Nvidia HPC SDK on WSL2

Accelerated Computing CUDA CUDA on Windows Subsystem for Linux
1

I successfully managed to install the Windows11 CUDA driver and the CUDA toolkit for WSL2, i.e. I can successfuly run CUDA program from within WSL2 and also compile programs with nvcc and run them on the GPU.

I would like to develop OpenACC applications, for which I need the NVIDIA HPC SDK. I see that there is no official distribution for WSL2. I installed in WSL2 the Ubuntu package distributed by NVIDIA. I can compile OpenACC applications correctly and I see that loops gets parallelised for the target GPU. However, if I run the compiled program, it does not run on the GPU.

Are there succesful experiences of installing the NVIDIA HPC SDK on WSL2 and running the compiled applications on the GPU from within WSL2?

Thanks a lot for your help!

2

Yes, I installed HPC SDK without problems don’t install the driver of Linux.
It works but some limitations.
For OpenACC, works fine for multicore but not optimal for cuda. You have to tune
it. nvc++ works stdpar, autopar, manged memory (unified memory) openACC
openMP . I faced no problems for compiling the source code.