AMD OpenVX Extension API

AMD OpenVX Extension API#

MIVisionX: Extension: AMD OpenVX Extension API

AMD OpenVX Extension to enhance base, vision, and neural network feature sets. More...

Data Structures
struct	AgoImageFormatDescription
	AMD data structure for image format information. More...

struct	AgoTargetAffinityInfo
	AMD data structure to specify target affinity. More...

struct	AgoContextTextMacroInfo
	AMD data structure to set a text macro. More...

struct	AgoGraphImportInfo
	AMD data structure to import a graph from a text.text: "macro <macro-name>" to use a pre-defined macro "file <file-name>" to load from a file otherwise use the text as is. More...

struct	AgoGraphExportInfo
	AMD data structure to export a graph to a text. More...

struct	AgoGraphPerfInternalInfo
	AMD data structure to get internal performance data. More...

struct	AgoNodeMergeRule_t
	AMD data structure to specify node merge rule. More...

Macros
#define	AGO_TARGET_AFFINITY_CPU 0x0010
	AMD target affinity enumerations for AgoTargetAffinityInfo.device_type.

#define	AGO_MAX_PARAMS 32
	AMD internal parameters. [TODO: This needs to be moved to ago_internal.h].

#define	VX_MAX_STRING_BUFFER_SIZE_AMD 256
	Maximum size of scalar string buffer. The local buffers used for accessing scalar strings should be of size VX_MAX_STRING_BUFFER_SIZE_AMD and the maximum allowed string length is VX_MAX_STRING_BUFFER_SIZE_AMD-1.

#define	VX_CONTEXT_ATTRIBUTE_NONLINEAR_MAX_DIMENSION VX_CONTEXT_NONLINEAR_MAX_DIMENSION
	The attributes for vx_context.

#define	VX_GRAPH_ATTRIBUTE_STATE VX_GRAPH_STATE
	The attributes for vx_graph.

#define	VX_OBJECT_ARRAY_ATTRIBUTE_ITEMTYPE VX_OBJECT_ARRAY_ITEMTYPE
	The attributes for vx_object_array.

Typedefs
typedef struct _vx_tensor_t *	vx_tensor
	The multidimensional data object (Tensor). More...

typedef struct AgoNodeMergeRule_t	AgoNodeMergeRule
	AMD data structure to specify node merge rule.

Enumerations
enum	ago_type_public_e { VX_TYPE_STRING_AMD = 0x011 , AGO_TYPE_KEYPOINT_XYS = VX_TYPE_VENDOR_STRUCT_START , AGO_TYPE_MEANSTDDEV_DATA = VX_TYPE_VENDOR_OBJECT_START , AGO_TYPE_MINMAXLOC_DATA , AGO_TYPE_CANNY_STACK , AGO_TYPE_SCALE_MATRIX }
	The type enumeration lists all the AMD specific types in OpenVX. More...

enum	vx_context_attribute_amd_e { VX_CONTEXT_ATTRIBUTE_AMD_OPENCL_CONTEXT = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_CONTEXT) + 0x01 , VX_CONTEXT_ATTRIBUTE_AMD_AFFINITY = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_CONTEXT) + 0x02 , VX_CONTEXT_ATTRIBUTE_AMD_SET_TEXT_MACRO = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_CONTEXT) + 0x03 , VX_CONTEXT_ATTRIBUTE_AMD_SET_MERGE_RULE = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_CONTEXT) + 0x04 , VX_CONTEXT_MAX_TENSOR_DIMENSIONS = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_CONTEXT) + 0x05 , VX_CONTEXT_CL_QUEUE_PROPERTIES = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_CONTEXT) + 0x06 , VX_CONTEXT_ATTRIBUTE_AMD_HIP_DEVICE = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_CONTEXT) + 0x07 }
	The AMD context attributes list. More...

enum	vx_kernel_attribute_amd_e { VX_KERNEL_ATTRIBUTE_AMD_QUERY_TARGET_SUPPORT = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_KERNEL) + 0x01 , VX_KERNEL_ATTRIBUTE_AMD_OPENCL_CODEGEN_CALLBACK = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_KERNEL) + 0x02 , VX_KERNEL_ATTRIBUTE_AMD_NODE_REGEN_CALLBACK = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_KERNEL) + 0x03 , VX_KERNEL_ATTRIBUTE_AMD_OPENCL_GLOBAL_WORK_UPDATE_CALLBACK = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_KERNEL) + 0x04 , VX_KERNEL_ATTRIBUTE_AMD_GPU_BUFFER_ACCESS_ENABLE = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_KERNEL) + 0x05 , VX_KERNEL_ATTRIBUTE_AMD_GPU_BUFFER_UPDATE_CALLBACK = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_KERNEL) + 0x06 }
	The AMD kernel attributes list. More...

