Collatz on GPUs with Linux
log in

Advanced search

Message boards : Unix/Linux : Collatz on GPUs with Linux

Author Message
Profile Gandalf_the_Grey
Avatar
Send message
Joined: 2 May 13
Posts: 40
Credit: 336,096,852
RAC: 0
Message 19054 - Posted: 13 Apr 2014, 12:44:14 UTC

I have two PC running Linux. However, I can not download any WUs for the GPUs.
Am I missing something? Do I also need to download something to make this work?
The OSs are Trisquel and Linux Mint (Cinnamon).

Thanks
____________

Profile Gandalf_the_Grey
Avatar
Send message
Joined: 2 May 13
Posts: 40
Credit: 336,096,852
RAC: 0
Message 19056 - Posted: 14 Apr 2014, 2:00:10 UTC - in response to Message 19054.

I just rediscovered that when I look into the /var/lib/boinc-client/projects/boinc.thesonntags.com_collatz folder, it is empty. That is why I can't run any WUs.

Why is it empty?
____________

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19057 - Posted: 14 Apr 2014, 18:15:43 UTC

Could be several reasons...
First, do you have the GPU drivers installed? Linux distros often include only the part that handles the display and not the CUDA or OpenCL parts that allow crunching. Those have to be added manually (or through the package manager) or from the GPU vendor's site (AMD or nVidia). Does BOINC show that you have GPUs that can run CUDA or OpenCL in the event log?

Check your prefences on the Collatz size to make sure you have GPU crunching enabled.

Check that you have <use_all_gpus>1</use_all_gpus> in your cc_config.xml

Exit and restart BOINC. I've had several instances where it was "stuck" and restarting BOINC fixed it.

Finally, BOINC will NOT ask for work if it thinks it has enough work already. In recent versions, it includes work for suspended projects. So, you you have your cache set for 2 days and have 2 days of PG or SETI or some other project, it won't ask for work from Collatz even if those projects are suspended. No, it doesn't make sense to me, but that's the way they [the BOINC developers and a few project admins] want it. The same thing can happen if you shorten the cache length. BOINC doesn't care that you can finish other tasks after the challenge is over because it doesn't look at the deadline, only the cache size. My solution is to just abort the other tasks which I will do until it gets changed back so that suspended means suspended rather than paused.

Hint: enable sched_op_debug and html_debug in the cc_config.xml logging section. If it isn't asking for > 0 seconds of GPU work, then it is either a driver or a cache is too full issue. If it asks for > 0 seconds but doesn't get any, it is likely a preferences setting issue.

Profile Gandalf_the_Grey
Avatar
Send message
Joined: 2 May 13
Posts: 40
Credit: 336,096,852
RAC: 0
Message 19059 - Posted: 14 Apr 2014, 20:38:16 UTC

Thanks Slicker for the reply.

I add <use_all_gpus>1</use_all_gpus>

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


boinc-amd-opencl was included/installed with the boinc package.

I do not know how to implement your "Hint".
Here is the current file...
<cc_config>
<log_flags>
<task>1</task>
<file_xfer>1</file_xfer>
<sched_ops>1</sched_ops>
</log_flags>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>


Here is the Event log...
Mon 14 Apr 2014 01:29:18 PM PDT | Collatz Conjecture | update requested by user
Mon 14 Apr 2014 01:29:21 PM PDT | Collatz Conjecture | Sending scheduler request: Requested by user.
Mon 14 Apr 2014 01:29:21 PM PDT | Collatz Conjecture | Not requesting tasks: don't need
Mon 14 Apr 2014 01:29:22 PM PDT | Collatz Conjecture | Scheduler request completed


I'll reinstall the Collatz project and cross my fingers.
____________

Profile Gandalf_the_Grey
Avatar
Send message
Joined: 2 May 13
Posts: 40
Credit: 336,096,852
RAC: 0
Message 19060 - Posted: 14 Apr 2014, 20:51:16 UTC

I removed and reinstalled Collatz, but no joy. The folder is still empty.
Even if I didn't have the correct drivers, would that explain why the folder is empty?

