AMD AOMP 11.5 Released For OpenMP Offloading To Radeon GPUs

Released on Wednesday was AOMP 11.5 as the latest version of the AMD/ROCm compiler based off LLVM Clang and focused on OpenMP offloading to Radeon GPUs.

AOMP is part of the ROCm umbrella for Radeon OpenMP device offloading. This is a branched version of upstream LLVM Clang that recently was re-based to LLVM 11 development code and thus the version number for AOMP.

AOMP 11.5 is tracking the state of LLVM 11 as of a few days ago and AMD ultimately is trying to work some of their changes upstream. Besides rebasing against upstream, AOMP 11.5 consolidates more code into their libomptarget HSA plug-in, improved test coverage, support for -g compilation of C/C++ code for target debugging, and HIPCC will now use the AOMP environment variable for pointing to compiler installations in non-default locations. Not a big release but good to see they are closely tracking LLVM 11 and ultimately aiming to upstream at least some of their changes.

More details on the AOMP 11.5 compiler release via the ROCm AOMP GitHub repository.