TI Deep Learning Product User Guide
TIDL_bufParams3D_t Struct Reference

Data Fields

uint32_t data_type
 
uint32_t dim_x
 
uint32_t dim_y
 
int32_t stride_y
 
uint32_t dim_z
 
int32_t stride_z
 
float32_tidltensorScaleList
 
int32_t * tensorZeroPointList
 

Field Documentation

◆ data_type

uint32_t TIDL_bufParams3D_t::data_type

Data type of input TIDL_data_type_e

◆ dim_x

uint32_t TIDL_bufParams3D_t::dim_x

Width of buffer in X dimension in elements. This corresponds to width of the tensor which is going to be processed in a given kernel call

◆ dim_y

uint32_t TIDL_bufParams3D_t::dim_y

Height of buffer in Y dimension in elements. This corresponds to height of the tensor which is going to be processed in a given kernel call

◆ stride_y

int32_t TIDL_bufParams3D_t::stride_y

Stride in Y dimension in bytes.

◆ dim_z

uint32_t TIDL_bufParams3D_t::dim_z

Depth of patch in Z dimension in elements.This corresponds to number of channels which is going to be processed in a given kernel call

◆ stride_z

int32_t TIDL_bufParams3D_t::stride_z

Stride in Z dimension in bytes.

◆ tensorScaleList

float32_tidl* TIDL_bufParams3D_t::tensorScaleList

Pointer to the array containing the tensor's tensor scales on a per channel basis, if quantization is per-tensor only one value is populated

◆ tensorZeroPointList

int32_t* TIDL_bufParams3D_t::tensorZeroPointList

Pointer to the array containing the tensor's zero points on a per channel basis, if quantization is per-tensor only one value is populated