AMD Developing Next-Gen Fortran Compiler Based On Flang, Optimized For AMD GPUs

AMD today went public with details on the "AMD Next-Gen Fortran Compiler" as a new Fortran compiler they are working on based on LLVM's Flang.

The AMD Next-Gen Fortran Compiler is a new open-source Fortran compiler focused on OpenMP offloading to AMD GPUs with a direct interface to ROCm and HIP.

This downstream of Flang is focused on AMD GPU offloading and interacting with the ROCm compute stack. This downstream is open-source and tested across various Linux distributions while using the ROCm driver.

This is akin to AMD's AOMP compiler efforts as their LLVM downstream focused on providing the latest OpenMP offload support to AMD GPUs... But at least for now this "next-gen" compiler is being developed via a separate codebase from AOMP.

This also isn't to be confused with AMD's CPU compiler efforts... Over in AOCC land for the AMD Optimizing C/C++ Compiler they have already been shipping a Flang version for Fortran support catered to the AMD Zen-based processors. But with this Next-Gen Fortran Compiler it's on the GPU front. Yes, AMD's multiple downstreams of LLVM for different product families can get a bit confusing and overlapping in some areas.

AMD today published this ROCm blog post outlining their Next-Gen Fortran Compiler. For now at least the GitHub repository being worked on for this new Fortran compiler is under AMD's InfinityHub CI tree.