enum	vx_graph_attribute_amd_e { VX_GRAPH_ATTRIBUTE_AMD_AFFINITY = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_GRAPH) + 0x01 , VX_GRAPH_ATTRIBUTE_AMD_IMPORT_FROM_TEXT = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_GRAPH) + 0x02 , VX_GRAPH_ATTRIBUTE_AMD_EXPORT_TO_TEXT = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_GRAPH) + 0x03 , VX_GRAPH_ATTRIBUTE_AMD_OPTIMIZER_FLAGS = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_GRAPH) + 0x04 , VX_GRAPH_ATTRIBUTE_AMD_PERFORMANCE_INTERNAL_LAST = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_GRAPH) + 0x05 , VX_GRAPH_ATTRIBUTE_AMD_PERFORMANCE_INTERNAL_AVG = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_GRAPH) + 0x06 , VX_GRAPH_ATTRIBUTE_AMD_PERFORMANCE_INTERNAL_PROFILE = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_GRAPH) + 0x07 , VX_GRAPH_ATTRIBUTE_AMD_OPENCL_COMMAND_QUEUE = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_GRAPH) + 0x08 , VX_GRAPH_ATTRIBUTE_AMD_CPU_NUM_THREADS = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_GRAPH) + 0x09 }
	The AMD graph attributes list. More...

enum	vx_node_attribute_amd_e { VX_NODE_ATTRIBUTE_AMD_AFFINITY = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_NODE) + 0x01 , VX_NODE_ATTRIBUTE_AMD_OPENCL_COMMAND_QUEUE = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_NODE) + 0x02 , VX_NODE_ATTRIBUTE_AMD_HIP_STREAM = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_NODE) + 0x03 , VX_NODE_ATTRIBUTE_AMD_CPU_NUM_THREADS = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_NODE) + 0x04 }
	The AMD node attributes list. More...

enum	vx_image_attribute_amd_e { VX_IMAGE_ATTRIBUTE_AMD_OPENCL_BUFFER = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_IMAGE) + 0x01 , VX_IMAGE_ATTRIBUTE_AMD_GPU_BUFFER_OFFSET = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_IMAGE) + 0x02 , VX_IMAGE_ATTRIBUTE_AMD_ENABLE_USER_BUFFER_GPU = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_IMAGE) + 0x03 , VX_IMAGE_ATTRIBUTE_AMD_GPU_BUFFER_STRIDE = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_IMAGE) + 0x04 , VX_IMAGE_ATTRIBUTE_AMD_HOST_BUFFER = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_IMAGE) + 0x05 , VX_IMAGE_ATTRIBUTE_AMD_HIP_BUFFER = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_IMAGE) + 0x06 }
	The AMD image attributes list. More...

enum	vx_tensor_attribute_amd_e { VX_TENSOR_STRIDE_GPU = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_TENSOR) + 0x5 , VX_TENSOR_OFFSET_GPU = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_TENSOR) + 0x6 , VX_TENSOR_BUFFER_OPENCL = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_TENSOR) + 0x7 , VX_TENSOR_BUFFER_HIP = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_TENSOR) + 0x8 , VX_TENSOR_MEMORY_TYPE = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_TENSOR) + 0x9 , VX_TENSOR_BUFFER_HOST = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_TENSOR) + 0x10 }
	The AMD tensor data attributes. More...

enum	vx_array_attribute_amd_e { VX_ARRAY_BUFFER_OPENCL = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_ARRAY) + 0x9 , VX_ARRAY_BUFFER_HIP = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_ARRAY) + 0x10 , VX_ARRAY_BUFFER = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_ARRAY) + 0x11 , VX_ARRAY_OFFSET_GPU = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_ARRAY) + 0x12 }
	The AMD array data attributes. More...

