Migrating CUDA capable container from JP4.6 to JP6.2

Accelerated Computing CUDA CUDA Programming and Performance
, , ,
1

Hello all,

I am trying to migrate CUDA capable container I built on JP4.6 (CUDA-10) on Jetson Nano, into Jetson Orin Nano with JP6.2 (CUDA-12).
After initial deployment, I see that some CUDA files from /usr/local/ are missing, and in general, the directory on Nano was about 3 GB, after moving to Orin it weights ~100MB. I suppose that some of the CUDA files were simply copied in the container build process.
What would be the easiest way to perform the migration? Somehow hard-copy the CUDA-10 files into container?
Or maybe migrate the application to CUDA12 (I’d probably need to re-build the container)?

2

Hi @bkawlatow,

I have a question to better understand the context of the problem, what docker image are you basing your container on?

Regards,
Francis Guindon

Embedded SW Engineer at RidgeRun
Contact us: support@ridgerun.com
Developers wiki: https://developer.ridgerun.com/
Website: http://www.ridgerun.com/

3

I had a gesture recognition app working on the Jetson Nano, and I was trying to migrate it to Jetson Orin.
It was based on one of the NVIDIA docker image. Eventually I realized that since the CUDA version is different, I had to rebuild all of the hardware dependent libraries.
It was just easier to build the new image from scratch.

1 Like