MMALIB User Guide
MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs Struct Reference

Detailed Description

Structure containing the parameters initialization of CNN convolution computation.

Definition at line 99 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

Data Fields

int8_t funcStyle
 Variant of the function refer to MMALIB_FUNCTION_STYLE
More...
 
int32_t No
 Number of output feature maps
More...
 
int32_t inChOffset
 offset of the input feature maps in B matrix. This is power of 2 for circular buffer and atleast 64 byte aligned for linear buffer
More...
 
int32_t validColsIn
 Valid columns of input feature maps in B matrix for one call of processing of the kernel for non strided convolution but including 1x1 stride 2. This is in units of data type
More...
 
int32_t validColsPerRowIn
 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. This will not be used in J7AM instead use inWidth
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
More...
 
int32_t inputPitchPerRow
 Valid pitch for each input rows of input feature maps for strided convolution. This is units of bytes
More...
 
int32_t outputPitchPerRow
 Valid output pitch for each output rows of output feature maps for strided convolution. This is units of bytes
More...
 
int32_t inWidth
 Width of each row of input feature map in units of data type
More...
 
int32_t pad
 Pad of each row of input feature map, specify pad only one one side. specific to J7ES
More...
 
int32_t padLeft
 Pad of each row of input feature map, specify pad only on left side
More...
 
int32_t padRight
 Pad of each row of input feature map, specify pad only on right side
More...
 
int32_t padTop
 Pad of each row of input feature map, specify pad only on top side
More...
 
int32_t padBottom
 Pad of each row of input feature map, specify pad only on bottom side
More...
 
int32_t validColsOutBottom
 Output pixels for stride 1 flow when padBotton > 0
More...
 
int32_t maxHeight
 Height of the input feature map in units of data type
More...
 
int32_t subMChannels
 Number of output channels per kernel call
More...
 
int32_t numGroupsPerKernel
 Scale parameter for output precision ‍/ uint8_t scale; /! More...
 
int32_t Fr
 coefficient rows (height)
More...
 
int32_t Fc
 coefficient columns (width)
More...
 
int32_t strideX
 stride of columns
More...
 
int32_t strideY
 stride of rows
More...
 
int32_t dilationX
 dilation of coefficients of columns
More...
 
int32_t dilationY
 dilation of coefficients of rows
More...
 
int32_t bias
 bias value in B matrix same as data type of B matrix
More...
 
uint8_t activationType
 activation RELU, SAT or none for output
More...
 
int32_t pSatMin
 Min value for parameterized saturation
More...
 
int32_t pSatMax
 Max value for parameterized saturation
More...
 
uint8_t mode
 mode for input feature map in Circular or Linear mode in B matrix
More...
 
int32_t col
 column number of input feature map
More...
 
int8_t packetizeMode
 packetizeMode is 0 then weights are not packetized. When 1 then they are packetized
More...
 

Field Documentation

◆ funcStyle

int8_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::funcStyle

Variant of the function refer to MMALIB_FUNCTION_STYLE

Definition at line 102 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ No

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::No

Number of output feature maps

Definition at line 104 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ inChOffset

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::inChOffset

offset of the input feature maps in B matrix. This is power of 2 for circular buffer and atleast 64 byte aligned for linear buffer

Definition at line 107 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ validColsIn

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::validColsIn

Valid columns of input feature maps in B matrix for one call of processing of the kernel for non strided convolution but including 1x1 stride 2. This is in units of data type

Definition at line 111 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ validColsPerRowIn

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::validColsPerRowIn

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. This will not be used in J7AM instead use inWidth

Definition at line 115 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ validRowsIn

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::validRowsIn

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

Definition at line 119 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ inputPitchPerRow

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::inputPitchPerRow

Valid pitch for each input rows of input feature maps for strided convolution. This is units of bytes

Definition at line 122 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ outputPitchPerRow

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::outputPitchPerRow

Valid output pitch for each output rows of output feature maps for strided convolution. This is units of bytes

Definition at line 125 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ inWidth

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::inWidth

Width of each row of input feature map in units of data type

Definition at line 127 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ pad

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::pad

Pad of each row of input feature map, specify pad only one one side. specific to J7ES

Definition at line 130 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ padLeft

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::padLeft

Pad of each row of input feature map, specify pad only on left side

Definition at line 133 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ padRight

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::padRight

Pad of each row of input feature map, specify pad only on right side

Definition at line 136 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ padTop

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::padTop

Pad of each row of input feature map, specify pad only on top side

Definition at line 139 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ padBottom

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::padBottom

Pad of each row of input feature map, specify pad only on bottom side

Definition at line 142 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ validColsOutBottom

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::validColsOutBottom

Output pixels for stride 1 flow when padBotton > 0

Definition at line 144 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ maxHeight

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::maxHeight

Height of the input feature map in units of data type

Definition at line 146 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ subMChannels

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::subMChannels

Number of output channels per kernel call

Definition at line 148 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ numGroupsPerKernel

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::numGroupsPerKernel

Scale parameter for output precision ‍/ uint8_t scale; /!

Shift parameter for output precision ‍/ int32_t shift; /! number of groups per kernel call > 1 will enable processing when No <= MMA size for non strided kernels and default value is 1

Definition at line 155 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ Fr

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::Fr

coefficient rows (height)

Definition at line 157 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ Fc

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::Fc

coefficient columns (width)

Definition at line 159 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ strideX

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::strideX

stride of columns

Definition at line 161 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ strideY

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::strideY

stride of rows

Definition at line 163 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ dilationX

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::dilationX

dilation of coefficients of columns

Definition at line 165 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ dilationY

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::dilationY

dilation of coefficients of rows

Definition at line 167 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ bias

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::bias

bias value in B matrix same as data type of B matrix

Definition at line 169 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ activationType

uint8_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::activationType

activation RELU, SAT or none for output

Definition at line 171 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ pSatMin

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::pSatMin

Min value for parameterized saturation

Definition at line 173 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ pSatMax

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::pSatMax

Max value for parameterized saturation

Definition at line 175 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ mode

uint8_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::mode

mode for input feature map in Circular or Linear mode in B matrix

Definition at line 178 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ col

int32_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::col

column number of input feature map

Definition at line 180 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.

◆ packetizeMode

int8_t MMALIB_CNN_convolveBias_row_ixX_ixX_oxX_InitArgs::packetizeMode

packetizeMode is 0 then weights are not packetized. When 1 then they are packetized

Definition at line 184 of file MMALIB_CNN_convolveBias_row_ixX_ixX_oxX.h.