Description of problem: The nVIDIA driver that is installed by Mageia 5 beta 3 is 346.35 and is OK. But I installed also nvidia-current-cuda-opencl-346.35-5.mga5.nonfree.x86_64.rpm And also installed the CUDA Toolkit directly from nVIDIA website su cuda_6.5.14_linux_64.run -override (-override is to be able to use Mageia's gcc 4.9.2) The simple toolkit sample deviceQuery returns 30 devices and gives error. [bert@Dell7720Mageia5 release]$ pwd /home/bert/Apps/NVIDIA_CUDA-6.5_Samples/bin/x86_64/linux/release [bert@Dell7720Mageia5 release]$ ./deviceQuery Then I created my own rpm's for nVIDIA with the script of Anssi: http://onse.fi/nvidia-mgabuild/ nvidia-current-cuda-opencl-346.35-1custom.nvidia.mga5.x86_64.rpm and 4 others and installed them with su rpm -Uvh --oldpackage *.rpm removed the 2 nvidia-current-kernel reboot Now CUDA samples work perfectly. CUDA in Mathematica 10.0.2 also works great. With the official package Mathematica behaved extremely strange. But I have no clue on what could be wrong with the offical rpm :-( Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Reproducible: Steps to Reproduce:
CC: (none) => tmbSource RPM: (none) => nvidia-current
CC: (none) => fri
Thanks. I followed your instruction and now I too have CUDA and OpenCL :)
CC: (none) => doktor5000
CC: (none) => danielosmari
Summary: Mageia 5 beta 3 nvidia-current-cuda-opencl-346.35-5.mga5.nonfree.x86_64.rpm is NOK => nvidia-current-cuda-opencl doesn't workSeverity: normal => majorSee Also: (none) => https://bugs.mageia.org/show_bug.cgi?id=15346Assignee: bugsquad => anssi.hannula
Is this bug still present in latest cauldron?
Keywords: (none) => NEEDINFO
WOOHO! Great, it works :) I cleaned skip.list, autoupdated and watched the locally built packages be replaced, rebooted, and the machine still crunch BOINC project with GPU (GPUGRID) Excerpt from start of BOINC log: lör 6 jun 2015 14:56:07 | | Starting BOINC client version 7.2.42 for x86_64-pc-linux-gnu lör 6 jun 2015 14:56:07 | | CUDA: NVIDIA GPU 0: GeForce GTX 760 (driver version unknown, CUDA version 7.0, compute capability 3.0, 2047MB, 1751MB available, 2469 GFLOPS peak) lör 6 jun 2015 14:56:07 | | OpenCL: NVIDIA GPU 0: GeForce GTX 760 (driver version 346.72, device version OpenCL 1.1 CUDA, 2047MB, 1751MB available, 2469 GFLOPS peak) lör 6 jun 2015 14:56:07 | | Processor: 8 GenuineIntel Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz [Family 6 Model 42 Stepping 7] lör 6 jun 2015 14:56:07 | | OS: Linux: 3.19.8-desktop-2.mga5 I close this. If anyone have problems please repoen.