enum	vx_directive_amd_e { VX_DIRECTIVE_AMD_READ_ONLY = VX_ENUM_BASE(VX_ID_AMD, VX_ENUM_DIRECTIVE) + 0x01 , VX_DIRECTIVE_AMD_COPY_TO_OPENCL = VX_ENUM_BASE(VX_ID_AMD, VX_ENUM_DIRECTIVE) + 0x02 , VX_DIRECTIVE_AMD_ENABLE_PROFILE_CAPTURE = VX_ENUM_BASE(VX_ID_AMD, VX_ENUM_DIRECTIVE) + 0x03 , VX_DIRECTIVE_AMD_DISABLE_PROFILE_CAPTURE = VX_ENUM_BASE(VX_ID_AMD, VX_ENUM_DIRECTIVE) + 0x04 , VX_DIRECTIVE_AMD_DISABLE_GPU_FLUSH = VX_ENUM_BASE(VX_ID_AMD, VX_ENUM_DIRECTIVE) + 0x05 , VX_DIRECTIVE_AMD_COPY_TO_HIPMEM = VX_ENUM_BASE(VX_ID_AMD, VX_ENUM_DIRECTIVE) + 0x02 }
	These enumerations are given to the `vxDirective` API to enable/disable platform optimizations and/or features. Directives are not optional and usually are vendor-specific, by defining a vendor range of directives and starting their enumeration from there. More...

enum	vx_memory_type_amd_e { VX_MEMORY_TYPE_OPENCL = VX_ENUM_BASE(VX_ID_KHRONOS, VX_ENUM_MEMORY_TYPE) + 0x2 , VX_MEMORY_TYPE_HIP = VX_ENUM_BASE(VX_ID_KHRONOS, VX_ENUM_MEMORY_TYPE) + 0x3 }
	An enumeration of additional memory type imports. More...

enum	vx_color_space_amd_e { VX_COLOR_SPACE_BT2020 = VX_ENUM_BASE(VX_ID_AMD, VX_ENUM_COLOR_SPACE) + 0x1 }
	The image color space list used by the `VX_IMAGE_SPACE` attribute of a `vx_image`. More...

enum	vx_df_image_amd_e { VX_DF_IMAGE_U1_AMD = VX_DF_IMAGE('U', '0', '0', '1') , VX_DF_IMAGE_F16_AMD = VX_DF_IMAGE('F', '0', '1', '6') , VX_DF_IMAGE_F32_AMD = VX_DF_IMAGE('F', '0', '3', '2') , VX_DF_IMAGE_F64_AMD = VX_DF_IMAGE('F', '0', '6', '4') , VX_DF_IMAGE_F32x3_AMD = VX_DF_IMAGE('F', '3', '3', '2') }
	Based on the VX_DF_IMAGE definition. More...

enum	vx_scalar_attribute_amd_e { VX_SCALAR_BUFFER = VX_ATTRIBUTE_BASE(VX_ID_AMD, VX_TYPE_SCALAR) + 0x1 }
	scalar data attributes. More...

Functions
VX_API_ENTRY vx_status VX_API_CALL	vxGetReferenceName (vx_reference ref, vx_char name[], vx_size size)
	Retrieve the name of a reference. More...

VX_API_ENTRY vx_status VX_API_CALL	vxSetModuleInternalData (vx_context context, const vx_char module, void ptr, vx_size size)
	Set module internal data. More...

VX_API_ENTRY vx_status VX_API_CALL	vxGetModuleInternalData (vx_context context, const vx_char module, void ptr, vx_size size)
	Retrieve module internal data. More...

VX_API_ENTRY vx_status VX_API_CALL	vxSetModuleHandle (vx_node node, const vx_char module, void ptr)
	Set module handle. More...

VX_API_ENTRY vx_status VX_API_CALL	vxGetModuleHandle (vx_node node, const vx_char module, void *ptr)
	Retrieve module handle. More...

VX_API_ENTRY vx_status VX_API_CALL	vxSetContextImageFormatDescription (vx_context context, vx_df_image format, const AgoImageFormatDescription *desc)
	Set custom image format description. More...

VX_API_ENTRY vx_status VX_API_CALL	vxGetContextImageFormatDescription (vx_context context, vx_df_image format, AgoImageFormatDescription *desc)
	Get custom image format description. More...

VX_API_ENTRY vx_status VX_API_CALL	vxAliasTensor (vx_tensor tensorMaster, vx_size offset, vx_tensor tensor)
	vxAliasTensor. More...

