Ort::Custom::CudaContext Struct Reference

#include <cuda_context.h>

Inheritance diagram for Ort::Custom::CudaContext:

Public Member Functions
void	Init (const OrtKernelContext &kernel_ctx)
template<typename T >
T	FetchResource (const OrtKernelContext &kernel_ctx, CudaResource resource_type)
void *	AllocDeferredCpuMem (size_t size) const
void	FreeDeferredCpuMem (void *mem) const
Public Member Functions inherited from CustomOpContext
	CustomOpContext ()=default
virtual	~CustomOpContext ()

Public Attributes
cudaStream_t	cuda_stream = {}
cudnnHandle_t	cudnn_handle = {}
cublasHandle_t	cublas_handle = {}
OrtAllocator *	deferred_cpu_allocator = {}
int16_t	device_id = 0
int32_t	arena_extend_strategy = 0
int32_t	cudnn_conv_algo_search = 0
bool	cudnn_conv_use_max_workspace = true
bool	cudnn_conv1d_pad_to_nc1d = false
bool	enable_skip_layer_norm_strict_mode = false
bool	prefer_nhwc = false
bool	use_tf32 = true
bool	fuse_conv_bias = true

Detailed Description

Definition at line 29 of file cuda_context.h.

Member Function Documentation

void* Ort::Custom::CudaContext::AllocDeferredCpuMem ( size_t  size ) const

inline

Definition at line 83 of file cuda_context.h.

template<typename T >

T Ort::Custom::CudaContext::FetchResource ( const OrtKernelContext &  kernel_ctx, CudaResource  resource_type  )

inline

Definition at line 65 of file cuda_context.h.

void Ort::Custom::CudaContext::FreeDeferredCpuMem ( void *  mem ) const

inline

Definition at line 96 of file cuda_context.h.

void Ort::Custom::CudaContext::Init ( const OrtKernelContext &  kernel_ctx )

inline

Definition at line 45 of file cuda_context.h.

Member Data Documentation

int32_t Ort::Custom::CudaContext::arena_extend_strategy = 0

Definition at line 36 of file cuda_context.h.

cublasHandle_t Ort::Custom::CudaContext::cublas_handle = {}

Definition at line 32 of file cuda_context.h.

cudaStream_t Ort::Custom::CudaContext::cuda_stream = {}

Definition at line 30 of file cuda_context.h.

bool Ort::Custom::CudaContext::cudnn_conv1d_pad_to_nc1d = false

Definition at line 39 of file cuda_context.h.

int32_t Ort::Custom::CudaContext::cudnn_conv_algo_search = 0

Definition at line 37 of file cuda_context.h.

bool Ort::Custom::CudaContext::cudnn_conv_use_max_workspace = true

Definition at line 38 of file cuda_context.h.

cudnnHandle_t Ort::Custom::CudaContext::cudnn_handle = {}

Definition at line 31 of file cuda_context.h.

OrtAllocator* Ort::Custom::CudaContext::deferred_cpu_allocator = {}

Definition at line 33 of file cuda_context.h.

int16_t Ort::Custom::CudaContext::device_id = 0

Definition at line 35 of file cuda_context.h.

bool Ort::Custom::CudaContext::enable_skip_layer_norm_strict_mode = false

Definition at line 40 of file cuda_context.h.

bool Ort::Custom::CudaContext::fuse_conv_bias = true

Definition at line 43 of file cuda_context.h.

bool Ort::Custom::CudaContext::prefer_nhwc = false

Definition at line 41 of file cuda_context.h.

bool Ort::Custom::CudaContext::use_tf32 = true

Definition at line 42 of file cuda_context.h.

---

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

• onnxruntime/core/providers/cuda/cuda_context.h