Updated: Nov 23
In this blog post we will install NVIDIA Container Runtime package. NVIDIA Container Runtime package is basically runtime layer of the NVIDIA based containers which generally need GPU support. If you have an application which use GPU and you want to run this in container, you have to use NVIDIA runtime package.
How can we install NVIDIA Container Runtime? Answer is SDK Manager. While you are installing your forecr, you should install "Jetson SDK Components" too.
If you did everything properly, you could see outputs like below.
On upper step you could notice that Default Runtime selected as "nvidia". This helps when you build container which needs NVIDIA Runtime support. To configure this you should add a line inside Docker configuration file then restart docker service.
Also, you can test your setup with NVIDIA 's base container image and sample CUDA application. Open a terminal and give necessary permissions to Xserver.
Then run NVIDIA base container. Docker probably can't find image locally and start download. When it ends , you can see container terminal.
On container terminal update&upgrade container system and install necessary build packages.
Lastly, copy samples folder to /tmp then compile and run it.
You could see an application window that run some particle effects.
Thanks for reading.