VX_API_ENTRY vx_bool VX_API_CALL	vxIsTensorAliased (vx_tensor tensorMaster, vx_size offset, vx_tensor tensor)
	vxIsTensorAliased. More...

Detailed Description

AMD OpenVX Extension to enhance base, vision, and neural network feature sets.

Typedef Documentation

◆ vx_tensor

typedef struct _vx_tensor_t* vx_tensor

The multidimensional data object (Tensor).

See also: vxCreateTensor

Enumeration Type Documentation

◆ ago_type_public_e

enum ago_type_public_e

The type enumeration lists all the AMD specific types in OpenVX.

Enumerator
VX_TYPE_STRING_AMD	AMD scalar data type for string.
AGO_TYPE_KEYPOINT_XYS	AGO struct data type for keypoint XYS.
AGO_TYPE_MEANSTDDEV_DATA	AGO data structure for AGO MeanStdDev kernels.
AGO_TYPE_MINMAXLOC_DATA	AGO data structure for AGO MinMaxLoc kernels.
AGO_TYPE_CANNY_STACK	AGO data structure for AGO Canny kernels.
AGO_TYPE_SCALE_MATRIX	AGO data structure for AGO Scale kernels.

◆ vx_array_attribute_amd_e

enum vx_array_attribute_amd_e

The AMD array data attributes.

Enumerator
VX_ARRAY_BUFFER_OPENCL	OpenCL buffer. `cl_mem`.
VX_ARRAY_BUFFER_HIP	HIP buffer. `hip_mem`.
VX_ARRAY_BUFFER	TODO:.
VX_ARRAY_OFFSET_GPU	TODO:.

◆ vx_color_space_amd_e

enum vx_color_space_amd_e

The image color space list used by the VX_IMAGE_SPACE attribute of a vx_image.

Enumerator
VX_COLOR_SPACE_BT2020	Use to indicate that the BT.2020 coefficients are used for conversions.

◆ vx_context_attribute_amd_e

enum vx_context_attribute_amd_e

The AMD context attributes list.

Enumerator
VX_CONTEXT_ATTRIBUTE_AMD_OPENCL_CONTEXT	OpenCL context. Use a `cl_context` parameter.
VX_CONTEXT_ATTRIBUTE_AMD_AFFINITY	Context affinity. Use a `AgoTargetAffinityInfo` parameter.
VX_CONTEXT_ATTRIBUTE_AMD_SET_TEXT_MACRO	Set a text macro definition. Use a `AgoContextMacroInfo` parameter.
VX_CONTEXT_ATTRIBUTE_AMD_SET_MERGE_RULE	Set a merge rule. Use a `AgoNodeMergeRule` parameter.
VX_CONTEXT_MAX_TENSOR_DIMENSIONS	Tensor Data max num of dimensions supported by HW.
VX_CONTEXT_CL_QUEUE_PROPERTIES	CL_QUEUE_PROPERTIES to be used for creating OpenCL command queue. Use a `cl_command_queue_properties` parameter.
VX_CONTEXT_ATTRIBUTE_AMD_HIP_DEVICE	HIP context. Use a `cl_context` parameter.

◆ vx_df_image_amd_e

enum vx_df_image_amd_e

Based on the VX_DF_IMAGE definition.

Note: Use vx_df_image to contain these values.

Enumerator
VX_DF_IMAGE_U1_AMD	AMD image with 1-bit data.
VX_DF_IMAGE_F16_AMD	AMD image with 16-bit floating-point (half).
VX_DF_IMAGE_F32_AMD	AMD image with 32-bit floating-point (float).
VX_DF_IMAGE_F64_AMD	AMD image with 64-bit floating-point (double).
VX_DF_IMAGE_F32x3_AMD	AMD image with THREE 32-bit floating-point channels in one buffer.

◆ vx_directive_amd_e

enum vx_directive_amd_e

These enumerations are given to the vxDirective API to enable/disable platform optimizations and/or features. Directives are not optional and usually are vendor-specific, by defining a vendor range of directives and starting their enumeration from there.

