Setting up Linux to crunch CUDA90 and above for Windows users

Message boards : Number crunching : Setting up Linux to crunch CUDA90 and above for Windows users
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 13 · 14 · 15 · 16 · 17 · 18 · 19 . . . 162 · Next

AuthorMessage
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1879397 - Posted: 21 Jul 2017, 7:03:52 UTC - in response to Message 1879395.  

With my Mint 17 computer (new build) I originally installed the Berkley version, checked what version it was (no setup done), then installed from the Repository and use that version (which was the same) and the Manager will never connect other than by IP.

It has never bugged me enough to really look at it, guess I should remove the files from my Home Folder and see if that works.
I have never worried much about it since I rarely restart the computer ....
ID: 1879397 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1879399 - Posted: 21 Jul 2017, 7:27:47 UTC
Last modified: 21 Jul 2017, 7:31:30 UTC

All I have to do is double click the boincmgr and everything runs.

It's possible you're missing a Dependency for the boinc file since that appears to be the one not starting.
To check dependencies open the Terminal and enter ldd then a space, then drag and drop the boinc app into the terminal window, click the window to make the terminal the front app, then hit enter. See if anything is missing, it will say File Not Found if something is missing. Do that to boincmgr too.
ID: 1879399 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1879401 - Posted: 21 Jul 2017, 7:56:28 UTC - in response to Message 1879399.  

OK, making progress. Removed the repository BOINC and am now running the 7.2.47 version in the Home folder. Had to copy the cc_config.xml file from my Windows machine as there wasn't one created in Linux. Manager complained about some options being invalid and I edited them out. Next, it complained it didn't have enough room and needed 32 MB more. Huh? It's a 120 GB drive. BOINC directory is only 238 MB. Figured out the global drive space limitation and fixed that.

Now the problem is that there is no OpenCL support. Huh? There was originally with the repository BOINC.
Fri 21 Jul 2017 12:38:38 AM PDT |  | Starting BOINC client version 7.2.47 for x86_64-pc-linux-gnu
Fri 21 Jul 2017 12:38:38 AM PDT |  | log flags: file_xfer, sched_ops, task
Fri 21 Jul 2017 12:38:38 AM PDT |  | Libraries: libcurl/7.47.0 OpenSSL/1.0.2g zlib/1.2.8 libidn/1.32 librtmp/2.3
Fri 21 Jul 2017 12:38:38 AM PDT |  | Data directory: /home/keith/BOINC_7.2.47
Fri 21 Jul 2017 12:38:38 AM PDT |  | CUDA: NVIDIA GPU 0: GeForce GTX 970 (driver version unknown, CUDA version 8.0, compute capability 5.2, 4034MB, 3881MB available, 6385 GFLOPS peak)
Fri 21 Jul 2017 12:38:38 AM PDT |  | CUDA: NVIDIA GPU 1: GeForce GTX 970 (driver version unknown, CUDA version 8.0, compute capability 5.2, 4037MB, 3973MB available, 6385 GFLOPS peak)
Fri 21 Jul 2017 12:38:38 AM PDT |  | CUDA: NVIDIA GPU 2: GeForce GTX 970 (driver version unknown, CUDA version 8.0, compute capability 5.2, 4037MB, 3973MB available, 6385 GFLOPS peak)
Fri 21 Jul 2017 12:38:38 AM PDT | SETI@home | Found app_info.xml; using anonymous platform
Fri 21 Jul 2017 12:38:38 AM PDT |  | App version needs OpenCL but GPU doesn't support it
Fri 21 Jul 2017 12:38:38 AM PDT |  | Host name: Darksider
Fri 21 Jul 2017 12:38:38 AM PDT |  | Processor: 8 AuthenticAMD AMD FX(tm)-8300 Eight-Core Processor [Family 21 Model 2 Stepping 0]
Fri 21 Jul 2017 12:38:38 AM PDT |  | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc extd_apicid aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 popcnt aes xsave avx f16c lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs xop skinit wdt lwp fma4 tce nodeid_msr tbm topoext perfctr_core perfctr_nb cpb hw_pstate vmmcall bmi1 arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold
Fri 21 Jul 2017 12:38:38 AM PDT |  | OS: Linux: 4.8.0-53-generic
Fri 21 Jul 2017 12:38:38 AM PDT |  | Memory: 15.57 GB physical, 15.90 GB virtual
Fri 21 Jul 2017 12:38:38 AM PDT |  | Disk: 93.77 GB total, 82.69 GB free
Fri 21 Jul 2017 12:38:38 AM PDT |  | Local time is UTC -7 hours
Fri 21 Jul 2017 12:38:38 AM PDT |  | Config: use all coprocessors
Fri 21 Jul 2017 12:38:38 AM PDT |  | Config: GUI RPC allowed from any host
Fri 21 Jul 2017 12:38:38 AM PDT | SETI@home | URL http://setiathome.berkeley.edu/; Computer ID 8306366; resource share 800
Fri 21 Jul 2017 12:38:38 AM PDT | SETI@home | General prefs: from SETI@home (last modified 21-Jul-2017 00:33:50)
Fri 21 Jul 2017 12:38:38 AM PDT | SETI@home | Host location: none
Fri 21 Jul 2017 12:38:38 AM PDT | SETI@home | General prefs: using your defaults
Fri 21 Jul 2017 12:38:38 AM PDT |  | Preferences:
Fri 21 Jul 2017 12:38:38 AM PDT |  | max memory usage when active: 7972.98MB
Fri 21 Jul 2017 12:38:38 AM PDT |  | max memory usage when idle: 14351.36MB
Fri 21 Jul 2017 12:38:38 AM PDT |  | max disk usage: 1.00GB
Fri 21 Jul 2017 12:38:38 AM PDT |  | (to change preferences, visit a project web site or select Preferences in the Manager)
Fri 21 Jul 2017 12:38:38 AM PDT |  | Not using a proxy
Fri 21 Jul 2017 12:38:39 AM PDT | SETI@home | Sending scheduler request: To fetch work.
Fri 21 Jul 2017 12:38:39 AM PDT | SETI@home | Requesting new tasks for CPU and NVIDIA


