AOMP 20.0-2 was released on Monday as the newest update to this AMD downstream of the LLVM/Clang/Flang code that is focused on delivering the latest staging/testing patches around OpenMP offloading to AMD GPUs using ROCm. Many of AMD’s AMDGPU/OpenMP patches end up being upstreamed into LLVM proper while AOMP is the staging area for those wanting to have the latest and best experience for Clang C/C++ and Flang Fortran offloading to AMD Instinct/Radeon hardware.
With AOMP 20.0-2 they have re-based against a newer LLVM 20 development snapshot. Additionally, there is now included a build of the math ROCm libraries, other new ROCm library options included, and most notably a pre-production “flang-new” compiler binary:
– Added build of math rocmlibs (aomp-hip-libraries). Currently only support the following architectures: gfx900;gfx906:xnack-;gfx908:xnack-;gfx90a;gfx942;gfx1010;gfx1012;gfx1030;gfx1100;gfx1101;gfx1102;gfx1151;gfx1200;gfx1201
– Added optional aomp-hip-libraries package. This contains libraries for rocBLAS, rocPRIM, rocSPARSE, rocSOLVER, and hipBLAS.
– Added preproduction flang-new executable. Flang-classic is still default with a flang to flang-classic symbolic link.
– Moved to ROCm 6.3.2 sources for non-compiler related repositories.
Flang-New is LLVM’s newer Fortran compiler. Flang-New is being renamed back to Flang for replacing the former Fortran Flang compiler. The new Flang compiler is a nice improvement over the former/classic option while for now AOMP 20.0 is still defaulting to that classic compiler by default.
Downloads of the source code and various Linux binaries of AOMP 20.0-2 are available from ROCm AOMP on GitHub.
