distributed arithmetic symmetry?

Ok, 

I think I understand now. My misunderstanding before was I thought the
summing was done AFTER the input went through the
parrallel-shift-register. But I guess you are saying the inputs are summed
together first, THEN that sum is serially sent into the LUT. Thanks a lot!

dtsao wrote:

> Ok, 
> 
> I think I understand now. My misunderstanding before was I thought the
> summing was done AFTER the input went through the
> parrallel-shift-register. But I guess you are saying the inputs are summed
> together first, THEN that sum is serially sent into the LUT. Thanks a lot!
> 

Yup,  The pre-adders and the LUT inputs are serial.  The output of the 
LUTs are a sum of 1xN partial products, which are then shifted and 
summed in a scaling accumulator over the N clock cycles to arrive at the 
full sum of MxN products.

dtsao wrote:

> Ok, 
> 
> I think I understand now. My misunderstanding before was I thought the
> summing was done AFTER the input went through the
> parrallel-shift-register. But I guess you are saying the inputs are summed
> together first, THEN that sum is serially sent into the LUT. Thanks a lot!
> 

Yup,  The pre-adders and the LUT inputs are serial.  The output of the
LUTs are a sum of 1xN partial products, which are then shifted and
summed in a scaling accumulator over the N clock cycles to arrive at the
full sum of MxN products.


Oh, and I should add that I have a page on my website explaining 
distributed arithmetic.  It doesn't go into the symmetry however, but it 
does go over the basics.  http://www.andraka.com/distribu.htm