GMP 4.4 remaining tasks

bodrato at bodrato at
Tue Dec 8 13:30:24 CET 2009

>   The "misuse" forbidden by the definition above are _very_ degenerate:
>    - (an+bn) < rn/2
> That might happen if we want to do full multiplication, given the nature
> of mpn_mulmod_bnm1_next_size...

Yes, but it _shouldn't_ happen if (an+bn) > 2*MUL_FFT_MODF_THRESHOLD...

Using mpn_mulmod_bnm1 to obtain a full product when the size is too small
to use FFT on the bnp1 part... is a _very_ degenerate "misuse" :-D

> Talking about mpn_mulmod_bnm1_next_size.  My simple-minded rounding to
> the next multiple of 16 for all sizes under MUL_FFT_MODF_THRESHOLD could
> perhaps be trimmed to have some size dependency:

Anyway... (multiple of 16 next to n) > 2*n can not happen if n > 8, can it?


More information about the gmp-devel mailing list