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19062 - Posted: 14 Apr 2014, 22:02:44 UTC - in response to Message 19060.

I removed and reinstalled Collatz, but no joy. The folder is still empty.
Even if I didn't have the correct drivers, would that explain why the folder is empty?


sched_ops and sched_op_debug are two different things. The latter will show how many seconds of work are requested and whether they are for CPU, ATI/CAL, CUDA, or OpenCL.

When you first start BOINC, does the event log show that you have a CUDA, CAL, or OpenCL coprocessor?

For example:

4/14/2014 5:01:17 PM | | Data directory: C:\ProgramData\BOINC
4/14/2014 5:01:17 PM | | Running under account Jon
4/14/2014 5:01:17 PM | | CUDA: NVIDIA GPU 0: GeForce GTX 570 (driver version 335.23, CUDA version 6.0, compute capability 2.0, 1280MB, 1176MB available, 1478 GFLOPS peak)
4/14/2014 5:01:17 PM | | CAL: ATI GPU 0: AMD Radeon HD 6900 series (Cayman) (CAL version 1.4.1848, 2048MB, 2016MB available, 6758 GFLOPS peak)
4/14/2014 5:01:17 PM | | OpenCL: NVIDIA GPU 0: GeForce GTX 570 (driver version 335.23, device version OpenCL 1.1 CUDA, 1280MB, 1176MB available, 1478 GFLOPS peak)
4/14/2014 5:01:17 PM | | OpenCL: AMD/ATI GPU 0: AMD Radeon HD 6900 series (Cayman) (driver version 1348.5 (VM), device version OpenCL 1.2 AMD-APP (1348.5), 2048MB, 2016MB available, 6758 GFLOPS peak)
4/14/2014 5:01:17 PM | | OpenCL CPU: Intel(R) Core(TM) i7 CPU 950 @ 3.07GHz (OpenCL driver vendor: Advanced Micro Devices, Inc., driver version 1348.5 (sse2), device version OpenCL 1.2 AMD-APP (1348.5))
4/14/2014 5:01:17 PM | | Host name: i7-950

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19063 - Posted: 14 Apr 2014, 22:03:33 UTC - in response to Message 19060.
Last modified: 14 Apr 2014, 22:03:44 UTC

I removed and reinstalled Collatz, but no joy. The folder is still empty.
Even if I didn't have the correct drivers, would that explain why the folder is empty?


If you have only GPU apps selected in your preferences and you have no GPUs visible to BOINC, then yes.

Profile Gandalf_the_Grey
Avatar
Send message
Joined: 2 May 13
Posts: 40
Credit: 336,096,852
RAC: 0
Message 19064 - Posted: 14 Apr 2014, 23:48:53 UTC

Are the files that should be in that folder the same files that are in my Win7 PC?

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19065 - Posted: 15 Apr 2014, 14:19:47 UTC - in response to Message 19064.

Are the files that should be in that folder the same files that are in my Win7 PC?


No. The executables on your windows box are for windows. Your linux box needs the linux versions. Copying the files manually will NOT help.


<cc_config>
<log_flags>
<task>1</task>
<file_xfer>1</file_xfer>
<sched_ops>1</sched_ops>
<sched_op_debug>1</sched_op_debug>
</log_flags>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

The complete list of options is found at: https://boinc.berkeley.edu/wiki/Client_configuration

Post the section of the BOINC event log showing how it identifies your CPU and coprocessors and also the lines where it requests X seconds of work. The latter should look something like:



