mersenneforum.org  

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

Reply
 
Thread Tools
Old 2007-09-22, 23:35   #1
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

3×7×167 Posts
Default Running DC on quad-core problem.

Can anyone think of a reason that a Linux box(Ubuntu Feisty Fawn) would quit running nice-19 projects on a random core, even though projects with a higher priority will run, but the higher priority projects don't even come close to maxing out the misbehaving core? It doesn't seem to matter what core it is, either. So far, I've seen it happen on cores 0 and 3.

It's an E6600 cpu, I forget which motherboard, with 2Gig of RAM. My dad thinks he has a vague idea what the problem is, based on what I've told him. I think he said something about the OS settings. I'm sorry that's so vague.

Basically, I've got three cores maxed out, and a fourth one that still operates, but refuses to run my DC project. Is it possible my Ubuntu program is reserving a core specifically for the OS?
jasong is offline   Reply With Quote
Old 2007-09-23, 00:06   #2
Xyzzy
 
Xyzzy's Avatar
 
Aug 2002

7×1,193 Posts
Default

If you run out of memory the kernel will kill processes.

http://linux-mm.org/OOM_Killer

Note the part about niceness.
Xyzzy is offline   Reply With Quote
Old 2007-09-23, 01:05   #3
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

3×7×167 Posts
Default

I appreciate the link Xyzzy, but I'm not sure if it applies to me. I have 4 instances of DC projects running, 2 have 100% utilization, and 2 have 50%. Altogether, they use about 5% of the memory. If I'm running out of memory, the culprit is probably Firefox. But I use Firefox much more on my 1Gig Linux box, and I don't think I've ever had memory problems with it. It's the same OS by the way, Ubuntu Feisty Fawn 7.04.
jasong is offline   Reply With Quote
Old 2007-09-23, 06:29   #4
Xyzzy
 
Xyzzy's Avatar
 
Aug 2002

7×1,193 Posts
Default

There are two files that may contain clues.

Code:
/var/log/syslog
/var/log/messages
If you can duplicate the behavior try watching these files to see what the kernel says when it terminates the process. Example:

Code:
tail -f /var/log/messages
Your first goal is finding the conditions to duplicate the behavior.

Both files track all sorts of stuff and you can view the archive of events. The archives are stored like messages.1.gz and can be read with zcat. Example:

Code:
zcat syslog.1.gz | less
What program is being killed? Perhaps it is just a software error. What does the terminal say when it is killed?

Nothing happens without a "paper trail" in Linux.

Note: You have to be root to view those files. Since you are using Ubuntu that means using sudo. Just be careful what you type in.
Xyzzy is offline   Reply With Quote
Old 2007-09-24, 02:40   #5
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

350710 Posts
Default

I really appreciate your help, Xyzzy, and I have looked at those files, though nothing jumps out at me.

I really don't think it comes from an out-of-memory problem, since I rebooted about 45 minutes ago, started my programs immediately, and still had a core that refused to run my DC project.

I'm wondering if the extra core would accept my DC projects if they were a higher priority. It could be I accidentally enabled a feature that reserved one core for high-priority tasks. I might try changing the niceness level to on one of the instances to see what happens.
jasong is offline   Reply With Quote
Old 2007-10-10, 00:48   #6
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

3·7·167 Posts
Default

I think I'm living in Bizzaro world. I wanted to get BOINC running under Wine, since a project called PrimeGrid only has Windows apps available. Well, I got it working, and the fourth core is running now.

I don't know if it has something to do with Wine, or if it was the restart. I'm going to watch it and see what happens.
jasong is offline   Reply With Quote
Old 2007-10-25, 22:30   #7
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

3×7×167 Posts
Default

I've come up with a possible theory(more generally used definition, rather than the mathematical or scientific definition) for why my computer seems to be randomly going down to only 3 cores.

My guess is that Ubuntu has some sort of safety feature for when the load average(I don't know of an equivalent in Windows, Google "linux load average" if you're curious :) ) gets too high. I'm thinking the load average could be indirectly related to the temperature of a cpu, so maybe if any of the load averages go more than 10-15% over the number of available cores, things get shut down.

