1 Reply by Veux

(449 replies, posted in Using SVP)

It makes no difference whether I try to run mpv with the SVPManager gui option, in a terminal with mpv, or in a terminal with optirun mpv, for all cases. Also, the error given when trying to use the 960M with SVP is only reported once, after the SVP settings are cleared to default.  After the first failure, SVP silently reverts to cpu rendering every time, ignoring the gpu acceleration setting. I'm not sure if that behavior is normal.

I forgot one big thing.  The 960M does not show up at all in svp with "optirun SVPManager" unless I add "alias nvidia-uvm nvidia-387-uvm" to /etc/modprobe.d/bumblebee.conf

Edit:
I finally managed to fix this. If anyone stumbles on this from google, I fixed it by installing the cuda toolkit .deb from nvidia's developer website https://developer.nvidia.com/cuda-relea … e-download rather than the cuda toolkit in the repos. This also reverted my drivers to 384. Install it with: 

sudo dpkg -i cuda-repo-*.deb
sudo apt-key add /var/cuda-repo-*/7fa2af80.pub
sudo apt-get update
sudo apt-get install cuda

In /etc/modprobe.d/ add: blacklist nvidia-384-drm or whatever driver version it is you're using. Don't use the alias from my previous posts.

Before running svp, use: 

sudo prime-select nvidia

I have no idea why luxmark was able to work while svp failed with my previous setup... shrug. This also fixed the intel igpu problems. Hope this helps someone.

Go to forum: Using SVP

2 Reply by Veux

(449 replies, posted in Using SVP)

I have an XPS15 9550 with optimus graphics. I'm trying to get SVP working with optirun. I have nvidia-387, bumblebee, bumblebee-nvidia, primus, nvidia-opencl-icd-387, and nvidia-modprobe for nvidia opencl, and beignet-opencl-icd for intel opencl.

clinfo shows: Intel(R) HD Graphics Skylake Halo GT2.
optirun clinfo shows both Intel(R) HD Graphics Skylake Halo GT2 ... GeForce GTX 960M ... Device Version    OpenCL 1.2 CUDA

I have run luxmark with and without optirun, and it behaves as expceted, both igpu and dgpu show up and work with optirun, so opencl appears to be set up correctly for both GPUs.

If I run SVPManager, Intel(R) HD Graphics Skylake Halo GT2 is detected, but when I assess system performance, it fails on the intel gpu with:

20:47:26.975 [W]: Video [gpuID=11] initialization failed
20:47:26.975 [W]: Performance: GPU-based frame rendering [gpuID=11] FAILED

and videos seem to default to cpu rendering when the intel gpu is selected, because cpu use is ~75% when I select the intel gpu and when I select cpu rendering.

I had to add "alias nvidia-uvm nvidia-387-uvm" to /etc/modprobe.d/bumblebee.conf to get the 960M to show up at all in svp with "optirun SVPManager".
If i run optimus SVPManager both Intel(R) HD Graphics Skylake Halo GT2 and GeForce GTX 960M are found, and assess system performance works for the 960M, but playing videos with mpv defaults to cpu rendering yet again and the log file shows:

04:20:51.153 [i]: Playback [f8c72de1]: playing at 59.94 [23.976 *5/2] 
04:20:51.331 [E]: Playback [f8c72de1]: VS - Script evaluation failed:
04:20:51.331 [E]: Playback [f8c72de1]: VS - Python exception: SVSmoothFps: unable to init GPU-based renderer - code 65536
04:20:51.332 [E]: Playback [f8c72de1]: VS - Traceback (most recent call last):
04:20:51.332 [E]: Playback [f8c72de1]: VS - File 'src/cython/vapoursynth.pyx', line 1810, in vapoursynth.vpy_evaluateScript (src/cython/vapoursynth.c:36455)
04:20:51.332 [E]: Playback [f8c72de1]: VS - File '/home/cake/.local/share/SVP4/scripts/f8c72de1.py', line 53, in <module>
04:20:51.332 [E]: Playback [f8c72de1]: VS - smooth = interpolate(clip)
04:20:51.332 [E]: Playback [f8c72de1]: VS - File '/home/cake/.local/share/SVP4/scripts/f8c72de1.py', line 36, in interpolate
04:20:51.332 [E]: Playback [f8c72de1]: VS - smooth = core.svp2.SmoothFps(input_m,super['clip'],super['data'],vectors['clip'],vectors['data'],smoothfps_params,src=input_um,fps=src_fps)
04:20:51.332 [E]: Playback [f8c72de1]: VS - File 'src/cython/vapoursynth.pyx', line 1702, in vapoursynth.Function.__call__ (src/cython/vapoursynth.c:34595)
04:20:51.332 [E]: Playback [f8c72de1]: VS - vapoursynth.Error: SVSmoothFps: unable to init GPU-based renderer - code 65536
04:20:51.381 [E]: Playback [f8c72de1]: VS - Script evaluation failed:
04:20:51.381 [E]: Playback [f8c72de1]: VS - Python exception: SVSmoothFps: unable to init GPU-based renderer - code 65536
04:20:51.381 [E]: Playback [f8c72de1]: VS - Traceback (most recent call last):
04:20:51.381 [E]: Playback [f8c72de1]: VS - File 'src/cython/vapoursynth.pyx', line 1810, in vapoursynth.vpy_evaluateScript (src/cython/vapoursynth.c:36455)
04:20:51.381 [E]: Playback [f8c72de1]: VS - File '/home/cake/.local/share/SVP4/scripts/f8c72de1.py', line 53, in <module>
04:20:51.381 [E]: Playback [f8c72de1]: VS - smooth = interpolate(clip)
04:20:51.381 [E]: Playback [f8c72de1]: VS - File '/home/cake/.local/share/SVP4/scripts/f8c72de1.py', line 36, in interpolate
04:20:51.381 [E]: Playback [f8c72de1]: VS - smooth = core.svp2.SmoothFps(input_m,super['clip'],super['data'],vectors['clip'],vectors['data'],smoothfps_params,src=input_um,fps=src_fps)
04:20:51.382 [E]: Playback [f8c72de1]: VS - File 'src/cython/vapoursynth.pyx', line 1702, in vapoursynth.Function.__call__ (src/cython/vapoursynth.c:34595)
04:20:51.382 [E]: Playback [f8c72de1]: VS - vapoursynth.Error: SVSmoothFps: unable to init GPU-based renderer - code 65536

So I'm unable to get either GPU working with SVP. I wonder if vapoursynth is being run somehow without being under 'optirun' and thus it can't find the nvidia opencl device? I have no idea what could be wrong with the intel gpu.

Go to forum: Using SVP