Class GXFExecutionContext

• Defined in File gxf_execution_context.hpp

Base Type

• public holoscan::ExecutionContext (Class ExecutionContext)

class GXFExecutionContext : public holoscan::ExecutionContext

Class to hold the execution context for GXF Operator.

This class provides the execution context for the operator using GXF.

Public Functions

GXFExecutionContext(gxf_context_t context, Operator *op)

Construct a new GXFExecutionContext object.

Parameters

• context – The pointer to the GXF context.

• op – The pointer to the operator.

GXFExecutionContext(gxf_context_t context, std::shared_ptr<GXFInputContext> gxf_input_context, std::shared_ptr<GXFOutputContext> gxf_output_context)

Construct a new GXFExecutionContext object.

Parameters

• context – The pointer to the GXF context.

• gxf_input_context – The shared pointer to the GXFInputContext object.

• gxf_output_context – The shared pointer to the GXFOutputContext object.

~GXFExecutionContext() override = default

inline std::shared_ptr<GXFInputContext> gxf_input()

Get the GXF input context.

Returns

The pointer to the GXFInputContext object.

inline std::shared_ptr<GXFOutputContext> gxf_output()

Get the GXF output context.

Returns

The pointer to the GXFOutputContext object.

virtual expected<cudaStream_t, RuntimeError> allocate_cuda_stream(const std::string &stream_name) override

allocate a new GXF CudaStream object and return the cudaStream_t corresponding to it

virtual void synchronize_streams(const std::vector<std::optional<cudaStream_t>> &cuda_streams, cudaStream_t target_cuda_stream) override

synchronize all of the streams in cuda_streams with target_cuda_stream

virtual expected<int, RuntimeError> device_from_stream(cudaStream_t stream) override

determine the CUDA device corresponding to the given stream

expected<gxf::CudaStreamHandle, RuntimeError> stream_handle_from_stream(cudaStream_t stream)

Return the CudaStreamHandle corresponding to a given cudaStream_t.

This will only work with a cudaStream_t that was allocated as a CudaStream object by GXF. The stream should correspond to a CudaStreamId that was received on one of the Operator’s input ports or a stream that was allocated via allocate_cuda_stream.

Parameters

stream_handle – A CUDA stream object.

Returns

The GXF CudaStream handle if found, or unexpected if not found.

inline std::shared_ptr<CudaObjectHandler> cuda_object_handler()

void init_cuda_object_handler(Operator *op)

initialize the CudaObjectHandler for the Operator

void release_internal_cuda_streams()

release any internal stream objects allocated by the operator

void clear_received_streams()

clear the handler’s received stream mappings from a prior Operator::compute call.

Protected Functions

expected<CudaStreamHandle, RuntimeError> allocate_cuda_stream_handle(const std::string &stream_name)

allocate a new GXF CudaStream object and return the GXF Handle to it

Protected Attributes

std::shared_ptr<GXFInputContext> gxf_input_context_ = {}

The GXF input context.

std::shared_ptr<GXFOutputContext> gxf_output_context_ = {}

The GXF output context.

std::shared_ptr<CudaObjectHandler> cuda_object_handler_ = {}

Friends

friend class holoscan::gxf::GXFWrapper

friend class holoscan::gxf::GXFInputContext

friend class holoscan::gxf::GXFOutputContext