I had the chance to give a presentation at the Easybuild User Meeting 2023 about the OpenMP Offloading capability in AMD’s ROCm compiler. You can find the recording of that talk on youtube.
OpenMP Offloading is the OpenMP model to use the GPU to accelerate computations. This is done via pragmas that use the target keyword. One of the nice things about this is that it does not require the programmer to write GPU-kernel code by hand but have instead the compiler do the work. This can be good for initial experimentation and incremental porting.
No responses yet