It appears the Tegra platform does not support libnvidia-ml.so and nvml.h? It would be incredibly helpful to have these available for the Jetson Nano where many CUDA applications that I work with depends on those.
The Jetson Nano is a really neat system on a board!
Logic would assume that if you had a raspberry pi supercomputer, then the next step
would be a one with gpu capacity which is yours.
These guys www.picocluster.com already sell it ; however, I don’t see HPC running on it. Just the standard hadoop, etc.
Lastly, you need to get python working with it for the same reasons.