It's just a theory, partially based on the two facts that (1) restarting only sometimes brings the core back, and (2) sometimes it comes back on it's own.
jasong is offline   Reply With Quote
Old 2007-10-25, 22:38   #8
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

3·7·167 Posts
Default

Quote:
Originally Posted by Xyzzy View Post
You have to be root to view those files. Since you are using Ubuntu that means using sudo. Just be careful what you type in.
I managed to open the files by right-clicking and clicking open. When I tried to open them through a bash prompt, the OS said they were write-protected. So I'm guessing if I had tried to edit them the OS would have complained.
jasong is offline   Reply With Quote
Old 2007-10-29, 00:58   #9
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

350710 Posts
Default

(sorry for the three posts in a row :( )

Yeah, I'm almost certain it's the load average. I remember reading something about a points system related to the load average. The two DC processes that are restricted to only using 50% power(100% being equal to one core) each have 4 and 1 point(s) when I grep them.

I'm going to Google "linux load average" to verify whether or not I'm correct.
jasong is offline   Reply With Quote
Old 2007-10-29, 01:49   #10
Tumo
 
Jan 2005

19 Posts
Default offer of help with multi-core cpus.

Quote:
Originally Posted by jasong View Post
(sorry for the three posts in a row :( )

Yeah, I'm almost certain it's the load average. I remember reading something about a points system related to the load average. The two DC processes that are restricted to only using 50% power(100% being equal to one core) each have 4 and 1 point(s) when I grep them.

I'm going to Google "linux load average" to verify whether or not I'm correct.
Jasong,
First... HI! Long time no chat! (and hi to all).

Regarding your quad core, I am a bit out of date with your hardware config. You do have some power-management / heat preservation features built into the cpu. If you, or anyone, would like, I am willing to share what I have/know and that I am successfully AMD dual and quad-core machines (Opties) on Fedora and win/2k. I successfully ran a prototype AMD cpu last summer with no problems.
The key is your power supply chain (from the wall through the PS to the mobo).
I offer my services to anyone in need... (Intel or AMD)
I will also research Ubuntu and make sure the appropriate drivers / modules are installed. I had a problem with Gentoo that sounds like what you are having now, and that was with simple, 3.06G dual-cpu, single-core Intels.
Again, anyone is welcome to PM me as needed (if preferred). I offer my services to all, in the thread or in private.

Chuck
Tumo is offline   Reply With Quote
Old 2007-10-29, 02:02   #11
Xyzzy
 
Xyzzy's Avatar
 
Aug 2002

7×1,193 Posts
Default

It isn't load average. If it was, how could you explain this box? (It has only one core.)

Code:
top - 22:00:18 up 12 days,  8:20,  5 users,  load average: 4.95, 4.28, 4.13
Tasks: 110 total,   5 running, 105 sleeping,   0 stopped,   0 zombie
Cpu0  :  2.7%us,  0.8%sy, 94.5%ni,  0.5%id,  0.0%wa,  0.6%hi,  0.9%si,  0.0%st
Mem:   1034872k total,   831636k used,   203236k free,    34924k buffers
Swap:  1034872k total,       84k used,  1034788k free,   389024k cached
http://en.wikipedia.org/wiki/Load_average
Xyzzy is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Dual Core to Quad Core Upgrade Rodrigo Hardware 6 2010-11-29 18:48
exclude single core from quad core cpu for gimps jippie Information & Answers 7 2009-12-14 22:04
Quad Core and P95 sgrupp Hardware 54 2008-01-25 22:01
Quad Core R.D. Silverman Hardware 76 2007-11-19 21:57
Optimising work for Intel Core 2 Duo or Quad Core S485122 Software 0 2007-05-13 09:15

All times are UTC. The time now is 03:05.


Thu Dec 2 03:05:10 UTC 2021 up 131 days, 21:34, 0 users, load averages: 1.20, 1.24, 1.26

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.