See also: vxDirective

Enumerator
VX_DIRECTIVE_AMD_READ_ONLY	Data object is readonly after this directive is given.
VX_DIRECTIVE_AMD_COPY_TO_OPENCL	Data object copy to OpenCL.
VX_DIRECTIVE_AMD_ENABLE_PROFILE_CAPTURE	Enable performance profile capture.
VX_DIRECTIVE_AMD_DISABLE_PROFILE_CAPTURE	Disable performance profile capture.
VX_DIRECTIVE_AMD_DISABLE_GPU_FLUSH	Disable node level flush for a graph.
VX_DIRECTIVE_AMD_COPY_TO_HIPMEM	Data object copy to HIP memory.

◆ vx_graph_attribute_amd_e

enum vx_graph_attribute_amd_e

The AMD graph attributes list.

Enumerator
VX_GRAPH_ATTRIBUTE_AMD_AFFINITY	Graph affinity. Use a `AgoNodeAffinityInfo` parameter.
VX_GRAPH_ATTRIBUTE_AMD_IMPORT_FROM_TEXT	Imports a graph from a text file. Use a `AgoGraphImportInfo` parameter.
VX_GRAPH_ATTRIBUTE_AMD_EXPORT_TO_TEXT	Export a graph into a text file. Use a `AgoGraphExportInfo` parameter.
VX_GRAPH_ATTRIBUTE_AMD_OPTIMIZER_FLAGS	Graph optimizer flags. Use a `vx_uint32` parameter.
VX_GRAPH_ATTRIBUTE_AMD_PERFORMANCE_INTERNAL_LAST	Graph last performance (internal). Use a `AgoGraphPerfInternalInfo` parameter.
VX_GRAPH_ATTRIBUTE_AMD_PERFORMANCE_INTERNAL_AVG	Graph avg performance (internal). Use a `AgoGraphPerfInternalInfo` parameter.
VX_GRAPH_ATTRIBUTE_AMD_PERFORMANCE_INTERNAL_PROFILE	Graph internal performance profile. Use a char * fileName parameter.
VX_GRAPH_ATTRIBUTE_AMD_OPENCL_COMMAND_QUEUE	OpenCL command queue. Use a `cl_command_queue` parameter.
VX_GRAPH_ATTRIBUTE_AMD_CPU_NUM_THREADS	CPU num_threads to be used in RPP. Use a `vx_uint32` parameter.

◆ vx_image_attribute_amd_e

enum vx_image_attribute_amd_e

The AMD image attributes list.

Enumerator
VX_IMAGE_ATTRIBUTE_AMD_OPENCL_BUFFER	Sync with user specified OpenCL buffer. Use a `cl_mem` parameter.
VX_IMAGE_ATTRIBUTE_AMD_GPU_BUFFER_OFFSET	GPU buffer offset. Use a `cl_uint` parameter.
VX_IMAGE_ATTRIBUTE_AMD_ENABLE_USER_BUFFER_GPU	Enable user kernel's own GPU buffer for virtual images. Supports only images with single color plane and stride should match framework's internal alignment. image ROI not supported. Use a `vx_bool` parameter.
VX_IMAGE_ATTRIBUTE_AMD_GPU_BUFFER_STRIDE	GPU buffer stride. Use a `cl_uint` parameter.
VX_IMAGE_ATTRIBUTE_AMD_HOST_BUFFER	Sync with user specified host buffer. Use a `cl_mem` parameter.
VX_IMAGE_ATTRIBUTE_AMD_HIP_BUFFER	Sync with user specified hip memory.

◆ vx_kernel_attribute_amd_e

enum vx_kernel_attribute_amd_e

The AMD kernel attributes list.

Enumerator
VX_KERNEL_ATTRIBUTE_AMD_QUERY_TARGET_SUPPORT	Kernel callback for query target support. Use a `amd_kernel_query_target_support_f` parameter.
VX_KERNEL_ATTRIBUTE_AMD_OPENCL_CODEGEN_CALLBACK	Kernel callback for OpenCL code generation. Use a `amd_kernel_opencl_codegen_callback_f` parameter.
VX_KERNEL_ATTRIBUTE_AMD_NODE_REGEN_CALLBACK	Kernel callback for node regeneration. Use a `amd_kernel_node_regen_callback_f` parameter.
VX_KERNEL_ATTRIBUTE_AMD_OPENCL_GLOBAL_WORK_UPDATE_CALLBACK	Kernel callback for OpenCL global work[]. Use a `amd_kernel_opencl_global_work_update_callback_f` parameter.
VX_KERNEL_ATTRIBUTE_AMD_GPU_BUFFER_ACCESS_ENABLE	Kernel flag to enable OpenCL buffer access (default OFF). Use a `vx_bool` parameter.
VX_KERNEL_ATTRIBUTE_AMD_GPU_BUFFER_UPDATE_CALLBACK	Kernel callback for GPU buffer update. Use a `AgoKernelGpuBufferUpdateInfo` parameter.

