Welcome to MilkyWay@home

OpenCL for Nvidia available for testing

Message boards : News : OpenCL for Nvidia available for testing
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4

AuthorMessage
Profile arkayn
Avatar

Send message
Joined: 14 Feb 09
Posts: 999
Credit: 74,932,619
RAC: 0
Message 44860 - Posted: 9 Dec 2010, 23:36:12 UTC - in response to Message 44859.  

You are copying the app_info.xml file as well I hope, that is what tells BOINC what app to use with the project.
ID: 44860 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
europa

Send message
Joined: 29 Oct 10
Posts: 89
Credit: 39,246,947
RAC: 0
Message 44863 - Posted: 10 Dec 2010, 0:20:35 UTC - in response to Message 44860.  

Matt,

Yes, indeed.

Here it is.

<app_info>
&#8722;
<app>
<name>milkyway</name>
<user_friendly_name>Milkyway@home Separation</user_friendly_name>
</app>
&#8722;
<file_info>
&#8722;
<name>
milkyway_separation_0.48_x86_64-pc-linux-gnu__cuda_opencl
</name>
<executable/>
</file_info>
&#8722;
<app_version>
<app_name>milkyway</app_name>
<version_num>48</version_num>
<plan_class>cuda_opencl</plan_class>
<avg_ncpus>0.05</avg_ncpus>
<max_ncpus>0.05</max_ncpus>
<flops>1.0e11</flops>
&#8722;
<coproc>
<type>CUDA</type>
<count>1</count>
</coproc>
&#8722;
<file_ref>
&#8722;
<file_name>
milkyway_separation_0.48_x86_64-pc-linux-gnu__cuda_opencl
</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>


It appears identical to the one on the first machine.


Here's the body of one of the stderr.txt's from machine #2. Seems like old times!! :)

<core_client_version>6.10.58</core_client_version>
<![CDATA[
<message>
process exited with code 1 (0x1, -255)
</message>
<stderr_txt>
Device index specified on the command line was 0
Looking for a Double Precision capable NVIDIA GPU
The device GeForce GTX 460 from the command line cannot be used because a device supporting compute capability 1.3 (Double Precision) is required
Found 1 CUDA cards
Found a GeForce GTX 460
Device cannot be used, it does not have compute capability 1.3 support
No compute capability 1.3 cards have been found, exiting...

//////
I installed the latest Nvidia driver from their website but am I missing some sort of cuda library file? The last time, it was the installation of the OpenCL app that solved things. I can't figure out why and where the cuda23 executable keeps coming from or gets called for.

But, I think that's symptomatic of it still trying to run cuda23 WU's rather than OpenCL ones. Is there any way to keep the cuda23 executable from returning to the MW folder?

I noticed that the WU's have changed to MilkyWay@Home N-Body Simulation v0.21 (sse2). Maybe the MW server has given up sending me cuda23 WUs?

Regards,
Steve
ID: 44863 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Matt Arsenault
Volunteer moderator
Project developer
Project tester
Project scientist

Send message
Joined: 8 May 10
Posts: 576
Credit: 15,979,383
RAC: 0
Message 44866 - Posted: 10 Dec 2010, 1:12:17 UTC - in response to Message 44853.  

I'm trying now the 0.48.2 and I am having slowdowns my PC that I was not with 0.48.1 ... In addition, the unit does not and this calculation is boosting the temperature of my card ..
Apparently the formula that I used to keep the system responsive fell apart for the 470's specifications. The temperature is most definitely expected to go up.
ID: 44866 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile arkayn
Avatar

Send message
Joined: 14 Feb 09
Posts: 999
Credit: 74,932,619
RAC: 0
Message 44870 - Posted: 10 Dec 2010, 3:07:49 UTC

What are you using to open the .xml files, it almost looks like a browser.

Make sure you are opening them in Gedit instead.

Also try the 48.2 version that is also in the first post of the thread.

The app_info should looks something like this.

<app_info>
  <app>
    <name>milkyway</name>
    <user_friendly_name>Milkyway@home Separation</user_friendly_name>
  </app>

  <file_info>
    <name>milkyway_separation_0.48.2_x86_64-pc-linux-gnu__cuda_opencl</name>
    <executable/>
  </file_info>

  <app_version>
    <app_name>milkyway</app_name>
    <version_num>48</version_num>
    <plan_class>cuda_opencl</plan_class>
    <avg_ncpus>0.05</avg_ncpus>
    <max_ncpus>0.05</max_ncpus>
    <flops>1.0e11</flops>
    <coproc>
      <type>CUDA</type>
      <count>1</count>
    </coproc>

    <file_ref>
      <file_name>milkyway_separation_0.48.2_x86_64-pc-linux-gnu__cuda_opencl</file_name>
      <main_program/>
    </file_ref>
  </app_version>
</app_info>

ID: 44870 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
europa

Send message
Joined: 29 Oct 10
Posts: 89
Credit: 39,246,947
RAC: 0
Message 44887 - Posted: 10 Dec 2010, 14:59:57 UTC - in response to Message 44870.  

