On Sunday 27 February 2005 21:18, Arjan Koek wrote: > > Version 24.6 is on the ftp sever (ftp://mersenne.org/gimps/),
Yes it is. Has been for ~ 3 months. > it's only > announced in the software forum (http://www.mersenneforum.org/) since > it's still in beta phase. Could I point out that a mail list, being a "push" communication medium, is a better method of making announcements than posts in a forum - where you will not see the announcement unless you happen to browse the correct thread. > It's major improvement is an 10 till 20% > speed gain on Athlons if I recall correctly. I think I got a bit more than that on my XP1700 system. > And also a small gain on > P3's. No! In fact a significant loss on everything except P4 (also current Celeron & Xeon based systems) (SSE2 code, unchanged) and Athlon. Nathan Russell <[EMAIL PROTECTED]> wrote: > Is the slowdown for the PIII significant enough for you to consider > splitting the code for them and other non-SSE2 686 machines? That's certainly one possibility - but Pentium, Pentium II and Pentium III systems (together with the corresponding Celeron & Xeon variants) are obsolete - like AMD K6 for which there never was an optimization. I honestly don't think it's worthwhile making the changes (involving an increase in executable size and complexity) to retain optimization for thoroughly obsolete architectures. The point is that systems up to PIII are more useful running trial factoring than LL testing, and trial factoring appears to be just the same in V24 as in V23. Regards Brian Beesley _______________________________________________ Prime mailing list [email protected] http://hogranch.com/mailman/listinfo/prime
