bradford 5 Posted April 10, 2019 Share Posted April 10, 2019 My issue is similar to the issue posted here: https://emby.media/community/index.php?/topic/71464-cannot-init-cuda-with-quadro-rtx-4000-on-ubuntu-1804/ The difference is I'm running it in a LXC container passing /dev/nvidia* through, as well as being on Debian 9.6. It's worked in the wild west days when I just had to use a ffmpeg build that supported nvenc - the link in my signature is my walkthrough of the process. To make sure the driver version isn't a culprit, I updated my driver on the host and in the container to 418.56 and nvidia-smi on both shows the card is detected with CUDA 10.1. Back in the day, nvidia-smi working was all it took to get ffmpeg and hardware transcoding working. hardware_detection.txt says "Cannot init CUDA", so I bet something about your CUDA detection isn't playing nice with container passthrough. I know this probably isn't supported, but I wonder if someone with insight into the hardware detection can give some insight (@softworkz?). Thanks for your help hardware_detection.txt embyserver.txt Link to comment Share on other sites More sharing options...
bradford 5 Posted April 12, 2019 Author Share Posted April 12, 2019 Any thoughts? Link to comment Share on other sites More sharing options...
Luke 36999 Posted April 12, 2019 Share Posted April 12, 2019 At the moment I'm not sure. We haven't tried doing this sort of thing yet. Link to comment Share on other sites More sharing options...
bradford 5 Posted April 12, 2019 Author Share Posted April 12, 2019 @@Luke can you give any insight into what your detect_hardware code does? If the source is available I can look at it myself. Is it just looking at /dev/ nodes? Or is it running a CUDA executable to test whether CUDA is available? Link to comment Share on other sites More sharing options...
Luke 36999 Posted April 12, 2019 Share Posted April 12, 2019 @@softworkz can answer that but it is essentially using the same methods that ffmpeg would use. Link to comment Share on other sites More sharing options...
softworkz 3325 Posted April 12, 2019 Share Posted April 12, 2019 @@Luke can you give any insight into what your detect_hardware code does? If the source is available I can look at it myself. Is it just looking at /dev/ nodes? Or is it running a CUDA executable to test whether CUDA is available? It is dynamically loading cuvid and nvenc libraries in the same way as ffmeg does. Link to comment Share on other sites More sharing options...
softworkz 3325 Posted April 12, 2019 Share Posted April 12, 2019 @@bradford - Are you saying that a recent 'official' ffmpeg build is working? Link to comment Share on other sites More sharing options...
bradford 5 Posted April 13, 2019 Author Share Posted April 13, 2019 @@bradford - Are you saying that a recent 'official' ffmpeg build is working? No, I haven't tried for a while. I'll see if I can get ffmpeg to work standalone and post back here with my results. Link to comment Share on other sites More sharing options...
bradford 5 Posted April 16, 2019 Author Share Posted April 16, 2019 So in order to use CUDA the /dev/nvidia-uvm device needs to be present, and isn't unless it's explicitly passed through from the host. In order to have the host load the kernel modules at boot, I made a few changes like those listed in this guide: https://medium.com/@MARatsimbazafy/journey-to-deep-learning-nvidia-gpu-passthrough-to-lxc-container-97d0bc474957 Works great now, I'll update my guide as well. Hope this helps someone else in the same boat. Do you guys do something similar in your docker container? Link to comment Share on other sites More sharing options...
Luke 36999 Posted April 17, 2019 Share Posted April 17, 2019 That's great info, thanks ! Someone else also posted a guide here: https://emby.media/community/index.php?/topic/72580-hardware-acceleration-emby-inside-lxcdebianubuntu-container-using-intel-igpu Link to comment Share on other sites More sharing options...
Luke 36999 Posted April 17, 2019 Share Posted April 17, 2019 And another here: https://emby.media/community/index.php?/topic/49680-howto-vaapi-transcoding-inside-lxc-container Link to comment Share on other sites More sharing options...
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now