Factorcyc

[ATH]

Can anyone please compile akruppa's program factorcyc for windows? I prefer Windows 64 bit, but 32bit will do:

http://www.mersenneforum.org/showpos...&postcount=149

[Brian Gladman]

Originally Posted by ATH

Can anyone please compile akruppa's program factorcyc for windows? I prefer Windows 64 bit, but 32bit will do:

http://www.mersenneforum.org/showpos...&postcount=149

This is possible but it will be quite hard to do becaause it contains x64 inline assembler sections without C or 32-bit assembler alternatives and this means it cannot be compiled in 32-bit mode.

Moreover, the inline assembler is in GCC format so it will need either a 64-bit GCC on Windows or the Intel compiler because the MSVC x64 compiler does not offer inline assembler capabilities.

The Intel compiler should succeed and Mingw64 might also work but I am not sure whether it can now compile GMP (or MPIR) - it failed last time I tried but this was some time ago and it is now more mature.

I have built GMP 4.2.x and MPIR using the Intel compiler but GMP 4.3 with fast assembler code support won't work on Windows x64 because the assembler code is only available using the Linux ABI conventions.

So it is very likely that this will need either the mingw64 GCC or the Intel compilers with MPIR or GMP 4.2.x (rather than GMP 4.3) if speed is important.

Brian

[ATH]

Last time I tried I couldn't get Mingw64 to compile GMP, guess I'll try again with new GMP version.

[akruppa]

There's a newer version at http://www.loria.fr/~kruppaal/factorcyc.20090612.c
It's faster (the ridiculous inversion modulo 2^64 was replaced by lifting) and primes used for sieving are not skipped in the trial divison any more. Note: if your divisors are <2^64, GMP is not used for the trial division (only for setup) so you can use any GMP version with no effect on speed.

Alex