Bug in Mul.v

[Rubiczhang]

Hello Zhang Wenting:

I have run rv64um isatest on your multiplier, but the result is wrong.

I have noticed that when you calculate middle result (mac_a * mac_b), you simply use a couple of 33 bits signed extension.

I'm not good at mathematic, however, what I have searched on internet told me two w bits signed extension timing together can only yield a correct answer when you cut it in w bits.

So, I think that's where the bug lies.