Keywords: NEEDINFO => (none)Resolution: (none) => FIXEDStatus: NEW => RESOLVED
It was still working before christmas. Then GPUGRID, the only BOINC project i run which use CUDA, ran out of work. Then we got a new kernel and other updates. (i block nothing in skip.list) Now GPUGRID send work but all fail with computation error I also had one and fist time ever complete system freeze yesterday There are currently several open mageia issues regarding CUDA and OpenCL, you see when searching here. This is part of BOINC log when it starts: fre 15 jan 2016 11:07:09 | | Starting BOINC client version 7.2.42 for x86_64-pc-linux-gnu fre 15 jan 2016 11:07:09 | | log flags: file_xfer, sched_ops, task fre 15 jan 2016 11:07:09 | | Libraries: libcurl/7.40.0 OpenSSL/1.0.2e zlib/1.2.8 libidn/1.32 libssh2/1.4.3 fre 15 jan 2016 11:07:09 | | Data directory: /var/lib/boinc fre 15 jan 2016 11:07:09 | | CUDA: NVIDIA GPU 0: GeForce GTX 760 (driver version unknown, CUDA version 7.0, compute capability 3.0, 2047MB, 1769MB available, 2469 GFLOPS peak) fre 15 jan 2016 11:07:09 | | OpenCL: NVIDIA GPU 0: GeForce GTX 760 (driver version 346.96, device version OpenCL 1.1 CUDA, 2047MB, 1769MB available, 2469 GFLOPS peak) fre 15 jan 2016 11:07:09 | | Host name: svarten fre 15 jan 2016 11:07:09 | | Processor: 8 GenuineIntel Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz [Family 6 Model 42 Stepping 7] fre 15 jan 2016 11:07:09 | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr pdcm pcid sse4_1 sse4_2 popcnt tsc_deadline_timer aes xsave avx lahf_lm ida arat epb pln pts dtherm tpr_shadow vnmi flexpriority ept vpid xsaveopt fre 15 jan 2016 11:07:09 | | OS: Linux: 4.1.15-desktop-1.mga5 fre 15 jan 2016 11:07:09 | | Memory: 7.78 GB physical, 3.91 GB virtual
Version: Cauldron => 5Severity: major => normalStatus: RESOLVED => REOPENEDResolution: FIXED => (none)
Well you are using upstream cuda, so complain upstream
Can you explain why you think that? (see #3 and 4)
To refresh, I uninstalled nvidia-current-cuda-opencl, and was going to uninstall nvidia-cuda-toolkit when i noticed it was not installed - Strange. - maybe because some clash with wine i have been trying Bug 15346 - Lib64openCL1 with latest update of Wine conflicts with nvidia-cuda, also a Wine dependency. - Or Bug 17059 - cuda conflicts with opencl At this point i rebooted and verified BOINC do not find CUDA. Installing only nvidia-current-cuda-opencl and rebooting do not make BOINC find CUDA Then i installed also nvidia-cuda-toolkit - see https://bugs.mageia.org/show_bug.cgi?id=14462#c19 ...and rebooted, and now BOINC find CUDA and GPUGRID use it successfully :) BOINC however do not list OpenCL anymore (like it did in #3 and 4 above) - maybe bug 17059 - but i do not need OpenCL currently. So i close this again; it is probably the other bugs i mentioned that struck.
(In reply to Morgan Leijström from comment #6) > Can you explain why you think that? (see #3 and 4) both comment 3 and 4 shows: driver version unknown, CUDA version 7.0, we only have 6.5 in mga5 and cauldron Anyway, I guess I should try and sort out the cuda mess
Strange, BOINC again reports version 7 Anyhow it works now. Probably best to leave this open
uninstalled nvidia-cuda-toolkit, and installed lib64opencl1 ( file /usr/lib64/libOpenCL.so.1 of nvidia-cuda-toolkit-6.5.14-6.mga5.nonfree.x86_64 conflicts that file of lib64opencl1-2.1.3-3.mga5.x86_64 ) And now BOINC finds both OpenCL and CUDA - and GPUGRID works OK :) ...At the moment ;)
Another user solved it by manually linking https://forums.mageia.org/en/viewtopic.php?f=15&t=8711&p=61839#p61839
That same system got upgraded online to cauldron and boinc just still works :)
.. since that, my son have borrewed the Nvidia card a few weeks, and now when i put it back Boinc could not use it. This time i sucessfully tried the solution provided by Yann https://bugs.mageia.org/show_bug.cgi?id=14462#c19 I did not try building locally as per #0 I also did not try #11 And did not try linking like have worked for another user: https://forums.mageia.org/en/viewtopic.php?f=15&t=8711&p=61839#p61839 so... there are several working possibilities Can we have one default working solution please?
Whiteboard: (none) => MGA5TOOVersion: 5 => Cauldron
*** Bug 14462 has been marked as a duplicate of this bug. ***
Working on fresh install of Mageia 7, nvidia-current, installed boinc and nvidia-current-cuda-opencl and it just works. boinc-client-7.14.2-3.mga7.src.rpm 64 bit, and i let boincmgr start boinc after i log into Plasma.
Status: REOPENED => RESOLVEDResolution: (none) => FIXED