It does.

It occurred to me that there are two things that I did last time that I haven't tried yet. I haven't tried re-installing Boinc from a BOINC app site. This installation is from the Ubuntu repository. I did both last time.

The other thing was that I attached to Collatz which I understand has some good cuda libraries included. I remember copying over one of their cuda files the last time.

I'll have to try both later today.

Regards,
Steve
ID: 44887 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
europa

Send message
Joined: 29 Oct 10
Posts: 89
Credit: 39,246,947
RAC: 0
Message 44890 - Posted: 10 Dec 2010, 17:37:45 UTC - in response to Message 44812.  

Matt,

After doing a couple of Boinc re-installs and detach/attach's to MW, I think I've spotted the problem on the second machine but I don't know what's causing it or how to fix it.

On the good machine (both are virtually identical including the exact same model GPU card) at the start of the messages on the BOINC Manager it says:

Fri 10 Dec 2010 12:21:13 PM EST NVIDIA GPU 0: GeForce GTX 460 (driver version unknown, CUDA version 3020, compute capability 2.1, 767MB, 673 GFLOPS peak)

Found app_info.xml; using anonymous platform

Fri 10 Dec 2010 12:21:13 PM EST Milkyway@home URL http://milkyway.cs.rpi.edu/milkyway/; Computer ID 243370; resource share 100


But, on the problem machine it is missing the reference line to "app_info.xml; using anonymous platform" at that spot.

Fri 10 Dec 2010 12:21:13 PM EST NVIDIA GPU 0: GeForce GTX 460 (driver version unknown, CUDA version 3020, compute capability 2.1, 767MB, 673 GFLOPS peak)

Fri 10 Dec 2010 12:21:13 PM EST Milkyway@home URL http://milkyway.cs.rpi.edu/milkyway/; Computer ID 243370; resource share 100

I've tried using both the original 48.0 version and the 48.2 but no luck. I just saw your posting about the new version being pushed out by the server and the removal of the cuda version. I'm hoping that will fix things on this second machine.

Regards,
Steve
ID: 44890 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Matt Arsenault
Volunteer moderator
Project developer
Project tester
Project scientist

Send message
Joined: 8 May 10
Posts: 576
Credit: 15,979,383
RAC: 0
Message 44893 - Posted: 10 Dec 2010, 17:56:43 UTC - in response to Message 44890.  

But, on the problem machine it is missing the reference line to "app_info.xml; using anonymous platform" at that spot.
Are you sure you're putting it in the right place? It needs to be in the milkyway directory under projects.
I just saw your posting about the new version being pushed out by the server and the removal of the cuda version. I'm hoping that will fix things on this second machine.
The CUDA removal hasn't happened yet.
ID: 44893 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
europa

Send message
Joined: 29 Oct 10
Posts: 89
Credit: 39,246,947
RAC: 0
Message 44898 - Posted: 10 Dec 2010, 19:00:31 UTC - in response to Message 44893.  

Matt,

It's there. That's the nice thing about having the working machine right next to it. I've checked everything multiple times.

So, right now, I'm hanging on for the server refresh with the OpenCl version and the removal of the cuda23. I'm hoping that will "force" whatever correction is needed.

Thanks for the suggestions.

Regards,
Steve
ID: 44898 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
europa

Send message
Joined: 29 Oct 10
Posts: 89
Credit: 39,246,947
RAC: 0
Message 44919 - Posted: 11 Dec 2010, 14:47:48 UTC - in response to Message 44893.  

Matt,

An update. Everything is now working!

I did an uninstall/re-boot/re-install of the installation from the Ubuntu repository. I did not un-install the manual install from the BOINC site.

I then extracted the last OpenCL tar and deleted the cuda23 executable. Re-started the BOINC-Client and was greeted by the previously openCL config line in the messages. A WU popped up and has the notation that it is using 0.5CPU and 1.0 GPU and is chugging along.

A bonus, is that this has fixed all of the problems on this machine. Previously, all of the other projects on this machine with the exception of Seti were returning bad WU's.

It's been 2 hrs. and I'm now seeing credits for all of the projects for this machine.

I guess there's something that I needed from both installations.

Hooray!

Regards,
Steve
ID: 44919 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4

Message boards : News : OpenCL for Nvidia available for testing

©2024 Astroinformatics Group