Toom-Cook methods and even long integers

[bodrato at mail.dm.unipi.it]

Hi everybody,

> Actually, for Toom-3 it seems that possible optimizations are not very
> effective, particularly when specific basic operations (A -= B/2) are

Early this year I experimented a slight optimization for Toom-3, saving
half an operation (the code for GMP 4.2.2 is available on my web site,
http://bodrato.it/software/toom.html#TC3 ). Unfortunately this
optimisation is in conflict with the even-Toom-3 proposed by Alberto...

> Toom-3.5, for unbalanced operands, seems instead to be a bit more
> promising, even if more investigation is needed.
>
> http://bodrato.it/papers/zanoni.html#CIVV2009

...on the other side the new sequences for the even-Toom-3.5 are perfectly
compatible with the strategy saving two half-operations used by the code I
announced one month ago. One more reason to start investigations from the
Toom-and-a-half.

Regards,
Marco

-- 
http://bodrato.it/toom-cook/