Wasting resources. (LHC Test4Theory)

Message boards : Projects : Wasting resources. (LHC Test4Theory)
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile adrianxw
Avatar

Send message
Joined: 2 Oct 05
Posts: 400
Denmark
Message 48998 - Posted: 5 May 2013, 21:26:22 UTC

I noticed that when an LHC2 multithreaded application was running, I was seeing the Windows inactivity process "running". Upon querying this, am informed that the mutithreaded app rarely utilizes the paralleled CPU's to the full extent, that this is normal, indeed, I was shown a list of processes that were sometimes wasting substantial resources.

I am deeply unhappy with this.
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.
ID: 48998 · Report as offensive
Profile Gary Charpentier
Avatar

Send message
Joined: 23 Feb 08
Posts: 2463
United States
Message 49013 - Posted: 7 May 2013, 2:40:41 UTC - in response to Message 48998.  

I noticed that when an LHC2 multithreaded application was running, I was seeing the Windows inactivity process "running". Upon querying this, am informed that the mutithreaded app rarely utilizes the paralleled CPU's to the full extent, that this is normal, indeed, I was shown a list of processes that were sometimes wasting substantial resources.

I am deeply unhappy with this.

Best ask on the LHC2 boards.

ID: 49013 · Report as offensive
Profile adrianxw
Avatar

Send message
Joined: 2 Oct 05
Posts: 400
Denmark
Message 49016 - Posted: 7 May 2013, 15:50:16 UTC
Last modified: 7 May 2013, 15:55:33 UTC

I did, the only response I had was one confirming what I had seen and giving a overview rank order of the efficiency levels of the various tasks.
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.
ID: 49016 · Report as offensive
SekeRob2

Send message
Joined: 6 Jul 10
Posts: 585
Italy
Message 49017 - Posted: 7 May 2013, 16:17:15 UTC - in response to Message 49016.  

Were you expecting some BOINC dev response from venting your discontent here? The projects are responsible for the efficiency... BOINC itself is really just a scheduler assigning work to available resources. How well the project apps use that resource is what BOINC has little influence on. E.g. WCG had the Help Conquer Cancer science app which had specific CPU and GPU phases. Default the app asked for 1 CPU and 1 GPU for the full duration, which is what BOINC allocated. Practically, the CPU was idling most of the full run time while the GPU was processing the bulk of the task, and the GPU was idling several minutes during startup and close of the task whilst the app did the post-processing and packaging for return of the result. Only manual tweaking allowed for greater utilization of the resources. Berkeley to blame? Emphatically, No!
Coelum Non Animum Mutant, Qui Trans Mare Currunt
ID: 49017 · Report as offensive
Profile adrianxw
Avatar

Send message
Joined: 2 Oct 05
Posts: 400
Denmark
Message 49019 - Posted: 7 May 2013, 17:44:52 UTC

Where did I say BOINC was to blame? I "vented" to spread a simple warning to others. I don't have a GPU so your further example has not been seen here.
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.
ID: 49019 · Report as offensive
SekeRob2

Send message
Joined: 6 Jul 10
Posts: 585
Italy
Message 49025 - Posted: 7 May 2013, 19:29:38 UTC - in response to Message 49019.  

To Moderator:

Propose

1) This thread be moved to Projects
2) Title be amended to include this is LHC specific
3) A Link to the discussion on LHC Forums is added, so readers can form their opinion on basis of the project owners and participants feedback.
Coelum Non Animum Mutant, Qui Trans Mare Currunt
ID: 49025 · Report as offensive
Profile adrianxw
Avatar

Send message
Joined: 2 Oct 05
Posts: 400
Denmark
Message 49055 - Posted: 8 May 2013, 19:15:30 UTC

1. Sure feel free, I was annoyed about the waste caused by LHC2. The thread here leads me to suspect other projects may be doing the same though.

2. See 1. above.

3. If you go to the LHC2 site forum, you will see that the posts are all old. However, if I go to my account, then to messages, I can find it. I suspect that is not relevent however, just curious. The thread is here.
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.
ID: 49055 · Report as offensive
Profile Gary Charpentier
Avatar

Send message
Joined: 23 Feb 08
Posts: 2463
United States
Message 49058 - Posted: 9 May 2013, 0:46:18 UTC - in response to Message 49055.  

1. Sure feel free, I was annoyed about the waste caused by LHC2.

What waste? BOINC only takes waste cycles to begin with.
The thread here leads me to suspect other projects may be doing the same though.

100% of projects that multi-core do this to some degree. You can't write a program that uses more than one thread to exactly split it work. There will always be contention for a critical resource. Nature of the beast. Just know that all modern processors really go to sleep and power down when idle.

ID: 49058 · Report as offensive
Profile adrianxw
Avatar

Send message
Joined: 2 Oct 05
Posts: 400
Denmark
Message 49064 - Posted: 9 May 2013, 13:47:42 UTC
Last modified: 9 May 2013, 13:57:21 UTC

>>> What waste? BOINC only takes waste cycles to begin with.

I leave this machine running 24/7 so it can run BOINC, I normally have another quad core here just for BOINC, (currently needs a new disk), other machines here were not connected to LHC2. I am paying for those cores to run BOINC, LHC2, is taking the resource, and not utilizing it "efficiently" (for want of a better word). Indeed, if you read the response I had at LHC, one of their apps hardly uses the second core at all.

"Waste cycles" is a convenient but imprecise term.

I agree that contentions exist and that a multi threaded app will never 100% use both cores - I have been programming multi threaded applications for 10+ years. The response I had in their thread says that the use % of the second core varies, with at least one of their applications taking the resource and barely using it at all.

>>> Just know that all modern processors really go to sleep and power down when idle.

When there are other tasks waiting to run, other BOINC jobs in this case, they should not be powering down, they should context switch - this is not the case with the LHC2 job I was watching.
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.
ID: 49064 · Report as offensive
Profile adrianxw
Avatar

Send message
Joined: 2 Oct 05
Posts: 400
Denmark
Message 49093 - Posted: 13 May 2013, 16:25:33 UTC

Another "quirk" is that the LHC2 jobs run at Windows "Normal" priority, not "Low" like the other BOINC jobs.
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.
ID: 49093 · Report as offensive

Message boards : Projects : Wasting resources. (LHC Test4Theory)

Copyright © 2024 University of California.
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.