16
|
protected |
Definition at line 145 of file GCNSubtarget.h.
Referenced by hasA16().
◆ HasAddMinMaxInsts
|
protected |
Definition at line 287 of file GCNSubtarget.h.
Referenced by hasAddMinMaxInsts().
◆ HasAddSubU64Insts
|
protected |
Definition at line 285 of file GCNSubtarget.h.
Referenced by hasAddSubU64Insts().
◆ HasAgentScopeFineGrainedRemoteMemoryAtomics
|
protected |
Definition at line 196 of file GCNSubtarget.h.
Referenced by supportsAgentScopeFineGrainedRemoteMemoryAtomics().
◆ HasApertureRegs
|
protected |
Definition at line 81 of file GCNSubtarget.h.
Referenced by hasApertureRegs().
◆ HasArchitectedFlatScratch
|
protected |
Definition at line 229 of file GCNSubtarget.h.
Referenced by flatScratchIsArchitected(), and getBaseReservedNumSGPRs().
◆ HasArchitectedSGPRs
|
protected |
Definition at line 231 of file GCNSubtarget.h.
Referenced by hasArchitectedSGPRs().
◆ HasAshrPkInsts
|
protected |
Definition at line 278 of file GCNSubtarget.h.
Referenced by hasAshrPkInsts().
◆ HasAtomicBufferGlobalPkAddF16Insts
|
protected |
Definition at line 189 of file GCNSubtarget.h.
Referenced by hasAtomicBufferGlobalPkAddF16Insts().
◆ HasAtomicBufferGlobalPkAddF16NoRtnInsts
|
protected |
Definition at line 188 of file GCNSubtarget.h.
Referenced by hasAtomicBufferGlobalPkAddF16NoRtnInsts().
◆ HasAtomicBufferPkAddBF16Inst
|
protected |
Definition at line 192 of file GCNSubtarget.h.
Referenced by hasAtomicBufferPkAddBF16Inst().
◆ HasAtomicCSubNoRtnInsts
|
protected |
Definition at line 190 of file GCNSubtarget.h.
Referenced by hasAtomicCSubNoRtnInsts().
◆ HasAtomicDsPkAdd16Insts
|
protected |
Definition at line 183 of file GCNSubtarget.h.
Referenced by hasAtomicDsPkAdd16Insts().
◆ HasAtomicFaddNoRtnInsts
|
protected |
Definition at line 186 of file GCNSubtarget.h.
Referenced by hasAtomicFaddInsts(), and hasAtomicFaddNoRtnInsts().
◆ HasAtomicFaddRtnInsts
|
protected |
Definition at line 185 of file GCNSubtarget.h.
Referenced by hasAtomicFaddInsts(), and hasAtomicFaddRtnInsts().
◆ HasAtomicFlatPkAdd16Insts
|
protected |
Definition at line 184 of file GCNSubtarget.h.
Referenced by hasAtomicFlatPkAdd16Insts().
◆ HasAtomicFMinFMaxF32FlatInsts
|
protected |
Definition at line 181 of file GCNSubtarget.h.
Referenced by hasAtomicFMinFMaxF32FlatInsts().
◆ HasAtomicFMinFMaxF32GlobalInsts
|
protected |
Definition at line 179 of file GCNSubtarget.h.
Referenced by hasAtomicFMinFMaxF32GlobalInsts().
◆ HasAtomicFMinFMaxF64FlatInsts
|
protected |
Definition at line 182 of file GCNSubtarget.h.
Referenced by hasAtomicFMinFMaxF64FlatInsts().
◆ HasAtomicFMinFMaxF64GlobalInsts
|
protected |
Definition at line 180 of file GCNSubtarget.h.
Referenced by hasAtomicFMinFMaxF64GlobalInsts().
◆ HasAtomicGlobalPkAddBF16Inst
|
protected |
Definition at line 191 of file GCNSubtarget.h.
Referenced by hasAtomicGlobalPkAddBF16Inst().
