mersenneforum.org  

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

Reply
 
Thread Tools
Old 2019-03-04, 15:39   #78
Cruelty
 
Cruelty's Avatar
 
May 2005

65216 Posts
Default

Was there any change in handling Affinity between 29.4 and 29.6?
Code:
[Worker #1]
Affinity=4,6

[Worker #2]
Affinity=0,2
Those settings does not seem to work in 29.6 as they were in 29.4, and CPU cores are chosen randomly and change during calculation?
Cruelty is offline   Reply With Quote
Old 2019-03-04, 20:16   #79
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

3×2,281 Posts
Default

Quote:
Originally Posted by Cruelty View Post
"Small FFTs" 12h stress test passed - I'd suggest lowering Max FFT size from 586K to maybe 128K,
Is this for 20 torture threads on 13.75MB L3 cache?
Prime95 is offline   Reply With Quote
Old 2019-03-04, 20:35   #80
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

3×2,281 Posts
Default

Quote:
Originally Posted by Cruelty View Post
Was there any change in handling Affinity between 29.4 and 29.6?
Not to my knowledge.


Quote:
Code:
[Worker #1]
Affinity=4,6

[Worker #2]
Affinity=0,2
Those settings does not seem to work in 29.6 as they were in 29.4, and CPU cores are chosen randomly and change during calculation?
29.6 is linked with a newer version of hwloc. Is it reporting an accurate description of your architecture in results.bench.txt?
Prime95 is offline   Reply With Quote
Old 2019-03-04, 23:34   #81
Cruelty
 
Cruelty's Avatar
 
May 2005

2·809 Posts
Default

Quote:
Originally Posted by Prime95 View Post
Is this for 20 torture threads on 13.75MB L3 cache?
Yes
Quote:
29.6 is linked with a newer version of hwloc. Is it reporting an accurate description of your architecture in results.bench.txt?
I don't see such file in prime95 directory CPU information in prime95 seems accurate though.
Cruelty is offline   Reply With Quote
Old 2019-03-05, 03:16   #82
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

3·2,281 Posts
Default

Quote:
Originally Posted by Cruelty View Post
Yes
I don't see such file in prime95 directory CPU information in prime95 seems accurate though.
You get results.bench.txt whenever you do any benchmark. Just start one and abort it.
Prime95 is offline   Reply With Quote
Old 2019-03-05, 03:22   #83
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

3·2,281 Posts
Default

Quote:
Originally Posted by Cruelty View Post
Is this for 20 torture threads on 13.75MB L3 cache?
Something is off. A small torture test trying to run in L3 cache should assume 13.75MB / 20 threads = 704KB / test. Each FFT word is an 8-byte double. Thus max FFT size should be under 88K.

Last fiddled with by Prime95 on 2019-03-05 at 03:25
Prime95 is offline   Reply With Quote
Old 2019-03-05, 07:11   #84
Cruelty
 
Cruelty's Avatar
 
May 2005

65216 Posts
Default

Quote:
Originally Posted by Prime95 View Post
You get results.bench.txt whenever you do any benchmark. Just start one and abort it.
Here you go.
Attached Files
File Type: txt results.bench.txt (3.1 KB, 36 views)
Cruelty is offline   Reply With Quote
Old 2019-03-05, 15:59   #85
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

11010101110112 Posts
Default

Quote:
Originally Posted by Cruelty View Post
Was there any change in handling Affinity between 29.4 and 29.6?
Code:
[Worker #1]
Affinity=4,6

[Worker #2]
Affinity=0,2
Those settings does not seem to work in 29.6 as they were in 29.4, and CPU cores are chosen randomly and change during calculation?
So 29.6 should bind worker #1 to PU#4 and PU#6 (which are logical CPUs on two different cores) as reported in results.bench.txt. Are you getting messages on screen saying what logical CPU each worker is being bound to?
Prime95 is offline   Reply With Quote
Old 2019-03-05, 18:43   #86
Cruelty
 
Cruelty's Avatar
 
May 2005

2×809 Posts
Default

Quote:
Originally Posted by Prime95 View Post
So 29.6 should bind worker #1 to PU#4 and PU#6 (which are logical CPUs on two different cores) as reported in results.bench.txt. Are you getting messages on screen saying what logical CPU each worker is being bound to?
Yes, but it's not what actually happens when I watch the graphs in task manager.
Attached Thumbnails
Click image for larger version

Name:	cpu-utilization1.png
Views:	48
Size:	195.5 KB
ID:	19996   Click image for larger version

Name:	cpu-utilization2.png
Views:	42
Size:	191.8 KB
ID:	19997  
Cruelty is offline   Reply With Quote
Old 2019-03-06, 00:22   #87
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

684310 Posts
Default

Quote:
Originally Posted by Cruelty View Post
Yes, but it's not what actually happens when I watch the graphs in task manager.
Hmmm. I looked at the code and see no problems (though that is far from conclusive!).

Are there any Windows system tools that tell you if a process' threads are bound to specific logical CPUs?

I'll try replicating on my non-hyperthreaded quad-core.
Prime95 is offline   Reply With Quote
Old 2019-03-06, 03:03   #88
Madpoo
Serpentine Vermin Jar
 
Madpoo's Avatar
 
Jul 2014

3·1,087 Posts
Default

Quote:
Originally Posted by harlee View Post
Madpoo,

It appears that JSON isn't displaying PRP-C results the same way as non-JSON - known factors are not being shown. Please see M5342101 and M7039433 and PrimeNet Most Recent Results page (sort by Type). Even the entry on my Account Results Details page is different:
Fixed. The latest builds have the known-factors in an array in the JSON data... not sure how the previous P95 builds would have included the known-factors if there were more than one, maybe just as a comma-delimited single value.
Madpoo is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Prime95 version 29.2 Prime95 Software 71 2017-09-16 16:55
Prime95 version 29.1 Prime95 Software 95 2017-08-22 22:46
Prime95 version 26.5 Prime95 Software 175 2011-04-04 22:35
Prime95 version 25.9 Prime95 Software 143 2010-01-05 22:53
Prime95 version 25.8 Prime95 Software 159 2009-09-21 16:30

All times are UTC. The time now is 01:52.

Sun Jun 7 01:52:21 UTC 2020 up 73 days, 23:25, 0 users, load averages: 1.43, 1.40, 1.35

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, 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.