V24.12 release candidate 3

[Prime95]

Originally Posted by ET_

you may as well check version 23.2: in many cases it is faster than verion 24 at factoring.

I think the v23.2 and v24 trial factoring code is the same for a PII and PIII.

[Prime95]

Originally Posted by smh

No 3rd one for me either.
Pentium M 1,4GHz

I'll have a fix for this too. SSE2 trial factoring has a bug for exponents above 580 million.

[Prime95]

Originally Posted by JHagerson

I upgraded one machine that was in the middle of a 2^68 factoring assignment from 24.11 to 24.12 RC3. When I restarted Prime95, that workstation immediately communicated with the server.

John, can you send the prime.log file to me at woltman@alum.mit.edu? Thanks.

[bayanne]

Originally Posted by Prime95

John, can you send the prime.log file to me at woltman@alum.mit.edu? Thanks.

I think that they have amended the factoring level from 2^68 to 2^67 ........

I'll send you a copy of the log file.

[ET_]

Originally Posted by Prime95

I think the v23.2 and v24 trial factoring code is the same for a PII and PIII.

George is right, I was referring to benchmarks again. My apologies

Luigi

[Prime95]

Originally Posted by kjaget

Not sure if this counts as a bug, but 24.12 is slower than 24.6 for exponents >= 1280K FFT size on my Celeron D machine (256K L2).

Can you follow the instructions in this thread, http://www.mersenneforum.org/showthread.php?t=4202, and post your Celeron D benchmarks. I thought the Celeron D would behave like my Willamette, but apparently it does not.

[kjaget]

Originally Posted by Prime95

Can you follow the instructions in this thread, http://www.mersenneforum.org/showthread.php?t=4202, and post your Celeron D benchmarks. I thought the Celeron D would behave like my Willamette, but apparently it does not.

I won't be able to get to the machine for a few days, but when I do I'll try the benchmark code from that thread and let you know what I find.

[Prime95]

Originally Posted by kjaget

I won't be able to get to the machine for a few days, but when I do I'll try the benchmark code from that thread and let you know what I find.

It wouldn't hurt for other Celeron D owners to run the special benchmarks. This will let me know if kjaget's computer is a special case or representative of all Celeron Ds.

[kjaget]

Originally Posted by Prime95

It wouldn't hurt for other Celeron D owners to run the special benchmarks. This will let me know if kjaget's computer is a special case or representative of all Celeron Ds.

Not sure if it's obvious from the results that I posted, but it's a 325J E0 stepping, based on the Prescott P4s. You know better than me, but I think that means it has a longer pipeline and worse cache latencies than a Northwood or Wilamette.

I should be able to get the timings on this machine towards the end of the week. Hopefully that will clear up some of the mystery.

Kevin

I just switched from RC1 to RC3 to resolve a bug, but I still have the problem, so I report it here. When starting P1-factoring stage 2, v.24.12 crashes when there is unsufficient memory. Changing DayMemory = 48 to 96 solved the problem, but the bug is still there.

[Prime95]

Originally Posted by filidoor

When starting P1-factoring stage 2, v.24.12 crashes when there is unsufficient memory. Changing DayMemory = 48 to 96 solved the problem, but the bug is still there.

It will be fixed in RC4. Thanks!