◆ HasBitOp3Insts
|
protected |
Definition at line 247 of file GCNSubtarget.h.
Referenced by hasBitOp3Insts().
◆ HasBVHDualAndBVH8Insts
|
protected |
Definition at line 252 of file GCNSubtarget.h.
Referenced by hasBVHDualAndBVH8Insts().
◆ HasClusters
|
protected |
Definition at line 299 of file GCNSubtarget.h.
Referenced by hasClusters().
◆ HasCubeInsts
|
protected |
Definition at line 169 of file GCNSubtarget.h.
Referenced by hasCubeInsts().
◆ HasCvtFP8Vop1Bug
|
protected |
Definition at line 177 of file GCNSubtarget.h.
Referenced by hasCvtFP8VOP1Bug().
◆ HasCvtNormInsts
|
protected |
Definition at line 173 of file GCNSubtarget.h.
Referenced by hasCvtNormInsts().
◆ HasCvtPkNormVOP2Insts
|
protected |
Definition at line 174 of file GCNSubtarget.h.
Referenced by hasCvtPkNormVOP2Insts().
◆ HasCvtPkNormVOP3Insts
|
protected |
Definition at line 175 of file GCNSubtarget.h.
Referenced by hasCvtPkNormVOP3Insts().
◆ HasDefaultComponentBroadcast
|
protected |
Definition at line 198 of file GCNSubtarget.h.
Referenced by hasDefaultComponentBroadcast().
◆ HasDefaultComponentZero
|
protected |
Definition at line 195 of file GCNSubtarget.h.
Referenced by hasDefaultComponentZero().
◆ HasDLInsts
|
protected |
Definition at line 151 of file GCNSubtarget.h.
Referenced by hasDLInsts().
◆ HasDot10Insts
|
protected |
Definition at line 162 of file GCNSubtarget.h.
Referenced by hasDot10Insts().
◆ HasDot11Insts
|
protected |
Definition at line 163 of file GCNSubtarget.h.
Referenced by hasDot11Insts().
◆ HasDot12Insts
|
protected |
Definition at line 164 of file GCNSubtarget.h.
Referenced by hasDot12Insts().
◆ HasDot13Insts
|
protected |
Definition at line 165 of file GCNSubtarget.h.
Referenced by hasDot13Insts().
◆ HasDot1Insts
|
protected |
Definition at line 153 of file GCNSubtarget.h.
Referenced by hasDot1Insts().
◆ HasDot2Insts
|
protected |
Definition at line 154 of file GCNSubtarget.h.
Referenced by hasDot2Insts().
◆ HasDot3Insts
|
protected |
Definition at line 155 of file GCNSubtarget.h.
Referenced by hasDot3Insts().
◆ HasDot4Insts
|
protected |
Definition at line 156 of file GCNSubtarget.h.
Referenced by hasDot4Insts().
◆ HasDot5Insts
|
protected |
Definition at line 157 of file GCNSubtarget.h.
Referenced by hasDot5Insts().
◆ HasDot6Insts
|
protected |
Definition at line 158 of file GCNSubtarget.h.
Referenced by hasDot6Insts().
◆ HasDot7Insts
|
protected |
Definition at line 159 of file GCNSubtarget.h.
Referenced by hasDot7Insts().
◆ HasDot8Insts
|
protected |
Definition at line 160 of file GCNSubtarget.h.
Referenced by hasDot8Insts().
◆ HasDot9Insts
|
protected |
Definition at line 161 of file GCNSubtarget.h.
Referenced by hasDot9Insts().
◆ HasDPALU_DPP
|
protected |
Definition at line 139 of file GCNSubtarget.h.
Referenced by hasDPALU_DPP().
◆ HasDPP
|
protected |
Definition at line 137 of file GCNSubtarget.h.
Referenced by hasDPP(), hasDPPBroadcasts(), and hasDPPWavefrontShifts().
