Ciao Paul,
Il 2022-03-07 10:28 Paul Zimmermann ha scritto:
Date: Sun, 06 Mar 2022 11:14:49 +0100
From: Marco Bodrato
Specifically I'd focus into a suspicious piece of code, shared by both
our current code and Vivien's.
if (cy >= 0)
cy = mpn_add_1 (tmp, tmp, Kl, cy);
Hi Marco,
> Since you deeply know how this code works, I ask you one more question.
> The last lines of the function mpn_fft_mul_2exp_modF (branch m < n)
> contain:
>
>/* now subtract cc and rd from r[m..n] */
>
>r[n] = -mpn_sub_1 (r + m, r + m, n - m, cc);
>r[n] -= mpn_sub_1
Ciao Paul,
Il 2022-03-08 12:56 Paul Zimmermann ha scritto:
Since you deeply know how this code works, I ask you one more
question.
The last lines of the function mpn_fft_mul_2exp_modF (branch m < n)
contain:
/* now subtract cc and rd from r[m..n] */
r[n] = -mpn_sub_1 (r + m, r + m, n -
Dear Marco,
> Uhm, the last line of code just before that ones is:
>
>cc = mpn_sub_1 (r, r, m, cc) + 1;
>/* add 1 to cc instead of rd since rd might overflow */
>
> So that I'd say that cc is 1 or 2.
> Then the case cc=2, m=n-1, r[m]=0, and rd=GMP_NUMB_MAX seems very very
> unlik
Ciao Paul,
Il 2022-03-08 16:20 Paul Zimmermann ha scritto:
Uhm, the last line of code just before that ones is:
cc = mpn_sub_1 (r, r, m, cc) + 1;
/* add 1 to cc instead of rd since rd might overflow */
it seems you are right!
Well, I pushed a clean-up for that portion of the code too
