View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0000763 | channel: elrepo/el7 | primus | public | 2017-07-26 17:23 | 2017-08-08 13:44 |
| Reporter | vychytraly | Assigned To | toracat | ||
| Priority | normal | Severity | minor | Reproducibility | always |
| Status | closed | Resolution | fixed | ||
| Summary | 0000763: New primus update probably disables GL support | ||||
| Description | Hello friends, I was using primus package from elrepo and it was working really great, some softwares had occasional glitches (not sure if caused by them or primus) but it was nothing unbearable. Last update made these small glitches disappear, which was great, but today going through some settings in one graphics software I noticed that it is using Intel GPU instead of Nvidia. I was not sure of the cause so I turned it off and started it manually via terminal with optirun command, but I got the same result. Then I tried running glxspheres64 with optirun command and it was using Intel GPU. I am not sure but I think that last update somehow disabled GL support for bumblebee (optirun) and the small glitches disappeared because it was using Intel GPU all the time. I think that support for other functionality remained ok, since I was running some CUDA demos with optirun command which seemed working fine (and they did not work without optirun command). I tried installing old primus package from akemi ( https://elrepo.org/people/akemi/testing/el7/primus/ ) and now glxshperes64 (when run with optirun command) and also the graphics softwares are using NVidia again. I am not sure if all this works as I described and new primus package is really the cause of this issues but it just seems to me this way and I wanted to ask you if you think this could be the problem. Thank you very much and have a nice day :) | ||||
| Tags | No tags attached. | ||||
| Reported upstream | |||||
|
|
I wanted to test it also with previous version of primus but I cant find it in repository anymore |
|
|
Or maybe is it all working properly, primus is using Nvidia GPU as it should, just the softwares (glxspheres64...) are somehow not able to recognize that they are actually using Nvidia GPU? |
|
|
Acknowledged. I did make a couple other small changes to the package, so it is not inconceivable that we could have broken something. Could you double check your findings, and if necessary we could build a test package reverting some of the changes. A full list of changes is available here to view: https://github.com/elrepo/packages/commit/064a7e2c20b20cee9f52e6c1df47dbbcc6930566#diff-738a9830e025c2886f497b16747588e5 |
|
|
I will try to test it as much as my skills will allow and report my findings. Testing packages like glxinfo, clinfo, glxspheres64, cuda demos come to my mind, if you had any more recommendations how can I test it please let me know :) |
|
|
Today I made few tests with last primus package and got these results: clinfo command shows 2 available devices: intel i7 cpu and intel hd graphics optirun clinfo command shows 3 available devices: nvidia gpu, intel i7 cpu and intel hd graphics (I think this output was expected) glxinfo command shows me these lines: OpenGL vendor string: Intel Open Source Technology Center OpenGL renderer string: Mesa DRI Intel(R) HD Graphics 530 (Skylake GT2) optirun glxinfo command shows me these lines: OpenGL vendor string: Intel Open Source Technology Center OpenGL renderer string: Mesa DRI Intel(R) HD Graphics 530 (Skylake GT2) (Here I expected that in case using optirun glxinfo command, OpenGL vendor and renderer string would be NVidia - I think in previous primus versions it worked that way) glxgears -info command shows these lines: GL_RENDERER = Mesa DRI Intel(R) HD Graphics 530 (Skylake GT2) GL_VERSION = 3.0 Mesa 11.2.2 GL_VENDOR = Intel Open Source Technology Center optirun glxgears -info shows these lines: GL_RENDERER = Mesa DRI Intel(R) HD Graphics 530 (Skylake GT2) GL_VERSION = 3.0 Mesa 11.2.2 GL_VENDOR = Intel Open Source Technology Center (Here again when running optirun glxgears command I expected NVidia in GL_RENDERER and GL_VENDOR lines) glxspheres64 command shows this: Context is Direct OpenGL Renderer: Mesa DRI Intel(R) HD Graphics 530 (Skylake GT2) optirun glxspheres64 command shows this: Context is Direct OpenGL Renderer: Mesa DRI Intel(R) HD Graphics 530 (Skylake GT2) (Here again when running optirun glxspheres64 command I expected NVidia in OpenGL Renderer line) nvidia-smi shows this: +-----------------------------------------------------------------------------+ | NVIDIA-SMI 