◆ vx_memory_type_amd_e

enum vx_memory_type_amd_e

An enumeration of additional memory type imports.

Enumerator
VX_MEMORY_TYPE_OPENCL	The memory type to import from the OpenCL.
VX_MEMORY_TYPE_HIP	The memory type to import from the HIP.

◆ vx_node_attribute_amd_e

enum vx_node_attribute_amd_e

The AMD node attributes list.

Enumerator
VX_NODE_ATTRIBUTE_AMD_AFFINITY	Node affinity. Use a `AgoTargetAffinityInfo` parameter.
VX_NODE_ATTRIBUTE_AMD_OPENCL_COMMAND_QUEUE	OpenCL command queue. Use a `cl_command_queue` parameter.
VX_NODE_ATTRIBUTE_AMD_HIP_STREAM	HIP stream. TBD.
VX_NODE_ATTRIBUTE_AMD_CPU_NUM_THREADS	TBD.

◆ vx_scalar_attribute_amd_e

enum vx_scalar_attribute_amd_e

scalar data attributes.

Enumerator
VX_SCALAR_BUFFER	scalar's buffer

◆ vx_tensor_attribute_amd_e

enum vx_tensor_attribute_amd_e

The AMD tensor data attributes.

Enumerator
VX_TENSOR_STRIDE_GPU	GPU buffer strides (array of `vx_size`).
VX_TENSOR_OFFSET_GPU	GPU buffer offset. `vx_size`.
VX_TENSOR_BUFFER_OPENCL	OpenCL buffer. `cl_mem`.
VX_TENSOR_BUFFER_HIP	HIP buffer. `cl_mem`.
VX_TENSOR_MEMORY_TYPE	Queries memory type if created using vxCreateTensorFromHandle. If vx_tensor was not created using vxCreateTensorFromHandle, VX_MEMORY_TYPE_NONE is returned. Use a `vx_memory_type_e` parameter.
VX_TENSOR_BUFFER_HOST	Sync with user specified host buffer.

Function Documentation

◆ vxAliasTensor()

VX_API_ENTRY vx_status VX_API_CALL vxAliasTensor	(	vx_tensor	tensorMaster,
		vx_size	offset,
		vx_tensor	tensor
	)

vxAliasTensor.

Note: - TBD

Returns: A vx_status_e enumeration.

Return values

VX_SUCCESS	No errors.
VX_ERROR_INVALID_REFERENCE	if reference is not valid.
VX_ERROR_INVALID_FORMAT	if format is already in use.

◆ vxGetContextImageFormatDescription()

VX_API_ENTRY vx_status VX_API_CALL vxGetContextImageFormatDescription	(	vx_context	context,
		vx_df_image	format,
		AgoImageFormatDescription *	desc
	)

Get custom image format description.

Parameters

[in]	context	The context.
[in]	format	The image format.
[out]	desc	The image format description.

Returns: A vx_status_e enumeration.

Return values

VX_SUCCESS	No errors.
VX_ERROR_INVALID_REFERENCE	if reference is not valid.
VX_ERROR_INVALID_FORMAT	if format is already in use.

◆ vxGetModuleHandle()

VX_API_ENTRY vx_status VX_API_CALL vxGetModuleHandle	(	vx_node	node,
		const vx_char *	module,
		void **	ptr
	)

Retrieve module handle.

This function is used to retrieve module specific handle of a graph from within a node.

Parameters

[in]	node	The node.
[in]	module	The name of the module used in vxLoadKernels.
[out]	ptr	The module internal buffer.

Returns: A vx_status_e enumeration.

Return values

VX_SUCCESS	No errors.
VX_ERROR_INVALID_REFERENCE	if reference is not valid.

◆ vxGetModuleInternalData()

VX_API_ENTRY vx_status VX_API_CALL vxGetModuleInternalData	(	vx_context	context,
		const vx_char *	module,
		void **	ptr,
		vx_size *	size
	)

Retrieve module internal data.

