MMALIB User Guide
MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_ExecInArgs Struct Reference

Detailed Description

Structure containing the parameters for input to the execute phase of CNN convolution computation These parameters will not exist in J7AM, kept for J7ES compatibility.

Definition at line 194 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

Data Fields

int32_t validColsIn
 Valid columns in a row of input feature maps for one call of processing of the kernel for non strided convolution and for 1x1 stride 2 for execute phase
More...
 
int32_t validColsPerRowIn
 Valid columns in a row of input feature maps for one call of processing of the kernel for strided convolution except for 1x1 stride 2 for execute phase
More...
 
int32_t validRowsIn
 Valid input rows of input feature maps for one call of processing of the kernel for strided convolution except for 1x1 stride 2 for execute phase
More...
 
int32_t col
 column offset of each row of input feature map enables with a flag in execute phase
More...
 
int32_t subMChannels
 Number of output channels per kernel call in execute phase
More...
 
int32_t quantMethod
 quantMethod 0 - per call 1 - per channel
More...
 
int32_t padFillValue
 pad fill value
More...
 
MMALIB_enableDynamicRange_type enableDynamicRange
 Flag indicating whether to enable dynamic range range (min,max) calculation. More...
 
MMALIB_initDynamicRange_type initDynamicRange
 Flag indicating whether to reset dynamic range (min,max) calculation at beginning of compute. More...
 

Field Documentation

◆ validColsIn

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_ExecInArgs::validColsIn

Valid columns in a row of input feature maps for one call of processing of the kernel for non strided convolution and for 1x1 stride 2 for execute phase

Definition at line 199 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ validColsPerRowIn

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_ExecInArgs::validColsPerRowIn

Valid columns in a row of input feature maps for one call of processing of the kernel for strided convolution except for 1x1 stride 2 for execute phase

Definition at line 203 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ validRowsIn

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_ExecInArgs::validRowsIn

Valid input rows of input feature maps for one call of processing of the kernel for strided convolution except for 1x1 stride 2 for execute phase

Definition at line 207 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ col

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_ExecInArgs::col

column offset of each row of input feature map enables with a flag in execute phase

Definition at line 210 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ subMChannels

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_ExecInArgs::subMChannels

Number of output channels per kernel call in execute phase

Definition at line 212 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ quantMethod

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_ExecInArgs::quantMethod

quantMethod 0 - per call 1 - per channel

Definition at line 214 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ padFillValue

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_ExecInArgs::padFillValue

pad fill value

Definition at line 216 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ enableDynamicRange

MMALIB_enableDynamicRange_type MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_ExecInArgs::enableDynamicRange

Flag indicating whether to enable dynamic range range (min,max) calculation.

Definition at line 218 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ initDynamicRange

MMALIB_initDynamicRange_type MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_ExecInArgs::initDynamicRange

Flag indicating whether to reset dynamic range (min,max) calculation at beginning of compute.

Definition at line 220 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.