384.59 Driver Version: 384.59 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | |===============================+======================+======================| | 0 GeForce GTX 960M Off | 00000000:02:00.0 Off | N/A | | N/A 42C P0 N/A / N/A | 0MiB / 4044MiB | 0% Default | +-------------------------------+----------------------+----------------------+ +-----------------------------------------------------------------------------+ | Processes: GPU Memory | | GPU PID Type Process name Usage | |=============================================================================| | No running processes found | +-----------------------------------------------------------------------------+ optirun nvidia-smi shows this: +-----------------------------------------------------------------------------+ | NVIDIA-SMI 384.59 Driver Version: 384.59 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | |===============================+======================+======================| | 0 GeForce GTX 960M Off | 00000000:02:00.0 Off | N/A | | N/A 42C P0 N/A / N/A | 5MiB / 4044MiB | 0% Default | +-------------------------------+----------------------+----------------------+ +-----------------------------------------------------------------------------+ | Processes: GPU Memory | | GPU PID Type Process name Usage | |=============================================================================| | 0 27319 G Xorg 4MiB | +-----------------------------------------------------------------------------+ (We can see here that when running optirun nvidia-smi, NVidia is running (or at least trying to run) Xorg) deviceQuery utility from CUDA samples shows this: CUDA Device Query (Runtime API) version (CUDART static linking) cudaGetDeviceCount returned 35 -> CUDA driver version is insufficient for CUDA runtime version Result = FAIL optirun deviceQuery shows this: CUDA Device Query (Runtime API) version (CUDART static linking) Detected 1 CUDA Capable device(s) Device 0: "GeForce GTX 960M" CUDA Driver Version / Runtime Version 9.0 / 8.0 CUDA Capability Major/Minor version number: 5.0 Total amount of global memory: 4044 MBytes (4240965632 bytes) ( 5) Multiprocessors, (128) CUDA Cores/MP: 640 CUDA Cores GPU Max Clock rate: 1176 MHz (1.18 GHz) Memory Clock rate: 2505 Mhz Memory Bus Width: 128-bit L2 Cache Size: 2097152 bytes Maximum Texture Dimension Size (x,y,z) 1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096) Maximum Layered 1D Texture Size, (num) layers 1D=(16384), 2048 layers Maximum Layered 2D Texture Size, (num) layers 2D=(16384, 16384), 2048 layers Total amount of constant memory: 65536 bytes Total amount of shared memory per block: 49152 bytes Total number of registers available per block: 65536 Warp size: 32 Maximum number of threads per multiprocessor: 2048 Maximum number of threads per block: 1024 Max dimension size of a thread block (x,y,z): (1024, 1024, 64) Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535) Maximum memory pitch: 2147483647 bytes Texture alignment: 512 bytes Concurrent copy and kernel execution: Yes with 1 copy engine(s) Run time limit on kernels: Yes Integrated GPU sharing Host Memory: No Support host page-locked memory mapping: Yes Alignment requirement for Surfaces: Yes Device has ECC support: Disabled Device supports Unified Addressing (UVA): Yes Device PCI Domain ID / Bus ID / location ID: 0 / 2 / 0 Compute Mode: < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) > deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 9.0, CUDA Runtime Version = 8.0, NumDevs = 1, Device0 = GeForce GTX 960M Result = PASS (I think this again works as expected) Looking at these results I think that primus runs great when trying to use OpenCL or CUDA - it uses NVidia GPU as expected, but when trying to run OpenGL software it does not make OpenGL run on NVidia but on Intel (before the last update it was running also on NVidia). I dont know any other tests how to try this, if I should run anything else please let me know :) |
|
|
I can confirm that when running optirun with virtualgl (not primus bridge) glxgears64 runs on NVidia: $ optirun -b virtualgl glxspheres64 Context is Direct OpenGL Renderer: GeForce GTX 960M/PCIe/SSE2 Same with glxgears $ optirun -b virtualgl glxgears -info GL_RENDERER = GeForce GTX 960M/PCIe/SSE2 GL_VERSION = 4.5.0 NVIDIA 384.59 GL_VENDOR = NVIDIA Corporation Same with glxinfo $ optirun -b virtualgl glxinfo OpenGL vendor string: NVIDIA Corporation OpenGL renderer string: GeForce GTX 960M/PCIe/SSE2 I think that this also shows that primus in current state does not work with OpenGL as it should :/ |
