mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Software

Reply
 
Thread Tools
Old 2017-09-15, 15:16   #45
rudi_m
 
rudi_m's Avatar
 
Jul 2005

2·7·13 Posts
Default

Quote:
Originally Posted by Prime95 View Post
I'll look at restarting the 12-hour Jacobi timer after a benchmark is run.
BTW shouldn't these timers be somehow related to the "running-time"?

For me it looks like the benchmark runs daily even on machines which only run 1 hour mprime per day. This means the 3-minute benchmark costs 5% throughput.
rudi_m is offline   Reply With Quote
Old 2017-09-15, 16:31   #46
GP2
 
GP2's Avatar
 
Sep 2003

258710 Posts
Default

One very minor point... I'm running mprime, so I don't know what the Prime95 GUI does in 29.3, but do the Advanced/P-1 and Advanced/ECM options still offer the exponent 1061 by default, as version 28 did?

That exponent is fully factored, so maybe the default could be changed to 1277.
GP2 is offline   Reply With Quote
Old 2017-09-22, 17:45   #47
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

11·673 Posts
Default

With no serious errors reported, I've made 29.3 the official release.
Prime95 is offline   Reply With Quote
Old 2017-09-22, 20:50   #48
Mark Rose
 
Mark Rose's Avatar
 
"/X\(‘-‘)/X\"
Jan 2013

29×101 Posts
Default

So I put this on three machines, and it seems mprime is assigning the helper threads to the same core:

Code:
[Main thread Sep 22 16:30] Mersenne number primality test program version 29.3
[Main thread Sep 22 16:30] Optimizing for CPU architecture: Core i3/i5/i7, L2 cache size: 256 KB, L3 cache size: 8 MB
[Main thread Sep 22 16:30] Starting worker.
[Work thread Sep 22 16:30] Worker starting
[Work thread Sep 22 16:30] Setting affinity to run worker on logical CPUs 0 (zero-based)
[Work thread Sep 22 16:30] Setting affinity to run helper thread 3 on logical CPUs 0 (zero-based)
[Work thread Sep 22 16:30] Setting affinity to run helper thread 2 on logical CPUs 0 (zero-based)
[Work thread Sep 22 16:30] Setting affinity to run helper thread 1 on logical CPUs 0 (zero-based)
[Work thread Sep 22 16:30] Running Jacobi error check.  Passed.  Time: 13.966 sec.
[Work thread Sep 22 16:31] Resuming primality test of M44491127 using FMA3 FFT length 2400K, Pass1=320, Pass2=7680, clm=2, 4 threads
[Work thread Sep 22 16:31] Iteration: 2767230 / 44491127 [6.21%].
I simply extracted the new tar over the existing directory, to keep my local.txt and prime.txt. If I moved worktodo.txt and the various assignment files to a clean install, the clean install resumes processing with the cores assigned correctly.

It seems the following legacy stanza, from when I was using affinityscramble2, was causing the issue:

Code:
[Worker #1]
Affinity=0
I hope this helps anyone else who runs into this problem.
Mark Rose is offline   Reply With Quote
Old 2017-09-23, 01:14   #49
storm5510
Random Account
 
storm5510's Avatar
 
Aug 2009

19·101 Posts
Default

Quote:
Originally Posted by Mark Rose View Post
...It seems the following legacy stanza, from when I was using affinityscramble2, was causing the issue:

Code:
[Worker #1]
Affinity=0
I hope this helps anyone else who runs into this problem.
I had a related issue when I started using this i7. After some experimentation, I ended up with this:

Code:
CoresPerTest=4
WorkerThreads=1

[Worker #1]
Affinity=0,1,2,3
Going beyond four threads was a bit self-defeating for a single worker and slowed everything slightly. I did not have an affinity scramble at the time. I didn't see the need.

I keep 29.2 B3 and 29.3 in the same folder by using modified file names. I found that using separate folders caused a bit of confusion, in my case.
storm5510 is offline   Reply With Quote
Old 2017-09-23, 18:00   #50
mackerel
 
mackerel's Avatar
 
Feb 2016
UK

25·13 Posts
Default

I think I've seen this on older versions also, but thought I'd post now as I'm doing more testing with 29.3. What I did was run benchmark mode, setting HT off, and size 64 to 8192. Results come back at different FFT sizes on my Ryzen 1700 system, than the Intel 6700k and 7800X systems. I don't know if it might be number of core related (the Intels are 4 and 6 core, Ryzen 8 core), or something else to do with CPU type.

