This week brought the release of AMD's ROCm 3.9 as their open-source compute stack for Radeon GPUs. With ROCm 3.9 the AOMP work for LLVM/Clang-based compiler with OpenMP offload capabilities to Radeon GPUs was integrated. AOMP though is still advancing independently of the ROCm releases with Friday night marking the release of AOMP 11.11.

AOMP 11.11 is the latest version of this LLVM Clang downstream that is carrying the Radeon OpenMP offloading support until all of it works its way into the upstream LLVM repository. AOMP 11.11 brings more consistency between the NVIDIA NVPTX and AMDGCN offloading and OpenMP / HIP / CUDA interfaces within the LLVM/Clang code-base. AOMP 11.11 also brings improvements to the Flang driver for Fortran OpenMP, run-time performance improvements for synchronous memory copies between host and device, various bug fixes, and other code improvements in preparation for ultimately upstreaming more of the code.

AOMP 11.11 can be downloaded from GitHub in source form as well as binaries for Red Hat Enterprise Linux / CentOS, SUSE LInux Enterprise Server, and Ubuntu LTS.