#include <op_kernel_context.h>

Public Types
using	ArgMap = std::unordered_map< std::string, size_t >

Public Member Functions
	OpKernelContext (_Inout_ IExecutionFrame frame, _In_ const OpKernel kernel, _In_ Stream stream, _In_opt_ concurrency::ThreadPool threadpool, _In_ const logging::Logger &logger)

virtual	~OpKernelContext ()=default

virtual int	NumVariadicInputs (size_t arg_num) const

virtual MLDataType	InputType (int index) const

virtual MLDataType	OutputType (int index) const

const OrtValue *	GetInputOrtValue (int index) const

template<typename T >
const T *	Input (int index) const

template<typename T >
const T &	RequiredInput (int index) const

template<typename T >
T *	Output (int index)

Tensor *	Output (int index, const TensorShape &shape)

Tensor *	Output (int index, const std::vector< int64_t > &shape)

Tensor *	Output (int index, const std::initializer_list< int64_t > &shape)

Tensor &	RequiredOutput (int index, const TensorShape &shape)

SparseTensor *	OutputSparse (int index, const TensorShape &shape)

template<typename T >
void	OutputOptionalWithoutData (int index)

virtual bool	TryGetInferredInputShape (int index, TensorShape &shape) const

virtual bool	TryGetInferredOutputShape (int index, TensorShape &shape) const

const logging::Logger &	Logger () const

virtual int	InputCount () const

virtual int	ImplicitInputCount () const

virtual int	OutputCount () const

virtual Status	GetTempSpaceAllocator (AllocatorPtr *output) const

Status	GetTempSpaceCPUAllocator (AllocatorPtr *output) const

virtual int	GetDeviceId () const

virtual Stream *	GetComputeStream () const

const std::string &	GetOpDomain () const

const std::string &	GetOpType () const

const std::string &	GetNodeName () const

_Ret_maybenull_ onnxruntime::concurrency::ThreadPool *	GetOperatorThreadPool () const

virtual bool	GetUseDeterministicCompute () const

template<>
Tensor *	Output (int index)

template<>
SparseTensor *	Output (int index)

Protected Member Functions
	OpKernelContext (concurrency::ThreadPool threadpool, const logging::Logger &logger, Stream stream)

onnxruntime::NodeIndex	GetNodeIndex () const

virtual const OrtValue *	GetInputMLValue (int index) const

const OrtValue *	GetImplicitInputMLValue (int index) const

OrtValue *	GetOutputMLValue (int index)

virtual OrtValue *	OutputMLValue (int index, const TensorShape &shape)

virtual OrtValue *	GetOrCreateOutputMLValue (int index)

Detailed Description

Definition at line 11 of file op_kernel_context.h.

Member Typedef Documentation

using onnxruntime::OpKernelContext::ArgMap = std::unordered_map<std::string, size_t>

Definition at line 13 of file op_kernel_context.h.

Constructor & Destructor Documentation

onnxruntime::OpKernelContext::OpKernelContext	(	_Inout_ IExecutionFrame *	frame,
		_In_ const OpKernel *	kernel,
		_In_ Stream *	stream,
		_In_opt_ concurrency::ThreadPool *	threadpool,
		_In_ const logging::Logger &	logger
	)

virtual onnxruntime::OpKernelContext::~OpKernelContext ( )

virtualdefault

onnxruntime::OpKernelContext::OpKernelContext	(	concurrency::ThreadPool *	threadpool,
		const logging::Logger &	logger,
		Stream *	stream
	)

protected

Member Function Documentation

virtual Stream* onnxruntime::OpKernelContext::GetComputeStream ( ) const

inlinevirtual

Return the compute stream associated with the EP that the kernel is partitioned to. For EPs that do not have a compute stream (e.g. CPU EP), a nullptr is returned.

Definition at line 152 of file op_kernel_context.h.

virtual int onnxruntime::OpKernelContext::GetDeviceId ( ) const

inlinevirtual

Return the device id that current kernel runs on.

Definition at line 144 of file op_kernel_context.h.

const OrtValue* onnxruntime::OpKernelContext::GetImplicitInputMLValue ( int index ) const

protected

virtual const OrtValue* onnxruntime::OpKernelContext::GetInputMLValue ( int index ) const

protectedvirtual

const OrtValue* onnxruntime::OpKernelContext::GetInputOrtValue ( int index ) const

inline

Definition at line 31 of file op_kernel_context.h.

onnxruntime::NodeIndex onnxruntime::OpKernelContext::GetNodeIndex ( ) const

protected

const std::string& onnxruntime::OpKernelContext::GetNodeName ( ) const

Returns the node name of the underlying kernel

const std::string& onnxruntime::OpKernelContext::GetOpDomain ( ) const

Returns the opset domain of the underlying kernel

_Ret_maybenull_ onnxruntime::concurrency::ThreadPool* onnxruntime::OpKernelContext::GetOperatorThreadPool ( ) const

inline

Returns the intra-op threadpool, if available.

