• AMD AOMP 23.0-0 Compiler Continues Enhancing Fortran Support AMD AOMP 23.0-0 was released overnight as the latest build of this LLVM/Clang downstream that continues to carry the very latest AMD patches focused on delivering the best support for GPU offloading to Radeon/Instinct hardware with the likes of the OpenMP and OpenACC APIs. • AOMP continues to serve as a great leading-edge compiler for the best AMD GPU offloading experience until the patches ultimately work their way into upstream LLVM. • AOMP 23.0-0 is re-based against the very latest LLVM/Clang/Flang 23 development code. • AOMP 23.0-0 also re-bases against the AMD ROCm 7.2 source code for the latest AMD capabilities there. • One fundamental change with AOMP 23.0-0 is rather than distributing Ubuntu / SUSE / RHEL / CentOS specific builds they are now providing a Tar file that was built on ManyLinux and aims to be the unified binary solution for using AOMP. • When it comes to the AOMP compiler improvements in this new release, it continues the recent theme of AOMP releases with much of the focus being around better Fortran programming language support with the Flang compiler front-end.
Article Summaries:
- AMD released AOMP 23.0‑0, the latest LLVM/Clang downstream compiler aimed at improving GPU offloading for Radeon and Instinct hardware. The new build is rebased on the current LLVM/Clang/Flang 23 development code and the ROCm 7.2 source, bringing the newest AMD capabilities. A significant change is the shift from platform‑specific binaries to a single ManyLinux‑built tarball, intended as a unified distribution. The release focuses on enhanced Fortran support via the Flang front‑end, adding numerous fixes and features. Full details, source code, and the new binary are available on the project’s GitHub repository.
Sources: