1.1. Data types used by cuDLA driver
Macros
- CUDLA_DATA_CATEGORY_BIAS
- CUDLA_DATA_CATEGORY_FEATURE
- CUDLA_DATA_CATEGORY_IMAGE
- CUDLA_DATA_CATEGORY_PLANAR
- CUDLA_DATA_CATEGORY_WEIGHT
- CUDLA_DATA_FORMAT_NCHW
- CUDLA_DATA_FORMAT_NCxHWx
- CUDLA_DATA_FORMAT_NHWC
- CUDLA_DATA_FORMAT_UNKNOWN
- CUDLA_DATA_TYPE_FLOAT
- CUDLA_DATA_TYPE_HALF
- CUDLA_DATA_TYPE_INT16
- CUDLA_DATA_TYPE_INT8
- CUDLA_DATA_TYPE_UINT16
- CUDLA_DATA_TYPE_UINT8
- CUDLA_DATA_TYPE_UNKNOWN
- CUDLA_LOADABLE_TENSOR_DESC_NUM_STRIDES
- CUDLA_PIXEL_FORMAT_A16B16G16R16
- CUDLA_PIXEL_FORMAT_A16B16G16R16_F
- CUDLA_PIXEL_FORMAT_A16Y16U16V16
- CUDLA_PIXEL_FORMAT_A16Y16U16V16_F
- CUDLA_PIXEL_FORMAT_A2B10G10R10
- CUDLA_PIXEL_FORMAT_A2R10G10B10
- CUDLA_PIXEL_FORMAT_A2Y10U10V10
- CUDLA_PIXEL_FORMAT_A8B8G8R8
- CUDLA_PIXEL_FORMAT_A8R8G8B8
- CUDLA_PIXEL_FORMAT_A8Y8U8V8
- CUDLA_PIXEL_FORMAT_B10G10R10A2
- CUDLA_PIXEL_FORMAT_B8G8R8A8
- CUDLA_PIXEL_FORMAT_B8G8R8X8
- CUDLA_PIXEL_FORMAT_FEATURE
- CUDLA_PIXEL_FORMAT_R10
- CUDLA_PIXEL_FORMAT_R10G10B10A2
- CUDLA_PIXEL_FORMAT_R12
- CUDLA_PIXEL_FORMAT_R16
- CUDLA_PIXEL_FORMAT_R16_F
- CUDLA_PIXEL_FORMAT_R16_I
- CUDLA_PIXEL_FORMAT_R8
- CUDLA_PIXEL_FORMAT_R8G8B8A8
- CUDLA_PIXEL_FORMAT_R8G8B8X8
- CUDLA_PIXEL_FORMAT_UNKNOWN
- CUDLA_PIXEL_FORMAT_V10U10Y10A2
- CUDLA_PIXEL_FORMAT_V16U16Y16A16
- CUDLA_PIXEL_FORMAT_V8U8Y8A8
- CUDLA_PIXEL_FORMAT_X16B16G16R16
- CUDLA_PIXEL_FORMAT_X8B8G8R8
- CUDLA_PIXEL_FORMAT_X8R8G8B8
- CUDLA_PIXEL_FORMAT_Y10___U10V10_N444
- CUDLA_PIXEL_FORMAT_Y10___V10U10_N444
- CUDLA_PIXEL_FORMAT_Y12___U12V12_N444
- CUDLA_PIXEL_FORMAT_Y12___V12U12_N444
- CUDLA_PIXEL_FORMAT_Y16___U16V16_N444
- CUDLA_PIXEL_FORMAT_Y16___V16U16_N444
- CUDLA_PIXEL_FORMAT_Y8___U8V8_N444
- CUDLA_PIXEL_FORMAT_Y8___V8U8_N444
- CUDLA_PIXEL_MAPPING_BLOCK_LINEAR
- CUDLA_PIXEL_MAPPING_PITCH_LINEAR
- CUDLA_RUNTIME_TENSOR_DESC_NAME_MAX_LEN
- CUDLA_VER_MAJOR
- CUDLA_VER_MINOR
- CUDLA_VER_PATCH
Enumerations
- cudlaAccessPermissionFlags
-
Access permission flags for importing NvSciBuffers.
- cudlaDevAttributeType
-
Device attribute type.
- cudlaFenceType
-
Supported fence types.
- cudlaMode
-
Device creation modes.
- cudlaModuleAttributeType
-
Module attribute types.
- cudlaModuleLoadFlags
-
Module load flags for cudlaModuleLoadFromMemory .