It looks like it is running the special app though, one each on each of the 3 970. Only one CPU task though. So making progress.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1879401 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1879402 - Posted: 21 Jul 2017, 8:18:46 UTC - in response to Message 1879401.  
Last modified: 21 Jul 2017, 8:30:42 UTC

I couldn't find where you listed the driver you have installed, and that version of BOINC only lists the driver in OpenCL. You might try installing a different driver, if you're running the Repo driver try the one from nVidia, and visa versa. The nVidia drivers are here, http://www.nvidia.com/Download/driverResults.aspx/118524/en-us that's the one I'm using in 16.04.

Depending on the packages you uninstalled, the OpenCL driver may have been uninstalled in the process. It might work if you just reinstalled the current driver.
ID: 1879402 · Report as offensive     Reply Quote
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1879403 - Posted: 21 Jul 2017, 8:30:53 UTC - in response to Message 1879402.  

For the NVidia driver ... Type, Device manager in the Mint Menu, install NVidia driver.
ID: 1879403 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1879404 - Posted: 21 Jul 2017, 9:13:33 UTC - in response to Message 1879402.  

I can't figure out where to locate the version number of the Nvidia drivers I'm using. I just accepted what the installer offered. I think it was 376.22 or something like that. I've downloaded the one you linked. Do I have to do the apt-get purge nvidia or whatever the command is .. before installing the new driver? I always used the Clean function in Windows in the installer.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1879404 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1879405 - Posted: 21 Jul 2017, 9:15:16 UTC - in response to Message 1879403.  

For the NVidia driver ... Type, Device manager in the Mint Menu, install NVidia driver.

There isn't a Device Manager in the Menu. Did you mean Driver Manager?
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1879405 · Report as offensive     Reply Quote
Profile tazzduke
Volunteer tester

Send message
Joined: 15 Sep 07
Posts: 190
Credit: 28,269,068
RAC: 5
Australia
Message 1879407 - Posted: 21 Jul 2017, 9:57:25 UTC - in response to Message 1879405.  

Greetings

Yes use that application.

Regards
ID: 1879407 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1879412 - Posted: 21 Jul 2017, 11:28:25 UTC - in response to Message 1879402.  

I couldn't find where you listed the driver you have installed, and that version of BOINC only lists the driver in OpenCL. You might try installing a different driver, if you're running the Repo driver try the one from nVidia, and visa versa. The nVidia drivers are here, http://www.nvidia.com/Download/driverResults.aspx/118524/en-us that's the one I'm using in 16.04.

Depending on the packages you uninstalled, the OpenCL driver may have been uninstalled in the process. It might work if you just reinstalled the current driver.