This function is used to retrieve module specific internal data. This is for use by vxUnpublishKernels().

Parameters

[in]	context	The context.
[in]	module	The name of the module used in vxLoadKernels.
[out]	ptr	The module internal buffer.
[out]	size	Size of the module internal buffer.

Returns: A vx_status_e enumeration.

Return values

VX_SUCCESS	No errors.
VX_ERROR_INVALID_REFERENCE	if reference is not valid.

◆ vxGetReferenceName()

VX_API_ENTRY vx_status VX_API_CALL vxGetReferenceName	(	vx_reference	ref,
		vx_char	name[],
		vx_size	size
	)

Retrieve the name of a reference.

This function is used to retrieve the name of a reference.

Parameters

[in]	ref	The reference.
[out]	name	Pointer to copy the name of the reference.
[in]	size	Size of the name buffer.

Returns: A vx_status_e enumeration.

Return values

VX_SUCCESS	No errors.
VX_ERROR_INVALID_REFERENCE	if reference is not valid.

◆ vxIsTensorAliased()

VX_API_ENTRY vx_bool VX_API_CALL vxIsTensorAliased	(	vx_tensor	tensorMaster,
		vx_size	offset,
		vx_tensor	tensor
	)

vxIsTensorAliased.

Note: - TBD

◆ vxSetContextImageFormatDescription()

VX_API_ENTRY vx_status VX_API_CALL vxSetContextImageFormatDescription	(	vx_context	context,
		vx_df_image	format,
		const AgoImageFormatDescription *	desc
	)

Set custom image format description.

This function is used to support custom image formats with single-plane by ISVs. Should be called from vxPublishKernels().

Parameters

[in]	context	The context.
[in]	format	The image format.
[in]	desc	The image format description.

Returns: A vx_status_e enumeration.

Return values

VX_SUCCESS	No errors.
VX_ERROR_INVALID_REFERENCE	if reference is not valid.
VX_ERROR_INVALID_FORMAT	if format is already in use.

◆ vxSetModuleHandle()

VX_API_ENTRY vx_status VX_API_CALL vxSetModuleHandle	(	vx_node	node,
		const vx_char *	module,
		void *	ptr
	)

Set module handle.

This function is used to set module specific of a graph from within a node.

Parameters

[in]	node	The node.
[in]	module	The name of the module used in vxLoadKernels.
[in]	ptr	The module internal buffer.

Returns: A vx_status_e enumeration.

Return values

VX_SUCCESS	No errors.
VX_ERROR_INVALID_REFERENCE	if reference is not valid.

◆ vxSetModuleInternalData()

VX_API_ENTRY vx_status VX_API_CALL vxSetModuleInternalData	(	vx_context	context,
		const vx_char *	module,
		void *	ptr,
		vx_size	size
	)

Set module internal data.

This function is used to set module specific internal data. This is for use by vxPublishKernels().

Parameters

[in]	context	The context.
[in]	module	The name of the module used in vxLoadKernels.
[in]	ptr	The module internal buffer.
[in]	size	Size of the module internal buffer.

Returns: A vx_status_e enumeration.

Return values

VX_SUCCESS	No errors.
VX_ERROR_INVALID_REFERENCE	if reference is not valid.

AMD OpenVX Extension API

AMD OpenVX Extension API#

Data Structures

Macros

Typedefs

Enumerations

Functions

Detailed Description

Typedef Documentation

◆ vx_tensor

Enumeration Type Documentation

◆ ago_type_public_e

◆ vx_array_attribute_amd_e

◆ vx_color_space_amd_e

◆ vx_context_attribute_amd_e

◆ vx_df_image_amd_e

◆ vx_directive_amd_e

◆ vx_graph_attribute_amd_e

◆ vx_image_attribute_amd_e

◆ vx_kernel_attribute_amd_e

◆ vx_memory_type_amd_e

◆ vx_node_attribute_amd_e

◆ vx_scalar_attribute_amd_e

◆ vx_tensor_attribute_amd_e

Function Documentation

◆ vxAliasTensor()

◆ vxGetContextImageFormatDescription()

◆ vxGetModuleHandle()

◆ vxGetModuleInternalData()

◆ vxGetReferenceName()

◆ vxIsTensorAliased()

◆ vxSetContextImageFormatDescription()

◆ vxSetModuleHandle()

◆ vxSetModuleInternalData()