Re: [Fastlibnt-devel] _Zpoly_mpn_add

[William H. <ha...@ya...>]

I got _Zpoly_mpn_add working in all cases with very
stringent test code. 

It allows the longer input polynomial to also be the
output polynomial, it allows all different lengths,
all different coefficient sizes, accomodates positive,
zero and negative coefficients.

It is relatively efficient I think, though some of the
branches can probably be replaced with some tricky
bitwise logic.

Because _Zpoly_mpn_add has to deal with negative
coefficients, _Zpoly_mpn_sub probably isn't that much
more difficult to do.

I think these functions are pretty hopelessly slow
though. There are so many unpredictable branches
related to the various possibilities for signs of
coefficients. I think we need some additional versions
which make assumptions about the lengths being the
same and the number of coefficient bits being the same
and the coefficients all being positive or something.

Bill.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com