Re: [PrimeNumbers] Comparison of spsp to multiple bases, BPSW and Frobenius with method B*
- --- On Fri, 1/21/11, Brian <btball@...> wrote:
> Timings on a XEON x5570 2.93 Ghz, 32GB ram:Then you're doing something wrong, or your library is optimised for one primitive (modular exponentiation), and horrifically slow for another (lucas sequence evaluation). What precisely was your input set?
> First place = SPSP to multiple bases. The clear winner at 4481.6 seconds
> Second place = Frobenius test at 20,070 seconds
> Third place = BPSW at 22,506 seconds
Are you comparing up-to-12th roots of unity in the multiple SPSP test? You ought to, that will make it a little quicker. (After the 2nd passed test there's a >50% chance that you'll have incompatible square roots of unity, and won't need to do a third test.)