sTIDLRT_Tensor_t Struct Reference

Data Fields
int8_t	name [TIDLRT_STRING_SIZE]
int32_t	elementType
int32_t	numDim
int32_t	dimValues [TIDLRT_DIM_MAX]
int32_t	pitch [TIDLRT_DIM_MAX-1]
int32_t	padValues [TIDLRT_PAD_DIM_MAX]
void *	ptr
int32_t	dataOffset
uint32_t	layout
int32_t	zeroPoint
float	scale
uint32_t	memType

Field Documentation

name

int8_t sTIDLRT_Tensor_t::name[TIDLRT_STRING_SIZE]

Unique name for the tensor holding feature vector , When we have multiple tensors as inputs, the tensors in the imported model and runtime may not be in same order. This name be part Model IO buffer descriptor file. This name would help in associating each tensor to model input

elementType

int32_t sTIDLRT_Tensor_t::elementType

Element type of the feature vector as defined in eTIDLRT_ElementType

numDim

int32_t sTIDLRT_Tensor_t::numDim

Number of dimensions supported by the tensor

dimValues

int32_t sTIDLRT_Tensor_t::dimValues[TIDLRT_DIM_MAX]

Values of dimensions as defined by eTIDLRT_DataDimIndex

pitch

int32_t sTIDLRT_Tensor_t::pitch[TIDLRT_DIM_MAX-1]

Pitch for each dimension as defined by eTIDL_PitchDimIndex

padValues

int32_t sTIDLRT_Tensor_t::padValues[TIDLRT_PAD_DIM_MAX]

Padding available in Spatial Dimension Left, right, top, Bot

ptr

void* sTIDLRT_Tensor_t::ptr

Pointer to the buffer Base address

dataOffset

int32_t sTIDLRT_Tensor_t::dataOffset

Actual tensor data start offset in elements from the base address – Excluding the padding

layout

uint32_t sTIDLRT_Tensor_t::layout

Layout of the data NCHW or NHWC define in eTIDLRT_TensorLayout

zeroPoint

int32_t sTIDLRT_Tensor_t::zeroPoint

Zero point for Asymmetric Fixed point Representation

scale

float sTIDLRT_Tensor_t::scale

Scale used for float to fixed point conversion

memType

uint32_t sTIDLRT_Tensor_t::memType

memory type – Linux user space memory or Shared memory define in eTIDLRT_MemType