- cudlaNvSciSyncAttributes
-
cuDLA NvSciSync attributes.
- cudlaScratchMemoryConfig
-
Scratch-memory configuration for cudlaCreateDevice, OR-ed with the device mode.
- cudlaStatus
-
Error codes.
- cudlaSubmissionFlags
-
Task submission flags for cudlaSubmitTask .
Structs
- CudlaFence
-
Fence description.
- cudlaExternalMemoryHandleDesc_t
-
External memory handle descriptor.
- cudlaExternalSemaphoreHandleDesc_t
-
External semaphore handle descriptor.
- cudlaModuleTensorDescriptor
-
Tensor descriptor.
- cudlaSignalEvents
-
Signal events for cudlaSubmitTask .
- cudlaTask
-
Structure of Task.
- cudlaWaitEvents
-
Wait events for cudlaSubmitTask .
Typedefs
Unions
- cudlaDevAttribute
-
Device attribute.
- cudlaModuleAttribute
-
Module attribute.
1.1.1. Macros
-
CUDLA_DATA_CATEGORY_BIAS 4U
-
CUDLA_DATA_CATEGORY_FEATURE 2U
-
CUDLA_DATA_CATEGORY_IMAGE 0U
-
CUDLA_DATA_CATEGORY_PLANAR 3U
-
CUDLA_DATA_CATEGORY_WEIGHT 1U
-
CUDLA_DATA_FORMAT_NCHW 1U
-
CUDLA_DATA_FORMAT_NCxHWx 3U
-
CUDLA_DATA_FORMAT_NHWC 2U
-
CUDLA_DATA_FORMAT_UNKNOWN 0U
-
CUDLA_DATA_TYPE_FLOAT 1U
-
CUDLA_DATA_TYPE_HALF 2U
-
CUDLA_DATA_TYPE_INT16 3U
-
CUDLA_DATA_TYPE_INT8 4U
-
CUDLA_DATA_TYPE_UINT16 6U
-
CUDLA_DATA_TYPE_UINT8 5U
-
CUDLA_DATA_TYPE_UNKNOWN 0U
-
CUDLA_LOADABLE_TENSOR_DESC_NUM_STRIDES 8U
-
CUDLA_PIXEL_FORMAT_A16B16G16R16 6U
-
CUDLA_PIXEL_FORMAT_A16B16G16R16_F 8U
-
CUDLA_PIXEL_FORMAT_A16Y16U16V16 9U
-
CUDLA_PIXEL_FORMAT_A16Y16U16V16_F 11U
-
CUDLA_PIXEL_FORMAT_A2B10G10R10 20U
-
CUDLA_PIXEL_FORMAT_A2R10G10B10 21U
-
CUDLA_PIXEL_FORMAT_A2Y10U10V10 24U
-
CUDLA_PIXEL_FORMAT_A8B8G8R8 12U
-
CUDLA_PIXEL_FORMAT_A8R8G8B8 13U
-
CUDLA_PIXEL_FORMAT_A8Y8U8V8 26U
-
CUDLA_PIXEL_FORMAT_B10G10R10A2 22U
-
CUDLA_PIXEL_FORMAT_B8G8R8A8 14U
-
CUDLA_PIXEL_FORMAT_B8G8R8X8 18U
-
CUDLA_PIXEL_FORMAT_FEATURE 36U
-
CUDLA_PIXEL_FORMAT_R10 1U
-
CUDLA_PIXEL_FORMAT_R10G10B10A2 23U
-
CUDLA_PIXEL_FORMAT_R12 2U
-
CUDLA_PIXEL_FORMAT_R16 3U
-
CUDLA_PIXEL_FORMAT_R16_F 5U
-
CUDLA_PIXEL_FORMAT_R16_I 4U
-
CUDLA_PIXEL_FORMAT_R8 0U
-
CUDLA_PIXEL_FORMAT_R8G8B8A8 15U
-
CUDLA_PIXEL_FORMAT_R8G8B8X8 19U
-
CUDLA_PIXEL_FORMAT_UNKNOWN 37U
-
CUDLA_PIXEL_FORMAT_V10U10Y10A2 25U
-
CUDLA_PIXEL_FORMAT_V16U16Y16A16 10U
-
CUDLA_PIXEL_FORMAT_V8U8Y8A8 27U
-
CUDLA_PIXEL_FORMAT_X16B16G16R16 7U
-
CUDLA_PIXEL_FORMAT_X8B8G8R8 16U
-
CUDLA_PIXEL_FORMAT_X8R8G8B8 17U
-
CUDLA_PIXEL_FORMAT_Y10___U10V10_N444 30U
-
CUDLA_PIXEL_FORMAT_Y10___V10U10_N444 31U
-
CUDLA_PIXEL_FORMAT_Y12___U12V12_N444 32U
-
CUDLA_PIXEL_FORMAT_Y12___V12U12_N444 33U
-
CUDLA_PIXEL_FORMAT_Y16___U16V16_N444 34U
-
CUDLA_PIXEL_FORMAT_Y16___V16U16_N444 35U
-
CUDLA_PIXEL_FORMAT_Y8___U8V8_N444 28U
-
CUDLA_PIXEL_FORMAT_Y8___V8U8_N444 29U
-
CUDLA_PIXEL_MAPPING_BLOCK_LINEAR 1U
-
CUDLA_PIXEL_MAPPING_PITCH_LINEAR 0U
-
CUDLA_RUNTIME_TENSOR_DESC_NAME_MAX_LEN 80U
-
CUDLA_VER_MAJOR 1U
-
CUDLA_VER_MINOR 5U
-
CUDLA_VER_PATCH 0U
1.1.2. Enumerations
-
enum cudlaAccessPermissionFlags
-
Access permission flags for importing NvSciBuffers.
