Detailed Description

Abstract interface for managing the actual inferencing implementation.

This interface abstracts away the low-level implementation details required for inferencing with implicit batch dimensions network/full dimensions network on GPU and inferencing on DLA.

Actual instance of a BackendContext can be created using createBackendContext function. This function will create the appropriate BackendContext (ImplicitTrtBackendContext/FullDimTrtBackendContext/DlaTrtBackendContext) based on the parameters used to build the network/engine.

Definition at line 167 of file sources/libs/nvdsinfer/nvdsinfer_backend.h.

Inheritance diagram for nvdsinfer::BackendContext:

[legend]

Public Member Functions
	BackendContext ()=default

virtual	~BackendContext ()=default

virtual NvDsInferStatus	initialize ()=0

virtual int	getNumBoundLayers ()=0

virtual const NvDsInferBatchDimsLayerInfo &	getLayerInfo (int bindingIdx)=0

virtual bool	isOutputLayer (const std::string &bindingName)=0

virtual bool	canSupportBatchDims (int bindingIdx, const NvDsInferBatchDims &batchDims)=0

virtual NvDsInferBatchDims	getMaxBatchDims (int bindingIdx)=0

virtual NvDsInferBatchDims	getMinBatchDims (int bindingIdx)=0

virtual NvDsInferBatchDims	getOptBatchDims (int bindingIdx)=0

virtual NvDsInferStatus	enqueueBuffer (const std::shared_ptr< InferBatchBuffer > &buffer, CudaStream &stream, CudaEvent *consumeEvent)=0

	BackendContext ()=default

virtual	~BackendContext ()=default

virtual NvDsInferStatus	initialize ()=0

virtual int	getNumBoundLayers ()=0

virtual const NvDsInferBatchDimsLayerInfo &	getLayerInfo (int bindingIdx)=0

virtual bool	isOutputLayer (const std::string &bindingName)=0

virtual bool	canSupportBatchDims (int bindingIdx, const NvDsInferBatchDims &batchDims)=0

virtual NvDsInferBatchDims	getMaxBatchDims (int bindingIdx)=0

virtual NvDsInferBatchDims	getMinBatchDims (int bindingIdx)=0

virtual NvDsInferBatchDims	getOptBatchDims (int bindingIdx)=0

virtual NvDsInferStatus	enqueueBuffer (const std::shared_ptr< InferBatchBuffer > &buffer, CudaStream &stream, CudaEvent *consumeEvent)=0

Constructor & Destructor Documentation

◆ BackendContext() [1/2]

nvdsinfer::BackendContext::BackendContext ( )

default

◆ ~BackendContext() [1/2]

virtual nvdsinfer::BackendContext::~BackendContext ( )

virtualdefault

◆ BackendContext() [2/2]

nvdsinfer::BackendContext::BackendContext ( )

default

◆ ~BackendContext() [2/2]

virtual nvdsinfer::BackendContext::~BackendContext ( )

virtualdefault

Member Function Documentation

◆ canSupportBatchDims() [1/2]

virtual bool nvdsinfer::BackendContext::canSupportBatchDims	(	int	bindingIdx,
		const NvDsInferBatchDims &	batchDims
	)

pure virtual

Implemented in nvdsinfer::TrtBackendContext, and nvdsinfer::TrtBackendContext.

◆ canSupportBatchDims() [2/2]

virtual bool nvdsinfer::BackendContext::canSupportBatchDims	(	int	bindingIdx,
		const NvDsInferBatchDims &	batchDims
	)

pure virtual

Implemented in nvdsinfer::TrtBackendContext, and nvdsinfer::TrtBackendContext.

◆ enqueueBuffer() [1/2]

virtual NvDsInferStatus nvdsinfer::BackendContext::enqueueBuffer	(	const std::shared_ptr< InferBatchBuffer > &	buffer,
		CudaStream &	stream,
		CudaEvent *	consumeEvent
	)

pure virtual

Implemented in nvdsinfer::DlaFullDimTrtBackendContext, and nvdsinfer::DlaFullDimTrtBackendContext.

◆ enqueueBuffer() [2/2]

virtual NvDsInferStatus nvdsinfer::BackendContext::enqueueBuffer	(	const std::shared_ptr< InferBatchBuffer > &	buffer,
		CudaStream &	stream,
		CudaEvent *	consumeEvent
	)