"balanced" unbalanced toom22

[David Harvey]

On Aug 7, 2008, at 6:54 AM, Paul Zimmermann wrote:

> Thus your results are not surprising. My guess is that you are too  
> near the
> quadratic region, where the modified cutting should not win. Also,  
> toom22
> performs only one unbalanced product, whereas the modified cutting  
> performs
> two, and I believe mpn_mul currently behaves much better for balanced
> multiplication.

Hi Paul,

Do you have any theoretical reason to think that it should win, or  
should not win, for certain sizes (assuming mpn_mul handled the  
unbalanced subproducts better)?

david