I32I8CONV

Convolution with int8 input and int32 output.

Description

Takes an image and a set of filters and performs a convolution using dot-product instruction.

Extension Interface

void i32i8conv(const QML_IMAGE_FORMAT inputFormat, const QML_IMAGE_FORMAT outputFormat,
              const bool biased, const qml_int8* src,
              const qml_int srcX, const qml_int srcY,
              const qml_int channels, const qml_int8* filters,
              const qml_int numFilters, const qml_int filterX, const qml_int filterY,
              const qml_int padLeft, const qml_int padTop,
              const qml_int strideX, const qml_int strideY,
              qml_int* output, const qml_int outputX, const qml_int outputY);

Arguments

inputFormat	Input image format (channel-major or image-major).
outputFormat	Output image format (channel-major or image-major).
biased	Whether or not a bias has already been applied to the output
src	Input image stored in the specified format
srcX	Width of the image
srcY	Height of the image
channels	Number of channels in the image
filters	Filters, assumed to be stored in channel major format
numFilters	Number of filters
filterX	Width of the filters
filterY	Height of the filters
padX	Amount of padding added to the left and right of the image
padY	Amount of padding added to the top and bottom of the image
strideX	Filter stride in the X dimension
strideY	Filter stride in the Y dimension
output	Output image stored in the specified format
outputX	Width of an individual output
outputy	Height of an individual output