◆ HasDPP8
|
protected |
Definition at line 138 of file GCNSubtarget.h.
Referenced by hasDPP8().
◆ HasDPPSrc1SGPR
|
protected |
Definition at line 140 of file GCNSubtarget.h.
Referenced by hasDPPSrc1SGPR().
◆ HasEmulatedSystemScopeAtomics
|
protected |
Definition at line 197 of file GCNSubtarget.h.
Referenced by hasEmulatedSystemScopeAtomics().
◆ HasExtendedImageInsts
|
protected |
Definition at line 143 of file GCNSubtarget.h.
Referenced by hasExtendedImageInsts().
◆ HasFlatAtomicFaddF32Inst
|
protected |
Definition at line 193 of file GCNSubtarget.h.
Referenced by hasFlatAtomicFaddF32Inst().
◆ HasFlatBufferGlobalAtomicFaddF64Inst
|
protected |
Definition at line 194 of file GCNSubtarget.h.
Referenced by hasFlatBufferGlobalAtomicFaddF64Inst().
◆ HasFlatSegmentOffsetBug
|
protected |
Definition at line 267 of file GCNSubtarget.h.
Referenced by hasFlatSegmentOffsetBug().
◆ HasFmacF64Inst
|
protected |
Definition at line 152 of file GCNSubtarget.h.
Referenced by hasFmacF64Inst().
◆ HasFmaMixBF16Insts
|
protected |
Definition at line 126 of file GCNSubtarget.h.
Referenced by hasFmaMixBF16Insts().
◆ HasFmaMixInsts
|
protected |
Definition at line 125 of file GCNSubtarget.h.
Referenced by hasFmaMixInsts().
◆ HasFP8ConversionInsts
|
protected |
Definition at line 168 of file GCNSubtarget.h.
Referenced by hasFP8ConversionInsts().
◆ HasFP8E5M3Insts
|
protected |
Definition at line 176 of file GCNSubtarget.h.
Referenced by hasFP8E5M3Insts().
◆ HasFP8Insts
|
protected |
Definition at line 167 of file GCNSubtarget.h.
Referenced by hasFP8Insts().
◆ HasG16
|
protected |
Definition at line 146 of file GCNSubtarget.h.
Referenced by hasG16().
◆ HasGDS
|
protected |
Definition at line 232 of file GCNSubtarget.h.
Referenced by hasGDS().
◆ HasGetWaveIdInst
|
protected |
Definition at line 217 of file GCNSubtarget.h.
Referenced by hasGetWaveIdInst().
◆ HasGloballyAddressableScratch
|
protected |
Definition at line 295 of file GCNSubtarget.h.
Referenced by hasGloballyAddressableScratch().
◆ HasGWS
|
protected |
Definition at line 233 of file GCNSubtarget.h.
Referenced by hasGWS().
◆ HasIEEEMinimumMaximumInsts
|
protected |
Definition at line 279 of file GCNSubtarget.h.
Referenced by hasIEEEMinimumMaximumInsts().
◆ HasImageGather4D16Bug
|
protected |
Definition at line 269 of file GCNSubtarget.h.
Referenced by hasImageGather4D16Bug().
◆ HasImageInsts
|
protected |
Definition at line 142 of file GCNSubtarget.h.
Referenced by hasImageInsts().
◆ HasImageStoreD16Bug
|
protected |
Definition at line 268 of file GCNSubtarget.h.
Referenced by hasImageStoreD16Bug().
◆ HasInstFwdPrefetchBug
|
protected |
Definition at line 258 of file GCNSubtarget.h.
Referenced by hasInstFwdPrefetchBug().
◆ HasIntClamp
|
protected |
Definition at line 124 of file GCNSubtarget.h.
Referenced by hasIntClamp().
◆ HasLdsBarrierArriveAtomic
|
protected |
Definition at line 290 of file GCNSubtarget.h.
Referenced by hasLdsBarrierArriveAtomic().
