How to Handle conv + bias in conv_3x3 for input(int8) * weight(int8) + bias(int32)?

hi @moimfeld 
I am currently working with the conv_3x3 function in the RISC-V assembly code, and I have a question regarding the incorporation of bias into the convolution operation. Specifically, I want to understand how to correctly handle the conv + bias operation where the input and weights are int8 and the bias is int32.