#include "AMDGPUInstrInfo.h"#include "AMDGPUTargetTransformInfo.h"#include "GCNSubtarget.h"#include "llvm/IR/IntrinsicsAMDGPU.h"#include "llvm/Transforms/InstCombine/InstCombiner.h"#include "InstCombineTables.inc"
Go to the source code of this file.
Macros | |
| #define | DEBUG_TYPE "AMDGPUtti" |
| #define | GET_AMDGPUImageDMaskIntrinsicTable_IMPL |
Functions | |
| static APFloat | fmed3AMDGCN (const APFloat &Src0, const APFloat &Src1, const APFloat &Src2) |
| static bool | canSafelyConvertTo16Bit (Value &V) |
| static Value * | convertTo16Bit (Value &V, InstCombiner::BuilderTy &Builder) |
| static Optional< Instruction * > | simplifyAMDGCNImageIntrinsic (const GCNSubtarget *ST, const AMDGPU::ImageDimIntrinsicInfo *ImageDimIntr, IntrinsicInst &II, InstCombiner &IC) |
| static Value * | simplifyAMDGCNMemoryIntrinsicDemanded (InstCombiner &IC, IntrinsicInst &II, APInt DemandedElts, int DMaskIdx=-1) |
| Implement SimplifyDemandedVectorElts for amdgcn buffer and image intrinsics. More... | |
Macro Definition Documentation
◆ DEBUG_TYPE
| #define DEBUG_TYPE "AMDGPUtti" |
Definition at line 25 of file AMDGPUInstCombineIntrinsic.cpp.
◆ GET_AMDGPUImageDMaskIntrinsicTable_IMPL
| #define GET_AMDGPUImageDMaskIntrinsicTable_IMPL |
Definition at line 33 of file AMDGPUInstCombineIntrinsic.cpp.
Function Documentation
◆ canSafelyConvertTo16Bit()
|
static |
Definition at line 61 of file AMDGPUInstCombineIntrinsic.cpp.
References llvm::APFloat::convert(), llvm::Value::getType(), llvm::APFloatBase::IEEEhalf(), llvm::Type::isHalfTy(), llvm::Type::isIntegerTy(), llvm::PatternMatch::m_FPExt(), llvm::PatternMatch::m_SExt(), llvm::PatternMatch::m_Value(), llvm::PatternMatch::m_ZExt(), llvm::PatternMatch::match(), and llvm::APFloatBase::rmTowardZero.
Referenced by simplifyAMDGCNImageIntrinsic().
◆ convertTo16Bit()
|
static |
Definition at line 88 of file AMDGPUInstCombineIntrinsic.cpp.
References Builder, llvm::Value::getContext(), llvm::Type::getHalfTy(), llvm::Type::getInt16Ty(), llvm::Value::getType(), llvm::Type::isFloatingPointTy(), llvm::Type::isIntegerTy(), and llvm_unreachable.
Referenced by simplifyAMDGCNImageIntrinsic().
◆ fmed3AMDGCN()
|
static |
Definition at line 42 of file AMDGPUInstCombineIntrinsic.cpp.
References assert(), llvm::APFloatBase::cmpEqual, llvm::APFloatBase::cmpUnordered, llvm::APFloat::compare(), and llvm::maxnum().
Referenced by llvm::GCNTTIImpl::instCombineIntrinsic().
◆ simplifyAMDGCNImageIntrinsic()
|
static |
Definition at line 101 of file AMDGPUInstCombineIntrinsic.cpp.
References llvm::AMDGPU::HSAMD::Kernel::Key::Args, llvm::CallBase::args(), assert(), llvm::InstCombiner::Builder, canSafelyConvertTo16Bit(), convertTo16Bit(), llvm::AMDGPU::ImageDimIntrinsicInfo::CoordStart, llvm::AMDGPU::ImageDimIntrinsicInfo::CoordTyArg, llvm::Instruction::copyFastMathFlags(), llvm::Instruction::copyMetadata(), llvm::IRBuilderBase::CreateCall(), llvm::CallBase::getCalledFunction(), llvm::Value::getContext(), llvm::Intrinsic::getDeclaration(), llvm::Type::getHalfTy(), llvm::Type::getInt16Ty(), llvm::IntrinsicInst::getIntrinsicID(), llvm::Intrinsic::getIntrinsicSignature(), llvm::Instruction::getModule(), llvm::User::getOperand(), llvm::Value::getType(), llvm::AMDGPU::ImageDimIntrinsicInfo::GradientStart, llvm::AMDGPU::ImageDimIntrinsicInfo::GradientTyArg, I, llvm::Type::isFloatingPointTy(), llvm::None, llvm::InstCombiner::replaceInstUsesWith(), llvm::ARM_MB::ST, llvm::Value::takeName(), and llvm::AMDGPU::ImageDimIntrinsicInfo::VAddrEnd.
Referenced by llvm::GCNTTIImpl::instCombineIntrinsic().
◆ simplifyAMDGCNMemoryIntrinsicDemanded()
|
static |
Implement SimplifyDemandedVectorElts for amdgcn buffer and image intrinsics.
Note: This only supports non-TFE/LWE image intrinsic calls; those have struct returns.
Definition at line 909 of file AMDGPUInstCombineIntrinsic.cpp.
References llvm::AMDGPU::HSAMD::Kernel::Key::Args, llvm::CallBase::args(), llvm::tgtok::Bit, llvm::InstCombiner::Builder, llvm::Instruction::copyMetadata(), llvm::countPopulation(), llvm::APInt::countPopulation(), llvm::APInt::countTrailingZeros(), llvm::IRBuilderBase::CreateAdd(), llvm::IRBuilderBase::CreateCall(), llvm::IRBuilderBase::CreateInsertElement(), llvm::IRBuilderBase::CreateShuffleVector(), llvm::ConstantInt::get(), llvm::FixedVectorType::get(), llvm::UndefValue::get(), llvm::APInt::getActiveBits(), llvm::CallBase::getArgOperand(), llvm::CallBase::getCalledFunction(), llvm::InstCombiner::getDataLayout(), llvm::Intrinsic::getDeclaration(), llvm::IntrinsicInst::getIntrinsicID(), llvm::Intrinsic::getIntrinsicSignature(), llvm::Instruction::getParent(), llvm::BasicBlock::getParent(), llvm::GlobalValue::getParent(), llvm::Type::getScalarType(), llvm::ConstantInt::getType(), llvm::Value::getType(), llvm::DataLayout::getTypeSizeInBits(), llvm::ConstantInt::getZExtValue(), llvm::APInt::isMask(), M, Offset, llvm::CallBase::setArgOperand(), llvm::IRBuilderBase::SetInsertPoint(), and llvm::Value::takeName().
Referenced by llvm::GCNTTIImpl::simplifyDemandedVectorEltsIntrinsic().