◆ HasLdsBranchVmemWARHazard
|
protected |
Definition at line 263 of file GCNSubtarget.h.
Referenced by hasLdsBranchVmemWARHazard().
◆ HasLerpInst
|
protected |
Definition at line 170 of file GCNSubtarget.h.
Referenced by hasLerpInst().
◆ HasLshlAddU64Inst
|
protected |
Definition at line 284 of file GCNSubtarget.h.
Referenced by hasLshlAddU64Inst().
◆ HasMADIntraFwdBug
|
protected |
Definition at line 273 of file GCNSubtarget.h.
Referenced by hasMADIntraFwdBug().
◆ HasMadU32Inst
|
protected |
Definition at line 286 of file GCNSubtarget.h.
Referenced by hasMadU32Inst().
◆ HasMAIInsts
|
protected |
Definition at line 166 of file GCNSubtarget.h.
Referenced by hasMAIInsts().
◆ HasMemoryAtomicFaddF32DenormalSupport
|
protected |
Definition at line 187 of file GCNSubtarget.h.
Referenced by hasMemoryAtomicFaddF32DenormalSupport().
◆ HasMFMAInlineLiteralBug
|
protected |
Definition at line 237 of file GCNSubtarget.h.
Referenced by hasMFMAInlineLiteralBug().
◆ HasMin3Max3PKF16
|
protected |
Definition at line 282 of file GCNSubtarget.h.
Referenced by hasMin3Max3PKF16().
◆ HasMinimum3Maximum3F16
|
protected |
Definition at line 281 of file GCNSubtarget.h.
Referenced by hasMinimum3Maximum3F16().
◆ HasMinimum3Maximum3F32
|
protected |
Definition at line 280 of file GCNSubtarget.h.
Referenced by hasMinimum3Maximum3F32().
◆ HasMinimum3Maximum3PKF16
|
protected |
Definition at line 283 of file GCNSubtarget.h.
Referenced by hasMinimum3Maximum3PKF16().
◆ HasMovrel
|
protected |
Definition at line 127 of file GCNSubtarget.h.
Referenced by hasMovrel().
◆ HasMSAALoadDstSelBug
|
protected |
Definition at line 270 of file GCNSubtarget.h.
Referenced by hasMSAALoadDstSelBug().
◆ HasNoDataDepHazard
|
protected |
Definition at line 222 of file GCNSubtarget.h.
Referenced by hasNoDataDepHazard().
◆ HasNoSdstCMPX
|
protected |
Definition at line 214 of file GCNSubtarget.h.
Referenced by hasNoSdstCMPX().
◆ HasNSAClauseBug
|
protected |
Definition at line 265 of file GCNSubtarget.h.
Referenced by hasNSAClauseBug().
◆ HasNSAEncoding
|
protected |
Definition at line 147 of file GCNSubtarget.h.
Referenced by hasNSAEncoding().
◆ HasNSAtoVMEMBug
|
protected |
Definition at line 264 of file GCNSubtarget.h.
Referenced by hasNSAtoVMEMBug().
◆ HasOffset3fBug
|
protected |
Definition at line 266 of file GCNSubtarget.h.
Referenced by hasOffset3fBug().
◆ HasPackedFP32Ops
|
protected |
Definition at line 141 of file GCNSubtarget.h.
Referenced by hasPackedFP32Ops().
◆ HasPackedTID
|
protected |
Definition at line 240 of file GCNSubtarget.h.
Referenced by hasPackedTID().
◆ HasPartialNSAEncoding
|
protected |
Definition at line 148 of file GCNSubtarget.h.
Referenced by hasPartialNSAEncoding().
◆ HasPermlane16Swap
|
protected |
Definition at line 253 of file GCNSubtarget.h.
Referenced by hasPermlane16Swap().
◆ HasPermlane32Swap
|
protected |
Definition at line 254 of file GCNSubtarget.h.
Referenced by hasPermlane32Swap().
◆ HasPkAddMinMaxInsts
|
protected |
Definition at line 288 of file GCNSubtarget.h.
