We made a number of additional changes to the program. * It now works (again) without longlong.h. We provide this option for code simplicity, at the expensive of performance.
* A new command line option `-w' enables weak primes testing. This is actually often a slowdown since 25 Miller-Rabin tests are currently often slower than the default of a combination of Miller-Rabin tests and Lucas tests. It will surely speed up some cases, in particular in the GMP range. * Speedup for prime_p by computing redcify of used bases more efficiently. * Cleanup and more comments. Pádraig's example should now run at about 40s on his machine. I believe we could make things about 50% faster for numbers < 2^128, mainly by improving powm, and by being more clever about how to compute the powers in Lucas. We could also speed Pollard rho by reducing the gcd call frequency. In the GMP range, there is more headroom. We can leave such improvements for the future. -- Torbjörn
