QlinearMacKernels

Collaboration diagram for QlinearMacKernels:

Classes
class	QlinearMacScalar< TT, TTPARAM, B, W, IS_RELU >
	QlinearMac scalar implementation, QlinearMacScalar<14,128,0> 418101 cycles. More...
class	QlinearMac< TT, TTPARAM, B, W, IS_RELU >
	QlinearMac vector implementation, requires W%16 == 0, QlinearMacScalar<14,128,0> 1256 cycles, 2692 for int32 scale. More...

Detailed Description

https://github.com/onnx/onnx/blob/main/docs/Operators.md#Mul https://github.com/onnx/onnx/blob/main/docs/Operators.md#Add https://github.com/onnx/onnx/blob/main/docs/Operators.md#Relu

• y = saturate ((x / y_scale) + y_zero)
• x = (qx - qx_zero) * qx_scale

Computation

• qz = (qx-qx_zero)*qx_scale * (qw-qw_zero)*qw_scale / qz_scale + qz_zero
• qy = (qz-qz_zero)*qz_scale + (qb-qb_zero)*qb_scale / qy_scale + qy_zero