Values:
-
enumerator CUDLA_READ_WRITE_PERM
-
Flag to import memory with read-write permission.
-
enumerator CUDLA_READ_ONLY_PERM
-
Flag to import memory with read-only permission.
-
enumerator CUDLA_TASK_STATISTICS
-
Flag to indicate buffer as layerwise statistics buffer.
-
enumerator CUDLA_READ_WRITE_PERM
-
enum cudlaDevAttributeType
-
Device attribute type.
Values:
-
enumerator CUDLA_UNIFIED_ADDRESSING
-
Flag to check for support for UVA.
-
enumerator CUDLA_DEVICE_VERSION
-
Flag to check for DLA HW version.
-
enumerator CUDLA_UNIFIED_ADDRESSING
-
enum cudlaFenceType
-
Supported fence types.
Values:
-
enumerator CUDLA_NVSCISYNC_FENCE
-
NvSciSync fence type for EOF.
-
enumerator CUDLA_NVSCISYNC_FENCE_SOF
-
enumerator CUDLA_NVSCISYNC_FENCE
-
enum cudlaMode
-
Device creation modes.
Values:
-
enumerator CUDLA_CUDA_DLA
-
Hyrbid mode.
-
enumerator CUDLA_STANDALONE
-
Standalone mode.
-
enumerator CUDLA_CUDA_DLA
-
enum cudlaModuleAttributeType
-
Module attribute types.
Values:
-
enumerator CUDLA_NUM_INPUT_TENSORS
-
Flag to retrieve number of input tensors.
-
enumerator CUDLA_NUM_OUTPUT_TENSORS
-
Flag to retrieve number of output tensors.
-
enumerator CUDLA_INPUT_TENSOR_DESCRIPTORS
-
Flag to retrieve all the input tensor descriptors.
-
enumerator CUDLA_OUTPUT_TENSOR_DESCRIPTORS
-
Flag to retrieve all the output tensor descriptors.
-
enumerator CUDLA_NUM_OUTPUT_TASK_STATISTICS
-
Flag to retrieve total number of output task statistics buffer.
-
enumerator CUDLA_OUTPUT_TASK_STATISTICS_DESCRIPTORS
-
Flag to retrieve all the output task statistics descriptors.
-
enumerator CUDLA_NUM_INPUT_TENSORS
-
enum cudlaModuleLoadFlags
-
Module load flags for cudlaModuleLoadFromMemory.
Values:
-
enumerator CUDLA_MODULE_DEFAULT
-
Default flag.
-
enumerator CUDLA_MODULE_ENABLE_FAULT_DIAGNOSTICS
-
Flag to load a module that is used to perform permanent fault diagnostics for DLA HW.
-
enumerator CUDLA_MODULE_DEFAULT
-
enum cudlaNvSciSyncAttributes
-
cuDLA NvSciSync attributes.
Values:
-
enumerator CUDLA_NVSCISYNC_ATTR_WAIT
-
Wait attribute.
-
enumerator CUDLA_NVSCISYNC_ATTR_SIGNAL
-
Signal attribute.
-
enumerator CUDLA_NVSCISYNC_ATTR_WAIT
-
enum cudlaScratchMemoryConfig
-
Scratch-memory configuration for cudlaCreateDevice, OR-ed with the device mode.
Bit 0 of ‘flags’ selects the cudlaMode; the scratch-memory configuration occupies bit 1 onward.
