miroslavkures wrote:Very nice. So, we use two tests parallelly: the complete test and the periodic test. The complete one tests cca 300-700 primes per second. Simultaneously, the periodic test implements the conjecture that binary expressions of Wieferich primes are periodic. The speed is multiplied by number of cores. As to Your testing: unpublished results are suspicious - although Your near W.p. is OK. Nevertheless, our strategy is not a linear searching; we search already for primes greater than 6*10^15 in one of four ranges. More on

www.elmath.org.

If You have more questions or even want cooperate, You can also use my e-mail

kures@fme.vutbr.cz.

The results are unpublished as I provided the software and some of the horsepower for someone else. They have the responsibility to publish the results,. which I why I won't reveal more information than what I know.

So if I understand correctly, you ignore certain primes because they do not periodic. But that is based upon an unproven conjecture. Your results are likely to be incomplete, especially WRT near Wieferichs.

300-700 tests per second? That speed seems slow to me. I would expect at least 50,000 per second per core on a Core 2 Duo at 2.4 GHz. Is the expmod written in C only or is it written in assembler?