limit CPU Cores

Questions and Answers : Preferences : limit CPU Cores
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Sutaru Tsureku
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 911625 - Posted: 26 Jun 2009, 12:26:49 UTC


A Quad-CPU and 4 GPUs.

If I would like to let run only 2 threads on the CPU and all 4 GPUs with full load, I should insert 6 or 2 in the computing preferences?

The GPUs are "CPUs"?


On multiprocessors, use at most X processors


Thanks! :-)

ID: 911625 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 911631 - Posted: 26 Jun 2009, 12:49:05 UTC - in response to Message 911625.  

You set your preferences to use 50% of your CPUs and add the following to your cc_config.xml file:

<cc_config>
<options>
<use_all_gpus>1</use_all_gpus> 
</options>
<cc_config>


Then you tell BOINC to use the changes in the cc_config.xml file, by either exiting BOINC and restarting it, or by BOINC Manager->Advanced view->Advanced->Read config file.

If you set the Use 50% of CPUs on the web-preferences, you have to update Seti:
BOINC Manager->Advanced view->Projects tab->Select Seti->Click Update.
If you use the local advanced preferences, these are used immediately upon saving.
ID: 911631 · Report as offensive
Profile Sutaru Tsureku
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 912583 - Posted: 29 Jun 2009, 11:31:42 UTC
Last modified: 29 Jun 2009, 12:11:56 UTC


Thanks! :-)


Hmm.. other question.

I would like to calculate only on the GPUs.

I have an app_info.xml only with CUDA entry.

But BOINC ask every second work request [EDIT: also] for new CPU work.

I unchecked CPU usage in the SETI@home preferences.

The calculation preferences are:

On multiprocessors, use at most --- % of the processors [saved 0]

Again CPU work request.

On multiprocessors, use at most --- processors [saved 0]

Again CPU work request.

O.K., I get never CPU WUs.. but BOINC ask. He don't need to..

It's a BUG?


I have set:
Use at most 100 percent of CPU time
But this is not of meaningful, or?



Thanks! :-)



EDIT:
Changed the title of this thread from:
GPU as "CPU" in preferences?
to:
limit CPU Cores

ID: 912583 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 912588 - Posted: 29 Jun 2009, 12:12:48 UTC - in response to Message 912583.  

I unchecked CPU usage in the SETI@home preferences.

And what's your setting for "If no work for selected applications is available, accept work from other applications?"

Will setting the amount of CPUs to use to 0% work with CUDA? You still need part of one CPU to supply the work for the GPU.

O.K., I get never CPU WUs.. but BOINC ask. He don't need to..

BOINC 6.6 has a separate CPU and GPU scheduler, so it may always try to contact home base. But is it asking for work? Can you post an example of such a request from your Messages?
ID: 912588 · Report as offensive
Profile Sutaru Tsureku
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 912613 - Posted: 29 Jun 2009, 14:29:17 UTC
Last modified: 29 Jun 2009, 14:32:39 UTC


It's look like this:

Run only the selected applications
SETI@home Enhanced: yes
Astropulse: no
Astropulse v5: no

If no work for selected applications is available, accept work from other applications? no
Use Graphics Processing Unit (GPU) if available yes
Use Central Processing Unit (CPU) no


Yes, CUDA is crunching well.
I made your hint with the <use_all_gpus>.

But I guess it's because of BOINC is confused and/or can't set 0 CPUs.
Because in the time I had set 50 %, in the messages was: CPUs 2 .
But now where were normally this 2, is now nothing. Like with 100 %.
So BOINC use again 4 CPUs - I guess.


To post the messages are not so easy..
I hope it's well if I write my memories.. ;-)

If work request for CPU and GPU, both have the same sec. (time)
Every 2nd work request CPU and GPU. Then again only GPU.
But of course I get only CUDA WUs because of app_info.xml .
(I have set <sched_op_debug> in cc_config.xml)


No <avg_ and <max_ncpus_ entry in app_info.xml .
Only CUDA entry in app_info.xml .
In BOINC '1 CPU, 1 CUDA' per CUDA WU.


Now you know everything.. ;-)
I guess - current BOINC can't do 0 CPUs.

Or where is maybe the problem?

ID: 912613 · Report as offensive
Profile Sutaru Tsureku
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 912618 - Posted: 29 Jun 2009, 14:40:16 UTC


BTW.
Maybe interesting also..

In the project overview of BOINC, I can click on 'properties' there I can see a counter for CPU and GPU when the next work request will be.
Also I see how big is the time between work requests.

The CPU stay at 24 hours.
And count backwards.

But like I said, in the messages every 2nd work request BOINC ask for CPU work also.

ID: 912618 · Report as offensive
Profile Gundolf Jahn

Send message
Joined: 19 Sep 00
Posts: 3184
Credit: 446,358
RAC: 0
Germany
Message 912649 - Posted: 29 Jun 2009, 16:02:52 UTC - in response to Message 912618.  

But like I said, in the messages every 2nd work request BOINC ask for CPU work also.

I think that's the same problem (just with CPU and GPU transposed) commented by Richard Haselgrove in Message 25581 on the BOINC dev boards:

That is the point. CP does not do CUDA, therefore I would like to be able to tell BOINC to stop asking CP (and other projects that don't do CUDA) for CUDA work units.

Does it really matter? Is it a measurable overhead for either your processing or your comms bandwidth?

On the other hand, many years ago we were told that contacting a project scheduler (and hence accessing the BOINC database) was a costly operation in server resource terms. Now we seem to be happy throwing these redundant RPCs all over the place (even if only twice/host/project/day). I still worry that this may be a contributing factor in the spate of server problems we've seen recently, but we'd need input from a BOINC project server administrator to evaluate that.

Computer sind nicht alles im Leben. (Kleiner Scherz)

SETI@home classic workunits 3,758
SETI@home classic CPU time 66,520 hours
ID: 912649 · Report as offensive

Questions and Answers : Preferences : limit CPU Cores


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.