Following table colums middle left is Ryzen 1700, middle right is Intel (6700k or 7800X). The ones on either side are just copies of the sizes performed on that system but not on the other. Is this expected behaviour?

Each system had a fresh download of 29.3 with no configuration outside the benchmark. If it makes a difference, the Ryzen system defaulted to testing 1, 2, 8 workers, and I removed 2 leaving only 1 and 8.

Code:
	64	64	
	72	72	
	80	80	
	84	84	
	96	96	
	100	100	
	112	112	
	120	120	
	128	128	
	144	140	
	160	144	
		160	160
	168	168	
	192	192	
	200	200	
	224	224	
	240	240	
	256	256	
280	280		
	288	288	
	320	320	
	336	336	
	384	384	
	400	400	
	448	448	
	480	480	
	512	512	
	560	560	
		576	576
	640	640	
	672	672	
		720	720
	768	768	
	800	800	
		864	864
	896	896	
	960	960	
	1024	1024	
	1120	1120	
	1152	1152	
		1200	1200
	1280	1280	
	1344	1344	
	1440	1440	
	1536	1536	
	1600	1600	
	1680	1680	
		1728	1728
	1792	1792	
	1920	1920	
		2016	2016
	2048	2048	
2240	2240		
	2304	2304	
	2400	2400	
	2560	2560	
	2688	2688	
2800	2800		
	2880	2880	
	3072	3072	
	3200	3200	
	3360	3360	
		3456	3456
	3584	3584	
	3840	3840	
	4096	4096	
	4480	4480	
	4608	4608	
	4800	4800	
	5120	5120	
	5376	5376	
5600	5600		
	5760	5760	
	6144	6144	
	6400	6400	
	6720	6720	
		6912	6912
	7168	7168	
	7680	7680	
8000	8000		
	8064	8064	
	8192	8192
mackerel is offline   Reply With Quote
Old 2017-09-23, 19:27   #51
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

11×673 Posts
Default

Quote:
Originally Posted by mackerel View Post
Following table colums middle left is Ryzen 1700, middle right is Intel (6700k or 7800X). The ones on either side are just copies of the sizes performed on that system but not on the other. Is this expected behaviour?
This is normal.

Take 160K FFT. The testing I did indicated the 160K FFT is slower than the 168K FFT on Intel CPUs. On Ryzen though the 160K is faster than the 168K (using the benchmarks emailed to me). Thus, in a default install, the Intel user will not see a 160K FFT selected, the Ryzen user will.

If your particular CPU does not behave like the typical Intel or Ryzen, then the throughput benchmarks that are run every 20 hours or so will decide whether the 160K FFT is useful.
Prime95 is offline   Reply With Quote
Old 2017-09-30, 06:25   #52
Citrix
 
Citrix's Avatar
 
Jun 2003

157910 Posts
Default

Is there a way to write conditional work in worktodo.txt?

For example run 100 ECM curves with bounds B1 and B2 and if no factors are found, run a PRP test.

Thanks.
Citrix is offline   Reply With Quote
Old 2017-09-30, 14:23   #53
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

11×673 Posts
Default

Quote:
Originally Posted by Citrix View Post
Is there a way to write conditional work in worktodo.txt?

For example run 100 ECM curves with bounds B1 and B2 and if no factors are found, run a PRP test.
Sorry, there is no way to do that.
Prime95 is offline   Reply With Quote
Old 2017-10-01, 05:03   #54
bayanne
 
bayanne's Avatar
 
"Tony Gott"
Aug 2002
Yell, Shetland, UK

3·107 Posts
Default

Are PRP tests likely to appear as an option in the Workers Windows box ...
bayanne is offline   Reply With Quote
Old 2017-10-01, 10:39   #55
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

11·673 Posts
Default

Quote:
Originally Posted by bayanne View Post
Are PRP tests likely to appear as an option in the Workers Windows box ...
Yes, in 29.4.
Prime95 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Prime95 version 27.3 Prime95 Software 148 2012-03-18 19:24
Prime95 version 26.3 Prime95 Software 76 2010-12-11 00:11
Prime95 version 25.5 Prime95 PrimeNet 369 2008-02-26 05:21
Prime95 version 25.4 Prime95 PrimeNet 143 2007-09-24 21:01
When the next prime95 version ? pacionet Software 74 2006-12-07 20:30

All times are UTC. The time now is 22:30.

Mon Apr 12 22:30:06 UTC 2021 up 4 days, 17:10, 1 user, load averages: 3.45, 3.21, 3.17

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.