4/15/2014 9:18:29 AM | Collatz Conjecture | Reporting 2 completed tasks
4/15/2014 9:18:29 AM | Collatz Conjecture | Requesting new tasks for CPU and NVIDIA and ATI
4/15/2014 9:18:29 AM | Collatz Conjecture | [sched_op] CPU work request: 8406.15 seconds; 0.00 devices
4/15/2014 9:18:29 AM | Collatz Conjecture | [sched_op] NVIDIA work request: 7227.86 seconds; 0.00 devices
4/15/2014 9:18:29 AM | Collatz Conjecture | [sched_op] ATI work request: 7570.63 seconds; 0.00 devices
4/15/2014 9:18:31 AM | Collatz Conjecture | Scheduler request completed: got 19 new tasks

Profile Crunch3r
Volunteer moderator
Project developer
Project tester
Avatar
Send message
Joined: 30 Jun 09
Posts: 219
Credit: 7,515,663,568
RAC: 8,898
Message 19066 - Posted: 15 Apr 2014, 18:34:10 UTC - in response to Message 19065.
Last modified: 15 Apr 2014, 18:35:19 UTC

It's quite obvious what's going wrong here. You don't have the proper graphics drivers installed on your machines and that's why boinc can't detect any usable machinesb (hence no work for your GPUs).

Download and install the drivers from AMD/Nvidia.

That'll fix the problem.
____________

Team BOINC United.Join Science that matters.

Profile Gandalf_the_Grey
Avatar
Send message
Joined: 2 May 13
Posts: 40
Credit: 336,096,852
RAC: 0
Message 19071 - Posted: 16 Apr 2014, 1:27:15 UTC

On one of my PCs, I attempted to install...
http://support.amd.com/en-us/download/desktop?os=Linux+x86

Received this error message...


File fglrx-install.log

Supported adapter detected.
Supported adapter detected.
Check if system has the tools required for installation.
fglrx installation requires that the system have kernel headers. /lib/modules/3.2.0-60-generic/build/include/linux/version.h cannot be found on this system.
fglrx installation requires that the system has make tool. make cannot be found on this system.
fglrx installation requires that the system has gcc tool. gcc cannot be found on this system.
One or more tools required for installation cannot be found on the system. Install the required tools before installing the fglrx driver.
Optionally, run the installer with --force option to install without the tools.
Forcing install will disable AMD hardware acceleration and may make your system unstable. Not recommended.


Where do I go from here?

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19077 - Posted: 16 Apr 2014, 23:13:31 UTC - in response to Message 19071.

On one of my PCs, I attempted to install...
http://support.amd.com/en-us/download/desktop?os=Linux+x86

Received this error message...


File fglrx-install.log
Supported adapter detected.
Supported adapter detected.
Check if system has the tools required for installation.
fglrx installation requires that the system have kernel headers. /lib/modules/3.2.0-60-generic/build/include/linux/version.h cannot be found on this system.
fglrx installation requires that the system has make tool. make cannot be found on this system.
fglrx installation requires that the system has gcc tool. gcc cannot be found on this system.
One or more tools required for installation cannot be found on the system. Install the required tools before installing the fglrx driver.
Optionally, run the installer with --force option to install without the tools.
Forcing install will disable AMD hardware acceleration and may make your system unstable. Not recommended.


Where do I go from here?


If I'm not mistaken, fglrx is only needed for ATI GPUs that use CAL. I wouldn't bother with it. I start with getting OpenCL working if it is AMD or CUDA if it is nVidia. On Debian and Ubuntu, the OpenCL and CUDA libraries do not get installed with the video driver. They have to be installed separately. I can't speak for other distros but that's probably the case for them as well.

Profile Gandalf_the_Grey
Avatar
Send message
Joined: 2 May 13
Posts: 40
Credit: 336,096,852
RAC: 0
Message 19079 - Posted: 17 Apr 2014, 0:31:22 UTC

I resolved my issue. I replaced the ATI GPUs with NV GPUs.


Post to thread

Message boards : Unix/Linux : Collatz on GPUs with Linux


Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.