Cant run more than one WU per GPU

Questions and Answers : GPU applications : Cant run more than one WU per GPU
Message board moderation

To post messages, you must log in.

AuthorMessage
Aus_Karlos

Send message
Joined: 28 Apr 10
Posts: 16
Credit: 2,679,860
RAC: 0
Australia
Message 1745800 - Posted: 30 Nov 2015, 10:23:55 UTC

As the title says.

I have the config file app_config.xml like so.

<app_config>
<app>
<name>setiathome_v7</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
<app>
<name>astropulse_v6</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>


After starting the BOINC manager it will still only run 1 WU per card.
I've also tried max_concurent. In the forum ive found some other threads stating about editing this line below

<coproc>
<type>CUDA</type>
<count>0.5</count>
</coproc>

Where is this located, in the replies no one mentioned the location.
ID: 1745800 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15183
Credit: 4,362,181
RAC: 3
Netherlands
Message 1745823 - Posted: 30 Nov 2015, 12:36:10 UTC - in response to Message 1745800.  

For everything in the app_config.xml file, see http://boinc.berkeley.edu/wiki/Client_configuration#Application_configuration.

The <coproc> entry is part of the <options> in the client configuration file (cc_config.xml) which can be made by the user and stored in the main BOINC data directory. See http://boinc.berkeley.edu/wiki/Client_configuration#Options. It does not need to be used here, though. One can use it to define a coprocessor, such as one that isn't automatically recognized by BOINC itself.

To run more than one task on the GPU:
<app_config>
<app>
<name>setiathome_v7</name>
<max_concurrent>2</max_concurrent>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.04</cpu_usage>
</gpu_versions>
</app>
<app>
<name>astropulse_v7</name>
<max_concurrent>2</max_concurrent>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.04</cpu_usage>
</gpu_versions>
</app>
</app_config>


Store this file in the Seti@Home projects directory in the BOINC data directory. E.g. c:\programdata\boinc\projects\setiathome.berkeley.edu\
ID: 1745823 · Report as offensive
Profile BilBg
Volunteer tester
Avatar

Send message
Joined: 27 May 07
Posts: 3720
Credit: 9,385,827
RAC: 0
Bulgaria
Message 1745840 - Posted: 30 Nov 2015, 14:45:36 UTC - in response to Message 1745800.  
Last modified: 30 Nov 2015, 15:23:58 UTC

The file app_config.xml will be ignored if:
- not put in the right place, as noted it should be in projects' directory (and not in BOINC Data directory). put app_config.xml in:
SETI@home directory (<BOINC_Data>\projects\setiathome.berkeley.edu\)

- the text format is wrong (Unicode, UTF-8), the text format should be ANSI
- the filename is wrong (app_config.xml.txt), should be app_config.xml

- the non-existent apps will be ignored:
<name>astropulse_v6</name>

I don't think you need max_concurent as it will affect the total number of tasks, including CPU tasks
You don't need to free CPU core for the SETI@home CUDA app

<app_config>

   <app>
      <name>setiathome_v7</name>
      <gpu_versions>
         <gpu_usage>0.5</gpu_usage>
         <cpu_usage>0.04</cpu_usage>
      </gpu_versions>
   </app>

   <app>
      <name>astropulse_v7</name>
      <gpu_versions>
         <gpu_usage>0.5</gpu_usage>
         <cpu_usage>1</cpu_usage>
      </gpu_versions>
   </app>

</app_config>



This will/should run 2 tasks/GPU (total 4 - you have 2x GeForce GTX 980 Ti)
Every running AstroPulse OpenCL GPU task will free 1 CPU core (one less CPU task will be run by BOINC)
So don't also reduce the general BOINC Setting:
Use at most 100% of the CPUs

And for the AstroPulse OpenCL app you may need also set/put in the file with similar name (I just don't know the exact name on your system, the file should already exist and be empty (have zero length)):
ap_cmdline_win_x86_SSE2_OpenCL_NV.txt

... the following switch/line:
-instances_per_device 2

 
 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 
ID: 1745840 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15183
Credit: 4,362,181
RAC: 3
Netherlands
Message 1746055 - Posted: 1 Dec 2015, 8:53:04 UTC - in response to Message 1745840.  

I don't think you need max_concurent as it will affect the total number of tasks, including CPU tasks

I put the line into the app_info.xml example to show where it went.
The value tells the application Z named in the <name>Z</name> tags how many it can run at maximum at the same time, so yes, because we don't state the app by its correct name, but merely by overall name, all applications of this sort will only run max 2 at a time, be it CPU or GPU applications.
ID: 1746055 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15183
Credit: 4,362,181
RAC: 3
Netherlands
Message 1746243 - Posted: 2 Dec 2015, 8:13:21 UTC - in response to Message 1745800.  
Last modified: 2 Dec 2015, 8:13:31 UTC

<coproc>
<type>CUDA</type>
<count>0.5</count>
</coproc>

I also found that the above are part of the anonymous platform app_info.xml file. While you can use both an app_info.xml and app_config.xml file in the Seti project directories, you cannot interchange all options within them between them.
ID: 1746243 · Report as offensive

Questions and Answers : GPU applications : Cant run more than one WU per GPU


 
©2022 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.