Re: SPRP's, and miller(n,k)

[Gerhard Niklasch <nikl@mathematik.tu-muenchen.de>]

P.S.:
> I have what I believe to be the complete list of 42 composites up
> to 3*2^46 which would (wrongly) pass the miller() test for the four
> bases chosen at k==16.  Without Atkin's `end-matching' check, there
> would be 86 of them.

I should have added that the search ran over composites not divisible
by any prime <100  (and that a few single-word gcds will take care of
composites with small prime divisors: this should be done before
calling miller()).

Gerhard