Before I buy half a dozen more Nano’s has anyone tested Julia with the GPU’s on the Nano’s? Or Xaviers?
Generally works fine. We don’t actively run CI for NVIDIA GPUs on ARM machines (anymore), but it’s supposed to be supported, so if there’s any issue let us know.
The only known issue is due to JIT code invalidation, Segfault during JIT memory invalidation · Issue #41852 · JuliaLang/julia · GitHub. It’s unclear what’s happening there, and it’s likely related to NVIDIA’s specific processors, so probably worth filing an issue with NVIDIA about if you run into it.
Linking this thread.