The AMD rocAL data transfer functions. More...

Functions
RocalStatus ROCAL_API_CALL	rocalCopyToOutput (RocalContext context, unsigned char *out_ptr, size_t out_size)
	copies data to output buffer More...

RocalStatus ROCAL_API_CALL	rocalToTensor (RocalContext rocal_context, void *out_ptr, RocalTensorLayout tensor_format, RocalTensorOutputType tensor_output_type, float multiplier0, float multiplier1, float multiplier2, float offset0, float offset1, float offset2, bool reverse_channels, RocalOutputMemType output_mem_type, int max_roi_height=0, int max_roi_width=0)
	converts data to a tensor More...

void ROCAL_API_CALL	rocalSetOutputs (RocalContext p_context, unsigned int num_of_outputs, std::vector< RocalTensor > &output_images)
	Sets the output images in the RocalContext. More...

RocalTensorList ROCAL_API_CALL	rocalGetOutputTensors (RocalContext p_context)
	gives the list of output tensors from rocal context More...

RocalStatus ROCAL_API_CALL	rocalExternalSourceFeedInput (RocalContext p_context, const std::vector< std::string > &input_images_names, bool is_labels, const std::vector< unsigned char * > &input_buffer, const std::vector< ROIxywh > &roi_xywh, unsigned int max_width, unsigned int max_height, unsigned int channels, RocalExternalSourceMode mode, RocalTensorLayout layout, bool eos)
	Creates ExternalSourceFeedInput for data transfer. More...

Detailed Description

The AMD rocAL data transfer functions.

Function Documentation

RocalStatus ROCAL_API_CALL rocalCopyToOutput	(	RocalContext	context,
		unsigned char *	out_ptr,
		size_t	out_size
	)

copies data to output buffer

Parameters

RocalStatus ROCAL_API_CALL rocalExternalSourceFeedInput	(	RocalContext	p_context,
		const std::vector< std::string > &	input_images_names,
		bool	is_labels,
		const std::vector< unsigned char * > &	input_buffer,
		const std::vector< ROIxywh > &	roi_xywh,
		unsigned int	max_width,
		unsigned int	max_height,
		unsigned int	channels,
		RocalExternalSourceMode	mode,
		RocalTensorLayout	layout,
		bool	eos
	)

Creates ExternalSourceFeedInput for data transfer.

Parameters

rocal_context	Rocal context
input_images	Strings pointing to the location on the disk
labels	Labels whose values is passed by the user using an external source
input_buffer	Compressed or uncompressed input buffer
roi_width	The roi width of the images
roi_height	The roi height of the images
max_width	The maximum width of the decoded images, larger or smaller will be resized to closest
max_height	The maximum height of the decoded images, larger or smaller will be resized to closest
channels	The number of channels for the image
mode	Determines the mode of the source passed from the user - file_names / uncompressed data / compressed data
layout	Determines the layout of the images - NCHW / NHWC

RocalTensorList ROCAL_API_CALL rocalGetOutputTensors ( RocalContext p_context )

gives the list of output tensors from rocal context

Parameters

[in] p_context Rocal Context

void ROCAL_API_CALL rocalSetOutputs	(	RocalContext	p_context,
		unsigned int	num_of_outputs,
		std::vector< RocalTensor > &	output_images
	)

Sets the output images in the RocalContext.

Parameters

RocalStatus ROCAL_API_CALL rocalToTensor	(	RocalContext	rocal_context,
		void *	out_ptr,
		RocalTensorLayout	tensor_format,
		RocalTensorOutputType	tensor_output_type,
		float	multiplier0,
		float	multiplier1,
		float	multiplier2,
		float	offset0,
		float	offset1,
		float	offset2,
		bool	reverse_channels,
		RocalOutputMemType	output_mem_type,
		int	max_roi_height = `0`,
		int	max_roi_width = `0`
	)

converts data to a tensor

Parameters

[in]	rocal_context	Rocal context
[in]	out_ptr	pointer to output buffer
[in]	tensor_format	the layout of the tensor data
[in]	tensor_output_type	the output type of the tensor data
[in]	multiplier0	the multiplier for channel 0
[in]	multiplier1	the multiplier for channel 1
[in]	multiplier2	the multiplier for channel 2
[in]	offset0	the offset for channel 0
[in]	offset1	the offset for channel 1
[in]	offset2	the offset for channel 2
[in]	reverse_channels	flag to reverse the channel orders
[in]	output_mem_type	the memory type of output tensor buffer