ECM users - version 30.9-30.12 (see post#235)

Prime95 · 2022-06-29, 21:09

I've been very slow at writing new code lately -- summertime I guess.

Is there interest in a pre-beta version of 30.9 that uses a GMP-ECM-like stage 2? It's been working and pretty stagnant for a month, but as with the 30.8 pre-beta there is a long way to go before it's ready for prime time.

Interestingly, in my limited testing it seems that, unlike P-1, B2 > ~1000 * B1 does not make sense. You are better off running more curves or increasing B1. In some situations, prime95 elects not to use all the available memory -- this will mean the default MaxHighMemWorkers setting might need to change.

BTW, the first and only success thusfar is here: 909107

storm5510 · 2022-06-29, 22:20

ECM has been my fallback for years. It's something I can start and leave running for day and not check on it.

I have used GMP-ECM a lot. My only issue with it is that it does not multi-thread. Other say it doesn't need to, run multiple instances instead. This is more than I want to keep up with.

Denial140 · 2022-06-29, 22:22

I would be interested in using it e.g. to speed up the ECM work on 2.22M-2.23M to take it to below 200 exponents unfactored.

kriesel · 2022-06-30, 00:34

Quote:

Originally Posted by Prime95

Is there interest in a pre-beta version of 30.9 that uses a GMP-ECM-like stage 2? It's been working and pretty stagnant for a month, but as with the 30.8 pre-beta there is a long way to go before it's ready for prime time.

First reactions:
You know it. If you build it, users will come. Whatever the next new or improved feature is. Happens every time.
Really, we've become quite accustomed over the decades, to you finding more ways to wring even more out of available hardware. Thanks again.
Not needing the prime95/gmp-ecm shuffle or gmp-ecm alone would be convenient and welcome.
Personally I'd like to see handling proof files on the server for p > 596M get automated to conserve your time for such other creativity. (Maybe a single karatsuba implemented length just there. 64M words via 3 x 32M fft, leveraged from 596M up to 1G+ exponent for proof processing on server.) But as always, your time your call.
Handling the "I want to change the proof residues location midflight without manually stopping the program and moving the residues file beforehand" case would be nice and surely save some otherwise lost proofs. Save us mortal users from our own lack of foresight.

lycorn · 2022-06-30, 07:08

Quote:

Originally Posted by Prime95

Is there interest in a pre-beta version of 30.9 that uses a GMP-ECM-like stage 2?

Count me in.

Zhangrc · 2022-06-30, 11:45

I've seen SUMOUT errors occur when running ECM on very small exponents and cofactors. (M1277 for example)
I'm not trying to reproduce the error (I have more PRPs to do), and it seems that these errors are not recorded in the log.

axn · 2022-06-30, 12:06

Quote:

Originally Posted by Prime95

Interestingly, in my limited testing it seems that, unlike P-1, B2 > ~1000 * B1 does not make sense. You are better off running more curves or increasing B1. In some situations, prime95 elects not to use all the available memory -- this will mean the default MaxHighMemWorkers setting might need to change.

Have you thrown a 100GB at the problem?

If I can get access to the pre-beta (with instructions, if any), I can do some high memory testing and report back on the behavior

masser · 2022-06-30, 15:00

Quote:

Originally Posted by Prime95

Is there interest in a pre-beta version of 30.9 that uses a GMP-ECM-like stage 2? It's been working and pretty stagnant for a month, but as with the 30.8 pre-beta there is a long way to go before it's ready for prime time.

I will test it, too. Will this version also include P+1 with improved stage 2?

Prime95 · 2022-07-01, 02:42

Quote:

Originally Posted by masser

I will test it, too. Will this version also include P+1 with improved stage 2?

P+1 is not improved.

I'll see if I can make a build sometime soon.

Prime95 · 2022-07-03, 00:25

30.9 build 1

If you aren't doing ECM, do not download! PRE-BETA software!