I've uninstalled the 375.66 package multiple times now and reinstalled selecting the OpenCL version. Still no sign of OpenCL in startup.

Tried to install the 385.22 package you referred. Finally figured out how to at least get it started. Then the installer complained it couldn't install without stopping the X server. Haven't figured that out yet.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1879412 · Report as offensive     Reply Quote
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22164
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1879415 - Posted: 21 Jul 2017, 12:03:43 UTC

If you are running the "special sauce" application there is no need to have OpenCL installed.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1879415 · Report as offensive     Reply Quote
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1879426 - Posted: 21 Jul 2017, 14:01:43 UTC - in response to Message 1879384.  
Last modified: 21 Jul 2017, 14:12:19 UTC

. . Hi Keith,

. . Anything I thought of has been covered ...

Stephen

good luck
ID: 1879426 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1879434 - Posted: 21 Jul 2017, 15:39:09 UTC - in response to Message 1879415.  

If you are running the "special sauce" application there is no need to have OpenCL installed.

Except I won't be able to run any AP tasks that might show up. I don't normally crunch for just SETI either. I also crunch for MilkyWay and Einstein and both need an OpenCL driver.

Anyone ever figure out how to get app_config.xml working in Linux? It reads the file fine at startup, just doesn't do anything with it.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1879434 · Report as offensive     Reply Quote
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1879436 - Posted: 21 Jul 2017, 15:42:23 UTC - in response to Message 1879434.  

app_config works the same as windows. I use it to reserve cores.
ID: 1879436 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1879439 - Posted: 21 Jul 2017, 15:55:29 UTC - in response to Message 1879436.  

app_config works the same as windows. I use it to reserve cores.

Can you point out where I'm wrong, please.
Fri 21 Jul 2017 08:45:10 AM PDT |  | Re-reading cc_config.xml
Fri 21 Jul 2017 08:45:10 AM PDT |  | Not using a proxy
Fri 21 Jul 2017 08:45:10 AM PDT |  | Config: use all coprocessors
Fri 21 Jul 2017 08:45:10 AM PDT |  | Config: GUI RPC allowed from any host
Fri 21 Jul 2017 08:45:10 AM PDT |  | log flags: file_xfer, sched_ops, task
Fri 21 Jul 2017 08:45:10 AM PDT | SETI@home | Found app_config.xml
Fri 21 Jul 2017 08:45:31 AM PDT | SETI@home | Computation for task 20dc08af.28130.20931.5.32.116_1 finished
Fri 21 Jul 2017 08:45:31 AM PDT | SETI@home | Starting task 15au08aa.7335.20113.12.39.181_1


It reads it without errors apparently.
<app_config>
  </app_version>
    <app_name>setiathome_v8</app_name>
    <plan_class>cuda80</plan_class>
    <avg_ncpus>1</avg_ncpus>
    <ngpus>1</ngpus>
  </app_version>

  <app_version>
    <app_name>astropulse_v7</app_name>
    <plan_class>opencl_nvidia_100</plan_class>
    <avg_ncpus>1</avg_ncpus>
    <ngpus>1</ngpus>
  </app_version>
<project_max_concurrent>8</project_max_concurrent>
</app_config>


I was trying to limit the CPU tasks running to keep the temps down and provide each GPU task a full core for feeding. It was running on all cores with 3 special sauce GPU tasks and 8 CPU tasks. As the system has a lowly FX-8300 which is really a 4 core machine with hyperthreading, I was trying to stay sensible.

I just redid the ncpu value in app_info.xml to reduce the number of CPU tasks running to five. It doesn't seem to obey the <project_max_concurrent>8</project_max_concurrent> option at all.

Any ideas as to what is wrong?
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1879439 · Report as offensive     Reply Quote
dallasdawg

Send message
Joined: 19 Aug 99
Posts: 49
Credit: 142,692,438
RAC: 2
United States
Message 1879442 - Posted: 21 Jul 2017, 16:03:46 UTC - in response to Message 1879434.  
Last modified: 21 Jul 2017, 16:10:16 UTC

If you are running the "special sauce" application there is no need to have OpenCL installed.

Except I won't be able to run any AP tasks that might show up. I don't normally crunch for just SETI either. I also crunch for MilkyWay and Einstein and both need an OpenCL driver.

Anyone ever figure out how to get app_config.xml working in Linux? It reads the file fine at startup, just doesn't do anything with it.


