Quantized multiplication framework

A framework to accelerate Quantized multiplication in Synet Framework. More...

Functions
SIMD_API void *	SimdSynetQuantizedMulInit (const size_t *aShape, size_t aCount, SimdTensorDataType aType, const float *aScale, int32_t aZero, const size_t *bShape, size_t bCount, SimdTensorDataType bType, const float *bScale, int32_t bZero, SimdTensorDataType dstType, const float *dstScale, int32_t dstZero)
	Initializes element-wise quantized multiplication of two tensors. More...
SIMD_API void	SimdSynetQuantizedMulForward (void *context, const uint8_t *a, const uint8_t *b, uint8_t *dst)
	Performs element-wise quantized multiplication. More...

Detailed Description

A framework to accelerate Quantized multiplication in Synet Framework.

Function Documentation

SimdSynetQuantizedMulInit()

void * SimdSynetQuantizedMulInit	(	const size_t *	aShape,
		size_t	aCount,
		SimdTensorDataType	aType,
		const float *	aScale,
		int32_t	aZero,
		const size_t *	bShape,
		size_t	bCount,
		SimdTensorDataType	bType,
		const float *	bScale,
		int32_t	bZero,
		SimdTensorDataType	dstType,
		const float *	dstScale,
		int32_t	dstZero
	)

Initializes element-wise quantized multiplication of two tensors.

The current implementation supports compatible input shapes (equal or broadcast). For each element it dequantizes UINT8 inputs as (value - zero)*scale, multiplies the two values and converts the result to FP32 or UINT8 output. FP32 inputs and outputs ignore the corresponding quantization zero.

Note

This function has a C++ wrapper: Simd::SynetQuantizedMul.

Parameters

[in]	aShape	- a pointer to shape of input A tensor.
[in]	aCount	- a count of dimensions of input A tensor.
[in]	aType	- a type of input A tensor. It can be SimdTensorData32f or SimdTensorData8u.
[in]	aScale	- a pointer to quantization scale of input A tensor. Can be NULL (scale is 1.0).
[in]	aZero	- a quantization zero of input A tensor.
[in]	bShape	- a pointer to shape of input B tensor.
[in]	bCount	- a count of dimensions of input B tensor.
[in]	bType	- a type of input B tensor. It can be SimdTensorData32f or SimdTensorData8u.
[in]	bScale	- a pointer to quantization scale of input B tensor. Can be NULL (scale is 1.0).
[in]	bZero	- a quantization zero of input B tensor.
[in]	dstType	- a type of output tensor. It can be SimdTensorData32f or SimdTensorData8u.
[in]	dstScale	- a pointer to output quantization scale. Can be NULL (scale is 1.0).
[in]	dstZero	- an output quantization zero.

Returns

a pointer to quantized multiplication context. On error it returns NULL. It must be released with using of function SimdRelease. This pointer is used in function SimdSynetQuantizedMulForward.

SimdSynetQuantizedMulForward()

void SimdSynetQuantizedMulForward	(	void *	context,
		const uint8_t *	a,
		const uint8_t *	b,
		uint8_t *	dst
	)

Performs element-wise quantized multiplication.

Algorithm's details for UINT8 output:

for(i = 0; i < size; ++i)
{
    _a = (a[i] - aZero)*aScale;
    _b = (b[i] - bZero)*bScale;
    dst[i] = RestrictRange(Round((_a * _b)/dstScale) + dstZero, 0, 255);
}

Note

This function has a C++ wrapper: Simd::SynetQuantizedMul.

Parameters

[in]	context	- a pointer to quantized multiplication context. It must be created by function SimdSynetQuantizedMulInit and released by function SimdRelease.
[in]	a	- a pointer to input A tensor data. Its type is defined by parameter aType of SimdSynetQuantizedMulInit.
[in]	b	- a pointer to input B tensor data. Its type is defined by parameter bType of SimdSynetQuantizedMulInit.
[out]	dst	- a pointer to output tensor data. Its type is defined by parameter dstType of SimdSynetQuantizedMulInit.