Save files during ECM stage 2 are broken (probably a crash)
Stage 2 time estimates and optimal B2 bounds could be off
Accurate estimates of stage 2 memory consumed may be off
Further stage 2 multithreading improvements are needed.
Stage 2 is pretty verbose, there's lots of code cleanup in my future.

Windows 64-bit: https://mersenne.org/ftp_root/gimps/p95v309b1.win64.zip
Linux 64-bit: https://mersenne.org/ftp_root/gimps/...linux64.tar.gz

Glenn · 2022-07-04, 07:31

I’ve downloaded 30.9b1 but am continuing to use 30.8b15,
which is now “official” on the main Mersenne.org page.

I had a single round off error over 0.4 on a recent test for
M113782777, but successfully completed that test anyway.
Since then my tests (two more so far) include a “round off:”
column in my results for each group of 10,000 iterations.

Please let me know when it’s safe to use 30.9.

2022-06-29, 21:09	#1
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 5·13·127 Posts	ECM users - version 30.9-30.12 (see post#235) I've been very slow at writing new code lately -- summertime I guess. Is there interest in a pre-beta version of 30.9 that uses a GMP-ECM-like stage 2? It's been working and pretty stagnant for a month, but as with the 30.8 pre-beta there is a long way to go before it's ready for prime time. Interestingly, in my limited testing it seems that, unlike P-1, B2 > ~1000 * B1 does not make sense. You are better off running more curves or increasing B1. In some situations, prime95 elects not to use all the available memory -- this will mean the default MaxHighMemWorkers setting might need to change. BTW, the first and only success thusfar is here: 909107

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
That's a Lot of Users!!!	jinydu	Lounge	9	2006-11-10 00:14
Beta version 24.6 - Athlon users wanted	Prime95	Software	139	2005-03-30 12:13
For Old Users	Citrix	Prime Sierpinski Project	15	2004-08-22 16:43
Opportunity! Retaining new users post-M40	GP2	Lounge	55	2003-11-21 21:08
AMD USERS	ET_	Lounge	3	2003-10-11 16:52

2022-06-29, 22:20	#2
storm5510 Random Account Aug 2009 Not U. + S.A. 31×89 Posts	ECM has been my fallback for years. It's something I can start and leave running for day and not check on it. I have used GMP-ECM a lot. My only issue with it is that it does not multi-thread. Other say it doesn't need to, run multiple instances instead. This is more than I want to keep up with.

2022-06-29, 22:22	#3
Denial140 Dec 2021 2⁴×5 Posts	I would be interested in using it e.g. to speed up the ECM work on 2.22M-2.23M to take it to below 200 exponents unfactored.

2022-06-30, 11:45	#6
Zhangrc "University student" May 2021 Beijing, China 269 Posts	I've seen SUMOUT errors occur when running ECM on very small exponents and cofactors. (M1277 for example) I'm not trying to reproduce the error (I have more PRPs to do), and it seems that these errors are not recorded in the log.

2022-07-03, 00:25	#10
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 5×13×127 Posts	30.9 build 1 If you aren't doing ECM, do not download! PRE-BETA software! Save files during ECM stage 2 are broken (probably a crash) Stage 2 time estimates and optimal B2 bounds could be off Accurate estimates of stage 2 memory consumed may be off Further stage 2 multithreading improvements are needed. Stage 2 is pretty verbose, there's lots of code cleanup in my future. Windows 64-bit: https://mersenne.org/ftp_root/gimps/p95v309b1.win64.zip Linux 64-bit: https://mersenne.org/ftp_root/gimps/...linux64.tar.gz

2022-07-04, 07:31	#11
Glenn "Glenn Leider" Apr 2021 Carlsbad, CA 3³ Posts	I’ve downloaded 30.9b1 but am continuing to use 30.8b15, which is now “official” on the main Mersenne.org page. I had a single round off error over 0.4 on a recent test for M113782777, but successfully completed that test anyway. Since then my tests (two more so far) include a “round off:” column in my results for each group of 10,000 iterations. Please let me know when it’s safe to use 30.9.