[Openmp-dev] Nvidia GPU Offloading on ARM

[Lukas Sommer via Openmp-dev]

Hi everyone,

we have a Nvidia Jetson TX2 (Nvidia GPU + ARM64 CPU) as experimenting
platform. I wanted to test OpenMP 4.0/4.5 offloading with Clang on that
platform, unfortunately cmake is telling me that this is not supported:

> -- LIBOMPTARGET: Not building CUDA offloading plugin: only support
> CUDA in Linux x86_64 or ppc64le hosts.
What is the reason for ARM not being supported for offloading? Is there
any technical limitation that inhibits using the offloading tool-chains
for x86/Power in a similar fashion?

Is there some sort of experimental support for offloading on ARM that I
could use, e.g. by setting some flags or similar?

Thanks in advance,

Regards

Lukas