FermatPRP vs. M-R

James Wanless bearnol at gmail.com
Wed Mar 16 11:17:53 CET 2011

(Seems I sent this from my wrong email address first :)

In case you're wondering ;-)
M-R calculates its probabilities in the wrong (reverse) direction to what
one needs.
And as anyone who knows Bayes' theorem understands, if something works
better in the wrong/reverse direction, then that means it will work _worse_
in the correct direction.
(Incidentally the same applies to the results of recent tests into

On 16 Mar 2011, at 09:37, James Wanless wrote:

Hi all,

[Congrats from me too on GMP's recent birthday!]

Is there a (simple) FermatPRP currently implemented in GMP?

I ask because FermatPRP is more accurate than M-R.

[I know one can easily code one oneself, but that's not the point :)]


More information about the gmp-discuss mailing list