Pipelined signed multipliers

Hello,

I am trying to design a signed multiplier and I want to make it pipelined so I can improve its performance. I am experimenting with a Baugh-wooley multiplier but I am not quite sure how to go about this.

How is it possible to specify the number of pipelining stages in the multiplier or do I have to add registers at every input and output?

Any suggestions for this or other signed high performance multipliers is appreciated.

Thanks
Kaf

khodorf