Dear Developers community,
I am new on this forum and I hope my post is relevant to your community (apologies otherwise). I have recently bought an AMD RADEON PRO W6800 (RDNA architecture Navi GPUs) and I have installed this card on a DELL Precision 7920 Workstation operated by Ubuntu 20.04.4 LTS. My objective is to offload scientific calculations to this card from an OpenMP parallelized Fortran 95 code that I have written. To do so, my first attempt was to install Rocm v5.2 and try to compile my code using (Classic) FLANG frontend, but FLANG is based on pgfortran which does not support various functionalities supported by gfortran and critical to my code, so I had compilation errors that drove me to find an alternative solution. Unfortunately, I cannot use GNU Fortran (gfortran) either, because the LLVM backend supported by this compiler was designed for GCN architectures (Fiji, Vega) and not yet for RDNA. So I wonder if anyone could help me find a solution to this compiler issue so that I can offload calculations from my Fortran code to my RADEON PRO W6800 card.
Many thanks for your help.
