Inheritance diagram for llvm::abi::AMDGPUTargetInfo:

[legend]

Public Member Functions
	AMDGPUTargetInfo (TypeBuilder &TypeBuilder, const ABICompatInfo &Compat, bool CoerceGenericPtrArgToGlobal)
const ABICompatInfo &	getABICompatInfo () const override
	Return this target's ABI compatibility flags.
unsigned	getAllocaAddrSpace () const override
	Indirect arguments live in the private (alloca) address space on AMDGPU.
void	computeInfo (FunctionInfo &FI) const override
	Populate FI with the target's ABI-lowering decisions for each argument and return value.
Public Member Functions inherited from llvm::abi::DefaultTargetInfo
ArgInfo	classifyArgumentType (const Type *Ty) const
ArgInfo	classifyReturnType (const Type *RetTy) const
void	computeInfo (FunctionInfo &FI) const override
	Populate FI with the target's ABI-lowering decisions for each argument and return value.
	TargetInfo (TypeBuilder &Builder)
Public Member Functions inherited from llvm::abi::TargetInfo
	TargetInfo (TypeBuilder &Builder)
virtual	~TargetInfo ()=default
virtual bool	isPassByRef (const Type *Ty) const
virtual bool	hasInt128Type () const
	Whether the target has a 128-bit integer type. Default true.
virtual unsigned	getLongLongWidth () const
	Width of long long in bits. Default 64.

Additional Inherited Members
Protected Member Functions inherited from llvm::abi::TargetInfo
LLVM_ABI RecordArgABI	getRecordArgABI (const RecordType *RT) const
LLVM_ABI RecordArgABI	getRecordArgABI (const Type *Ty) const
LLVM_ABI bool	isPromotableInteger (const IntegerType *IT) const
unsigned	getBitIntRegThreshold () const
	Bit width above which a _BitInt cannot stay in registers.
LLVM_ABI ArgInfo	getNaturalAlignIndirect (const Type *Ty, unsigned AddrSpace, bool ByVal=true) const
LLVM_ABI bool	isAggregateTypeForABI (const Type *Ty) const
LLVM_ABI const Type *	useFirstFieldIfTransparentUnion (const Type *Ty) const
	If Ty is a transparent union, return its first field type; otherwise return Ty unchanged.
LLVM_ABI const Type *	isSingleElementStruct (const Type *Ty) const
	Returns the scalar a single-element struct reduces to, else null.
LLVM_ABI bool	maybeCommonClassifyReturnType (FunctionInfo &FI) const
	Apply rules for classifying return types that are common to all targets.
virtual bool	isHomogeneousAggregateBaseType (const Type *Ty) const
	Return true if `Ty` is a valid base type for a homogeneous aggregate.
virtual bool	isHomogeneousAggregateSmallEnough (const Type *Base, uint64_t Members) const
	Return true if a homogeneous aggregate with `Members` copies of `Base` is small enough to be passed in registers for this ABI.
virtual bool	isZeroLengthBitfieldPermittedInHomogeneousAggregate () const
	Return true if zero-length bitfields should be ignored when deciding whether an aggregate is homogeneous.
virtual bool	isPermittedToBeHomogeneousAggregate (const RecordType *RT) const
	Return true if the C++ ABI permits `RT` to be a homogeneous aggregate.
LLVM_ABI bool	isHomogeneousAggregate (const Type Ty, const Type &Base, uint64_t &Members) const
	Return true if `Ty` is an ELFv2-style homogeneous aggregate.
Protected Attributes inherited from llvm::abi::TargetInfo
TypeBuilder &	TB

Detailed Description

Definition at line 24 of file AMDGPU.cpp.

Constructor & Destructor Documentation

◆ AMDGPUTargetInfo()

llvm::abi::AMDGPUTargetInfo::AMDGPUTargetInfo	(	TypeBuilder &	TypeBuilder,
		const ABICompatInfo &	Compat,
		bool	CoerceGenericPtrArgToGlobal )

inline

Definition at line 43 of file AMDGPU.cpp.

Member Function Documentation

◆ computeInfo()

void llvm::abi::AMDGPUTargetInfo::computeInfo ( FunctionInfo & FI ) const

overridevirtual

Populate FI with the target's ABI-lowering decisions for each argument and return value.

Implements llvm::abi::TargetInfo.

Definition at line 232 of file AMDGPU.cpp.

References llvm::CallingConv::AMDGPU_KERNEL, llvm::abi::FunctionInfo::arguments(), llvm::abi::FunctionInfo::getCallingConvention(), llvm::abi::FunctionInfo::getNumRequiredArgs(), llvm::abi::FunctionInfo::getReturnInfo(), llvm::abi::FunctionInfo::getReturnType(), and llvm::abi::TargetInfo::maybeCommonClassifyReturnType().

◆ getABICompatInfo()

const ABICompatInfo & llvm::abi::AMDGPUTargetInfo::getABICompatInfo ( ) const

inlineoverridevirtual

Return this target's ABI compatibility flags.

Targets with extra flags store a derived object and return that as an ABICompatInfo reference.

Implements llvm::abi::TargetInfo.

Definition at line 48 of file AMDGPU.cpp.

◆ getAllocaAddrSpace()

unsigned llvm::abi::AMDGPUTargetInfo::getAllocaAddrSpace ( ) const

inlineoverridevirtual

Indirect arguments live in the private (alloca) address space on AMDGPU.

Reimplemented from llvm::abi::TargetInfo.

Definition at line 51 of file AMDGPU.cpp.

References llvm::AMDGPUAS::PRIVATE_ADDRESS.

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

lib/ABI/Targets/AMDGPU.cpp

Public Member Functions