Give this a try in the terminal: sudo apt-get install nvidia-opencl-dev
This will install the complete opencl development package. Should fix any missing dependencies.

If that does not work, you can also try: sudo apt-get install nvidia-cuda-toolkit
I know it says CUDA, but from what I have read, it also installs opencl drivers as well.

You can always check the nvidia hardware and driver status by excuting nvidia-smi in a terminal.

You can also check driver info from the Software & Updates window via the Additional Drivers tab. Make sure you are on the Nvidia driver and not the Nouveau Driver.

Matt
ID: 1879442 · Report as offensive     Reply Quote
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1879446 - Posted: 21 Jul 2017, 16:21:37 UTC - in response to Message 1879439.  
Last modified: 21 Jul 2017, 16:24:15 UTC

I keep my app_config simple:
<app_config>
	<app>
	<name>setiathome_v8</name>
	<gpu_versions>
		<gpu_usage>1</gpu_usage>
		<cpu_usage>1.0</cpu_usage>
	</gpu_versions>
	</app>

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


I'm not certain, but you could have a fight going on with values stated in app_info.
To get around max concurrent you could over reserve the CPU with <cpu_usage>1.4</cpu_usage> with should reserve 4 cores for 3 tasks.

EDIT: For NVidia this is what I always do:
On initial Mint install, install NVidia drivers
sudo apt-get purge 'nvidia*'
sudo apt-get autoremove
reboot
Install the latest nvidia driver with the driver manager
reboot


NOTE: Suspend ALL GPU TASKS FIRST. The reboot and auto BOINC start WILL error out tasks on you.
ID: 1879446 · Report as offensive     Reply Quote
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1879450 - Posted: 21 Jul 2017, 16:30:43 UTC - in response to Message 1879439.  

Keith, As you seen not all thing work in BOINC 7.2 that worked in 7.6, as you seen by copying your cc_config over - you probably had errors on prosess_priority stuff, and other things. I'm not sure when max_concurrent was introduced, would have to check wiki.
ID: 1879450 · Report as offensive     Reply Quote
TBar
Volunteer tester

Send message
Joined: 22 May 99
Posts: 5204
Credit: 840,779,836
RAC: 2,768
United States
Message 1879453 - Posted: 21 Jul 2017, 16:47:13 UTC

I have 3 Quad core machines, one of them running 3 GPUs. All I do is set the Preferences to use 74% of the CPUs. That's it. Done. The machines run two Irreverent CPU tasks and the GPUs are happy. I'm amazed at the problems some people create for themselves. It really is very simple if you chose to make it so. I don't use an app_config as I don't have any use for it. I believe in making things as simple as possible, such as just downloading a package, expanding it, double clicking on an App, and having it work. Oh well ;-)
ID: 1879453 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1879462 - Posted: 21 Jul 2017, 17:50:59 UTC - in response to Message 1879450.  

Keith, As you seen not all thing work in BOINC 7.2 that worked in 7.6, as you seen by copying your cc_config over - you probably had errors on prosess_priority stuff, and other things. I'm not sure when max_concurrent was introduced, would have to check wiki.

Correct, there was no original cc_config.xml file so I copied the one I use on the Windows machines. I had about six commands that weren't recognized as valid and removed them. It never complained about max_project_concurrent. I will have to research the wiki to see at what version level it appeared.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1879462 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1879463 - Posted: 21 Jul 2017, 17:54:30 UTC - in response to Message 1879453.  

I have 3 Quad core machines, one of them running 3 GPUs. All I do is set the Preferences to use 74% of the CPUs. That's it. Done. The machines run two Irreverent CPU tasks and the GPUs are happy. I'm amazed at the problems some people create for themselves. It really is very simple if you chose to make it so. I don't use an app_config as I don't have any use for it. I believe in making things as simple as possible, such as just downloading a package, expanding it, double clicking on an App, and having it work. Oh well ;-)

Won't work for me with my machines which are quad core and octa core. Couldn't use the global CPU usage preferences. Would have to resort to venues and local preferences. I never could get venues to work across all my projects correctly. I crunch projects other than SETI BTW.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1879463 · Report as offensive     Reply Quote
Previous · 1 . . . 13 · 14 · 15 · 16 · 17 · 18 · 19 . . . 162 · Next

Message boards : Number crunching : Setting up Linux to crunch CUDA90 and above for Windows users


 
©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.