Application/OS/CPU Architecture matrix for YAFU

[Dirk Broer]

Application
x86
x86-64


YAFU 134.01 (mt)


Linux/64


YAFU 134.05 (mt)

Windows 98 or later

Windows/64


YAFU for small composites 134.01 (mt)


Linux/64


YAFU for small composites 134.05 (mt)

Windows 98 or later

Windows/64

YAFU-4t 134.01 (4t)


Linux/64


YAFU-4t 134.05 (4t)

Windows 98 or later

Windows/64


YAFU-8t 134.01 (8t)


Linux/64


YAFU-8t 134.05 (8t)

Windows 98 or later

Windows/64


YAFU-16t 134.01 (16t)


Linux/64


YAFU-16t 134.05 (16t)

Windows 98 or later

Windows/64


YAFU-32t 134.01 (32t)


Linux/64


YAFU-32t 134.05 (32t)

Windows 98 or later

Windows/64

[NeoGen]

Nice table Dirk. I knew the project was multi-threaded but never looked at that tiny detail... they actually have specific apps for 4, 8 and 16 threads? I can't see the purpose of it unless there are optimizations in play for those specific values... otherwise the general multi-threaded should work fine and just as well.

Is it known if these applications require CPUs that support specific instructions sets like SSE4, FMA, or AVX? Being a mathematical factorization project they probably would benefit from it.

[Dirk Broer]

My guess is that they came up with the 4, 8 and 16 thread apps for people to be able to choose how much cores they would like to spend on the project, without app_config.xml hussle.
Nowhere do they mention the advantages of certain instruction sets, one can only hope that the app is so smart to choose the best path itself.