Referenced by hasPkAddMinMaxInsts().
◆ HasPkFmacF16Inst
|
protected |
Definition at line 178 of file GCNSubtarget.h.
Referenced by hasPkFmacF16Inst().
◆ HasPointSampleAccel
|
protected |
Definition at line 289 of file GCNSubtarget.h.
Referenced by hasPointSampleAccel().
◆ HasPrivEnabledTrap2NopBug
|
protected |
Definition at line 271 of file GCNSubtarget.h.
Referenced by hasPrivEnabledTrap2NopBug().
◆ HasPrngInst
|
protected |
Definition at line 251 of file GCNSubtarget.h.
Referenced by hasPrngInst().
◆ HasPseudoScalarTrans
|
protected |
Definition at line 243 of file GCNSubtarget.h.
Referenced by hasPseudoScalarTrans().
◆ HasQsadInsts
|
protected |
Definition at line 172 of file GCNSubtarget.h.
Referenced by hasQsadInsts().
◆ HasR128A16
|
protected |
Definition at line 144 of file GCNSubtarget.h.
Referenced by hasR128A16().
◆ HasRequiredExportPriority
|
protected |
Definition at line 276 of file GCNSubtarget.h.
Referenced by hasRequiredExportPriority().
◆ HasRestrictedSOffset
|
protected |
Definition at line 244 of file GCNSubtarget.h.
Referenced by hasRestrictedSOffset().
◆ HasSadInsts
|
protected |
Definition at line 171 of file GCNSubtarget.h.
Referenced by hasSadInsts().
◆ HasSafeCUPrefetch
|
protected |
Definition at line 261 of file GCNSubtarget.h.
Referenced by hasSafeCUPrefetch().
◆ HasSafeSmemPrefetch
|
protected |
Definition at line 260 of file GCNSubtarget.h.
Referenced by hasSafeSmemPrefetch().
◆ HasSALUFloatInsts
|
protected |
Definition at line 242 of file GCNSubtarget.h.
Referenced by hasSALUFloatInsts().
◆ HasScalarAtomics
|
protected |
Definition at line 131 of file GCNSubtarget.h.
Referenced by hasScalarAtomics().
◆ HasScalarDwordx3Loads
|
protected |
Definition at line 129 of file GCNSubtarget.h.
Referenced by hasScalarDwordx3Loads().
◆ HasScalarStores
|
protected |
Definition at line 130 of file GCNSubtarget.h.
Referenced by hasScalarStores().
◆ HasSDWAMac
|
protected |
Definition at line 135 of file GCNSubtarget.h.
Referenced by hasSDWAMac().
◆ HasSDWAOmod
|
protected |
Definition at line 132 of file GCNSubtarget.h.
Referenced by hasSDWAOmod().
◆ HasSDWAOutModsVOPC
|
protected |
Definition at line 136 of file GCNSubtarget.h.
Referenced by hasSDWAOutModsVOPC().
◆ HasSDWAScalar
|
protected |
Definition at line 133 of file GCNSubtarget.h.
Referenced by hasSDWAScalar().
◆ HasSDWASdst
|
protected |
Definition at line 134 of file GCNSubtarget.h.
Referenced by hasSDWASdst().
◆ HasSetPrioIncWgInst
|
protected |
Definition at line 291 of file GCNSubtarget.h.
Referenced by hasSetPrioIncWgInst().
◆ HasShaderCyclesHiLoRegisters
|
protected |
Definition at line 220 of file GCNSubtarget.h.
Referenced by hasShaderCyclesHiLoRegisters().
◆ HasShaderCyclesRegister
|
protected |
Definition at line 219 of file GCNSubtarget.h.
Referenced by hasShaderCyclesRegister().
◆ HasSMemRealTime
|
protected |
Definition at line 123 of file GCNSubtarget.h.
Referenced by hasSMemRealTime().