Definition at line 174 of file op_kernel_context.h.

const std::string& onnxruntime::OpKernelContext::GetOpType ( ) const

Returns the optype of the underlying kernel

virtual OrtValue* onnxruntime::OpKernelContext::GetOrCreateOutputMLValue ( int index )

protectedvirtual

OrtValue* onnxruntime::OpKernelContext::GetOutputMLValue ( int index )

protected

virtual Status onnxruntime::OpKernelContext::GetTempSpaceAllocator ( AllocatorPtr * output ) const

virtual

Return an allocator on device 0, with memtype of OrtMemTypeDefault.

Remarks: Use SafeInt when calculating the size of memory to allocate using AllocatorPtr->Alloc.

Status onnxruntime::OpKernelContext::GetTempSpaceCPUAllocator ( AllocatorPtr * output ) const

Return the allocator associated with the CPU EP with memtype of OrtMemTypeDefault.

Remarks: Use SafeInt when calculating the size of memory to allocate using AllocatorPtr->Alloc.

virtual bool onnxruntime::OpKernelContext::GetUseDeterministicCompute ( ) const

inlinevirtual

Returns whether deterministic computation is preferred.

Definition at line 179 of file op_kernel_context.h.

virtual int onnxruntime::OpKernelContext::ImplicitInputCount ( ) const

inlinevirtual

Definition at line 120 of file op_kernel_context.h.

template<typename T >

const T* onnxruntime::OpKernelContext::Input ( int index ) const

inline

Definition at line 36 of file op_kernel_context.h.

virtual int onnxruntime::OpKernelContext::InputCount ( ) const

inlinevirtual

Definition at line 115 of file op_kernel_context.h.

virtual MLDataType onnxruntime::OpKernelContext::InputType ( int index ) const

virtual

const logging::Logger& onnxruntime::OpKernelContext::Logger ( ) const

inline

Definition at line 110 of file op_kernel_context.h.

virtual int onnxruntime::OpKernelContext::NumVariadicInputs ( size_t arg_num ) const

virtual

Return the number of inputs for a variadic argument.

Parameters

arg_num The operator argument number.

Returns: Number of inputs the argument has.

template<typename T >

T* onnxruntime::OpKernelContext::Output ( int index )

inline

Definition at line 56 of file op_kernel_context.h.

Tensor* onnxruntime::OpKernelContext::Output	(	int	index,
		const TensorShape &	shape
	)

Tensor* onnxruntime::OpKernelContext::Output	(	int	index,
		const std::vector< int64_t > &	shape
	)

Tensor* onnxruntime::OpKernelContext::Output	(	int	index,
		const std::initializer_list< int64_t > &	shape
	)

template<>

Tensor* onnxruntime::OpKernelContext::Output ( int index )

inline

Definition at line 222 of file op_kernel_context.h.

template<>

SparseTensor* onnxruntime::OpKernelContext::Output ( int index )

inline

Definition at line 230 of file op_kernel_context.h.

virtual int onnxruntime::OpKernelContext::OutputCount ( ) const

inlinevirtual

Definition at line 125 of file op_kernel_context.h.

virtual OrtValue* onnxruntime::OpKernelContext::OutputMLValue	(	int	index,
		const TensorShape &	shape
	)

protectedvirtual

template<typename T >

void onnxruntime::OpKernelContext::OutputOptionalWithoutData ( int index )

inline

Definition at line 89 of file op_kernel_context.h.

SparseTensor* onnxruntime::OpKernelContext::OutputSparse	(	int	index,
		const TensorShape &	shape
	)

virtual MLDataType onnxruntime::OpKernelContext::OutputType ( int index ) const

virtual

template<typename T >

const T& onnxruntime::OpKernelContext::RequiredInput ( int index ) const

inline

Definition at line 48 of file op_kernel_context.h.

Tensor& onnxruntime::OpKernelContext::RequiredOutput	(	int	index,
		const TensorShape &	shape
	)

inline

Definition at line 72 of file op_kernel_context.h.

virtual bool onnxruntime::OpKernelContext::TryGetInferredInputShape	(	int	index,
		TensorShape &	shape
	)		const

virtual

virtual bool onnxruntime::OpKernelContext::TryGetInferredOutputShape	(	int	index,
		TensorShape &	shape
	)		const

virtual

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

onnxruntime/core/framework/op_kernel_context.h

Public Types

Public Member Functions

Protected Member Functions

Detailed Description

Member Typedef Documentation

Constructor & Destructor Documentation

Member Function Documentation