Jacobi symbol using Lehmer's algorithm.

bodrato at mail.dm.unipi.it bodrato at mail.dm.unipi.it
Thu Jan 28 20:31:21 CET 2010


> I would suggest to make use of the --enable-minithres mechanism, and
> then stick to operands that are not terribly large.

I assume Niels already did that. There are minimal thresholds defined in
his code, and one can use the current implementation as a reference for
small operands.

Comparing two sub-quadratic implementations, one can test larger (not
terribly) operands, where many iterations are needed. There also is a
bonus, I'm testing two new algorithms at once: Paul's and Niels'!

Anyway, no news on that side: they keep on agreeing on any test!
I'll try another range.

I know that this does not prove anything, one should not use random
operands, but generate some cleverly selected special values... Let's say
that this is a first level of tests, and both code seem very solid.



More information about the gmp-devel mailing list