◆ HasSMemTimeInst
|
protected |
Definition at line 218 of file GCNSubtarget.h.
Referenced by hasSMemTimeInst().
◆ HasSMEMtoVectorWriteHazard
|
protected |
Definition at line 257 of file GCNSubtarget.h.
Referenced by hasSMEMtoVectorWriteHazard().
◆ HasTanhInsts
|
protected |
Definition at line 248 of file GCNSubtarget.h.
Referenced by hasTanhInsts().
◆ HasTensorCvtLutInsts
|
protected |
Definition at line 249 of file GCNSubtarget.h.
Referenced by hasTensorCvtLutInsts().
◆ HasTransposeLoadF4F6Insts
|
protected |
Definition at line 250 of file GCNSubtarget.h.
Referenced by hasTransposeLoadF4F6Insts().
◆ HasUnpackedD16VMem
|
protected |
Definition at line 235 of file GCNSubtarget.h.
Referenced by hasUnpackedD16VMem().
◆ HasVALUTransUseHazard
|
protected |
Definition at line 275 of file GCNSubtarget.h.
Referenced by hasVALUTransUseHazard().
◆ HasVcmpxExecWARHazard
|
protected |
Definition at line 262 of file GCNSubtarget.h.
Referenced by hasVcmpxExecWARHazard().
◆ HasVcmpxPermlaneHazard
|
protected |
Definition at line 255 of file GCNSubtarget.h.
Referenced by hasVcmpxPermlaneHazard().
◆ HasVGPRIndexMode
|
protected |
Definition at line 128 of file GCNSubtarget.h.
Referenced by hasVGPRIndexMode().
◆ HasVmemPrefInsts
|
protected |
Definition at line 259 of file GCNSubtarget.h.
Referenced by hasVmemPrefInsts().
◆ HasVMemToLDSLoad
|
protected |
Definition at line 207 of file GCNSubtarget.h.
Referenced by hasVMemToLDSLoad().
◆ HasVMEMtoScalarWriteHazard
|
protected |
Definition at line 256 of file GCNSubtarget.h.
Referenced by hasVMEMtoScalarWriteHazard().
◆ HasVmemWriteVgprInOrder
|
protected |
Definition at line 277 of file GCNSubtarget.h.
Referenced by hasVmemWriteVgprInOrder().
◆ HasVOP3Literal
|
protected |
Definition at line 221 of file GCNSubtarget.h.
Referenced by hasVOP3Literal().
◆ HasVOPDInsts
|
protected |
Definition at line 274 of file GCNSubtarget.h.
Referenced by hasVOPDInsts().
◆ HasVscnt
|
protected |
Definition at line 215 of file GCNSubtarget.h.
Referenced by hasVscnt().
◆ HasWaitXcnt
|
protected |
Definition at line 216 of file GCNSubtarget.h.
Referenced by hasWaitXCnt().
◆ HasXF32Insts
|
protected |
Definition at line 199 of file GCNSubtarget.h.
Referenced by hasXF32Insts().
◆ InstrItins
|
protected |
Definition at line 65 of file GCNSubtarget.h.
Referenced by GCNSubtarget(), and getInstrItineraryData().
◆ KernargPreload
|
protected |
Definition at line 83 of file GCNSubtarget.h.
Referenced by hasKernargPreload().
◆ LDSBankCount
|
protected |
Definition at line 66 of file GCNSubtarget.h.
Referenced by getLDSBankCount(), and initializeSubtargetDependencies().
◆ LDSMisalignedBug
|
protected |
Definition at line 236 of file GCNSubtarget.h.
Referenced by hasLDSMisalignedBug().
◆ MaxHardClauseLength
|
protected |
The maximum number of instructions that may be placed within an S_CLAUSE, which is one greater than the maximum argument to S_CLAUSE.
A value of 0 indicates a lack of S_CLAUSE support.
Definition at line 203 of file GCNSubtarget.h.
Referenced by hasHardClauses(), and maxHardClauseLength().