|
|
But when looking at github changes I am not really sure what could be the problem :/ maybe we should somehow point primus to use nvidia GL libraries instead of Intel ones? |
|
|
Can you confirm the original primus package that Akemi built worked? If so, I will rebuild the package reverting the changes I made except to filter the libGL provides, and that should establish if it's something we broke in the latest package. |
|
|
Ok I installed Akemi's primus package and got these results: optirun glxspheres64 Context is Direct OpenGL Renderer: GeForce GTX 960M/PCIe/SSE2 optirun glxgears -info GL_RENDERER = GeForce GTX 960M/PCIe/SSE2 GL_VERSION = 4.5.0 NVIDIA 384.59 GL_VENDOR = NVIDIA Corporation optirun glxinfo OpenGL vendor string: NVIDIA Corporation OpenGL renderer string: GeForce GTX 960M/PCIe/SSE2 (Everything seems fine) |
|
|
OK, I'll revert the changes I made and lets see if we can figure out what caused it to break. Will post back here once done. |
|
|
Ok thank you very much, looking forward to test it :) |
|
|
I've reverted the changes other than filtering the Provides for libGL.so, and pushed primus-20150328-3.el7.elrepo.x86_64.rpm to the testing repo. It should show up here shortly: http://elrepo.org/linux/testing/el7/x86_64/RPMS/ Hopefully that should fix things. |
|
|
Thank you very much Phil, now it seems working with OpenGL applications as before :) but while dealing with this new primus package and testing it I probably discovered a bug regarding OpenCL running on NVidia - OpenCL is discovered by applications just fine, but they cant use it - I keep getting errors like this: 1 devices found supporting OpenCL: Error 221 in clCreateContext call !!! or: [opencl_init] could not create context for device 0: 221 [opencl_init] FINALLY: opencl is NOT AVAILABLE on this system. [opencl_init] initial status of opencl enabled flag is OFF. I also tested the previous primus package and I get the same results, so it seems that it is not an issue of primus. Im not sure if its bug of the new kmod-nvidia driver branch or it was present before. Is some additional configuration needed for OpenCL to work correctly with kmod-nvidia? I have installed ocl-icd, opencl-filesystem and opencl-header on my machine for a long time and with Intel CPU/GPU OpenCL seems working fine, its just NVidia which seems having problems. |
|
|
could it be caused by the fact that current nvidia driver uses CUDA 9 for OpenCL (this is output from clinfo): Platform Version OpenCL 1.2 CUDA 9.0.130 but NVidia only provides CUDA 8 for download? |
|
|
OK, pleased we seen to have resolved the initial primus issue. I've no idea on the OpenCL issue as I don't use it or CUDA I'm afraid. Can I suggest you try downgrading the nvidia drivers to 375.66 to see if they work, and then open a new bug if required for that issue. |
|
|
Thank you again very much Phil. And yes you are right, after downgrading kmod-nvidia driver OpenCL works fine, so its probably bug of current version, I will open a new thread. |
|
|
I think that we can mark this issue as resolved :) |
|
|
(New Primus seems working as before :) ) |
|
|
Thanks |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2017-07-26 17:23 | vychytraly | New Issue | |
| 2017-07-26 17:23 | vychytraly | Status | new => assigned |
| 2017-07-26 17:23 | vychytraly | Assigned To | => toracat |
| 2017-07-26 17:25 | vychytraly | Note Added: 0005341 | |
| 2017-07-26 17:46 | vychytraly | Note Added: 0005342 | |
| 2017-07-27 00:24 | pperry | Note Added: 0005346 | |
| 2017-07-27 14:07 | vychytraly | Note Added: 0005348 | |
| 2017-07-29 13:22 | vychytraly | Note Added: 0005355 | |
| 2017-07-29 13:37 | vychytraly | Note Added: 0005356 | |
| 2017-07-29 13:41 | vychytraly | Note Added: 0005357 | |
| 2017-07-29 13:49 | pperry | Note Added: 0005358 | |
| 2017-07-29 13:57 | vychytraly | Note Added: 0005359 | |
| 2017-07-29 14:16 | pperry | Note Added: 0005360 | |
| 2017-07-29 14:17 | vychytraly | Note Added: 0005361 | |
| 2017-07-29 14:37 | pperry | Note Added: 0005362 | |
| 2017-07-30 09:31 | vychytraly | Note Added: 0005363 | |
| 2017-07-30 09:52 | vychytraly | Note Added: 0005364 | |
| 2017-07-30 10:55 | pperry | Note Added: 0005365 | |
| 2017-07-30 11:26 | vychytraly | Note Added: 0005366 | |
| 2017-08-07 14:13 | vychytraly | Note Added: 0005379 | |
| 2017-08-07 14:16 | vychytraly | Note Added: 0005381 | |
| 2017-08-08 13:44 | pperry | Note Added: 0005382 | |
| 2017-08-08 13:44 | pperry | Status | assigned => closed |
| 2017-08-08 13:44 | pperry | Resolution | open => fixed |
