[RC5]V3...Lets beat the GIMPS effort

Harijs Ausmanis harijs at parks.lv
Mon Aug 31 17:34:03 EDT 1998


However, George's code is completely Pentium-optimized, it runs very slowly
on other X86 CPUs, also, the codes for other platforms they run GIMPS on -
Mac, Alpha - are far from being optimized, I seem to remmember that the last
release of the client for Alpha improved speed by 200%. A good point of
d.net is that they have not only Intel, but also AMD, Cyrix and non-X86
optimized cores for every contest that they have participated in.

>There are good reasons that hasn't happened yet; George's code is the
>end result of several years of constant optimization (it's on version
>16.xx right now). His FFT multiply is insanely complicated, and the source
>code for it contains about 200kB of assembly code.

>I've been working on a high-performance multiply, off and on, for the last
>few months. If it gets to within a factor of two of George's code I'll
>be surprised.

>jasonp

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list