Side channel silent karatsuba / mpn_addmul_2 karatsuba
bodrato at mail.dm.unipi.it
Thu Dec 13 19:36:02 UTC 2018
Il Gio, 13 Dicembre 2018 1:07 pm, Torbjörn Granlund ha scritto:
> nisse at lysator.liu.se (Niels Möller) writes:
> It will be very interesting to see what thresholds we get with that.
> Marco's code might be in shape to enlighten us about that!
I'll try it again as soon as I'll have the time :-)
> I am still playing in C with longlong.h, but I will likely try creating
Are you willing to add a mul_2 to longlong.h?
I feel we should at least add an umulhi macro.
We have too many umul_ppmm(used, dummy, a, b) in the code now...
Sometimes in the past we also played with an almost-branch-free C-only
umul_ppmm using karatsuba... but I don't remember when...
More information about the gmp-devel