Values:
-
enumerator CUDLA_SCRATCH_MEMORY_DEFAULT
-
enumerator CUDLA_SCRATCH_MEMORY_SHARED_STATIC
-
enumerator CUDLA_SCRATCH_MEMORY_CONFIG_MAX
-
enumerator CUDLA_SCRATCH_MEMORY_DEFAULT
-
enum cudlaStatus
-
Error codes.
Values:
-
enumerator cudlaSuccess
-
The API call returned with no errors.
-
enumerator cudlaErrorInvalidParam
-
This indicates that one or more parameters passed to the API is/are incorrect.
-
enumerator cudlaErrorOutOfResources
-
This indicates that the API call failed due to lack of underlying resources.
-
enumerator cudlaErrorCreationFailed
-
This indicates that an internal error occurred during creation of device handle.
-
enumerator cudlaErrorInvalidAddress
-
This indicates that the memory object being passed in the API call has not been registered before.
-
enumerator cudlaErrorOs
-
This indicates that an OS error occurred.
-
enumerator cudlaErrorCuda
-
This indicates that there was an error in a CUDA operation as part of the API call.
-
enumerator cudlaErrorUmd
-
This indicates that there was an error in the DLA runtime for the API call.
-
enumerator cudlaErrorInvalidDevice
-
This indicates that the device handle passed to the API call is invalid.
-
enumerator cudlaErrorInvalidAttribute
-
This indicates that an invalid attribute is being requested.
-
enumerator cudlaErrorIncompatibleDlaSWVersion
-
This indicates that the underlying DLA runtime is incompatible with the current cuDLA version.
-
enumerator cudlaErrorMemoryRegistered
-
This indicates that the memory object is already registered.
-
enumerator cudlaErrorInvalidModule
-
This indicates that the module being passed is invalid.
-
enumerator cudlaErrorUnsupportedOperation
-
This indicates that the operation being requested by the API call is unsupported.
-
enumerator cudlaErrorNvSci
-
This indicates that the NvSci operation requested by the API call failed.
-
enumerator cudlaErrorDriverNotFound
-
This indicates that the dependent driver is not found or loaded.
-
enumerator cudlaErrorDlaErrInvalidInput
-
DLA HW Error.
-
enumerator cudlaErrorDlaErrInvalidPreAction
-
DLA HW Error.
-
enumerator cudlaErrorDlaErrNoMem
-
DLA HW Error.
-
enumerator cudlaErrorDlaErrProcessorBusy
-
DLA HW Error.
-
enumerator cudlaErrorDlaErrTaskStatusMismatch
-
DLA HW Error.
-
enumerator cudlaErrorDlaErrEngineTimeout
-
DLA HW Error.
-
enumerator cudlaErrorDlaErrDataMismatch
-
DLA HW Error.
-
enumerator cudlaErrorUnknown
-
This indicates that an unknown error has occurred.
-
enumerator cudlaSuccess
-
enum cudlaSubmissionFlags
-
Task submission flags for cudlaSubmitTask.
Values:
-
enumerator CUDLA_SUBMIT_NOOP
-
Flag to specify that the submitted task must be bypassed for execution.
-
enumerator CUDLA_SUBMIT_SKIP_LOCK_ACQUIRE
-
Flag to specify that the global lock acquire must be skipped.
-
enumerator CUDLA_SUBMIT_DIAGNOSTICS_TASK
-
Flag to specify that the submitted task is to run permanent fault diagnostics for DLA HW.
-
enumerator CUDLA_SUBMIT_NOOP
1.1.3. Typedefs
-
typedef enum cudlaAccessPermissionFlags cudlaAccessPermissionFlags
-
typedef enum cudlaDevAttributeType cudlaDevAttributeType
-
typedef struct cudlaDevHandle_t *cudlaDevHandle
-
cuDLA Device Handle
-
typedef enum cudlaFenceType cudlaFenceType
-
typedef enum cudlaMode cudlaMode
-
typedef struct cudlaModule_t *cudlaModule
-
cuDLA Module Handle
-
typedef enum cudlaModuleAttributeType cudlaModuleAttributeType
-
typedef enum cudlaModuleLoadFlags cudlaModuleLoadFlags
-
typedef enum cudlaNvSciSyncAttributes cudlaNvSciSyncAttributes
-
typedef enum cudlaScratchMemoryConfig cudlaScratchMemoryConfig
-
typedef enum cudlaStatus cudlaStatus
-
typedef enum cudlaSubmissionFlags cudlaSubmissionFlags