◆ MaxPrivateElementSize
|
protected |
Definition at line 67 of file GCNSubtarget.h.
Referenced by getMaxPrivateElementSize(), and initializeSubtargetDependencies().
◆ MIMG_R128
|
protected |
Definition at line 106 of file GCNSubtarget.h.
Referenced by hasMIMG_R128().
◆ NegativeScratchOffsetBug
|
protected |
Definition at line 121 of file GCNSubtarget.h.
Referenced by hasNegativeScratchOffsetBug().
◆ NegativeUnalignedScratchOffsetBug
|
protected |
Definition at line 122 of file GCNSubtarget.h.
Referenced by hasNegativeUnalignedScratchOffsetBug().
◆ RelaxedBufferOOBMode
|
protected |
Definition at line 80 of file GCNSubtarget.h.
Referenced by hasRelaxedBufferOOBMode().
◆ RequiresAlignVGPR
|
protected |
Definition at line 208 of file GCNSubtarget.h.
Referenced by needsAlignedVGPRs().
◆ RequiresCOV6
|
protected |
Definition at line 293 of file GCNSubtarget.h.
Referenced by requiresCodeObjectV6().
◆ RequiresWaitsBeforeSystemScopeStores
|
protected |
Definition at line 300 of file GCNSubtarget.h.
Referenced by requiresWaitsBeforeSystemScopeStores().
◆ ScalarFlatScratchInsts
|
protected |
Definition at line 228 of file GCNSubtarget.h.
Referenced by hasScalarFlatScratchInsts().
◆ ScalarizeGlobal
|
protected |
Definition at line 241 of file GCNSubtarget.h.
Referenced by getScalarizeGlobalBehavior(), and setScalarizeGlobalBehavior().
◆ SGPRInitBug
|
protected |
Definition at line 119 of file GCNSubtarget.h.
Referenced by hasSGPRInitBug().
◆ SupportsSRAMECC
|
protected |
Definition at line 204 of file GCNSubtarget.h.
◆ SupportsXNACK
|
protected |
Definition at line 82 of file GCNSubtarget.h.
◆ TargetID
|
protected |
Definition at line 63 of file GCNSubtarget.h.
Referenced by d16PreservesUnusedBits(), GCNSubtarget(), getTargetID(), initializeSubtargetDependencies(), and isXNACKEnabled().
◆ TrapHandler
|
protected |
Definition at line 91 of file GCNSubtarget.h.
Referenced by isTrapHandlerEnabled().
◆ UnalignedAccessMode
|
protected |
Definition at line 79 of file GCNSubtarget.h.
Referenced by hasUnalignedAccessMode(), hasUnalignedBufferAccessEnabled(), hasUnalignedDSAccessEnabled(), and hasUnalignedScratchAccessEnabled().
◆ UnalignedBufferAccess
|
protected |
Definition at line 238 of file GCNSubtarget.h.
Referenced by hasUnalignedBufferAccess(), and hasUnalignedBufferAccessEnabled().
◆ UnalignedDSAccess
|
protected |
Definition at line 239 of file GCNSubtarget.h.
Referenced by hasUnalignedDSAccess(), and hasUnalignedDSAccessEnabled().
◆ UnalignedScratchAccess
|
protected |
Definition at line 78 of file GCNSubtarget.h.
Referenced by hasUnalignedScratchAccess(), and hasUnalignedScratchAccessEnabled().
◆ UseBlockVGPROpsForCSR
|
protected |
Definition at line 294 of file GCNSubtarget.h.
Referenced by useVGPRBlockOpsForCSR().
◆ UserSGPRInit16Bug
|
protected |
Definition at line 120 of file GCNSubtarget.h.
Referenced by hasUserSGPRInit16Bug().
The documentation for this class was generated from the following files:
- lib/Target/AMDGPU/GCNSubtarget.h
- lib/Target/AMDGPU/GCNSubtarget.cpp
Generated on for LLVM by
1.14.0