Radeon GCC Compute Back-End Approved For Merging In The Upcoming GCC9 Compiler

As a follow-up to the story a few days ago about interest in getting the Radeon GCC compiler back-end merged for this year's GNU Compiler Collection 9 (GCC9) release, it's been approved!

This is the back-end for the GCC computer targeting recent generations of AMD's Radeon GCN architecture. Ultimately it's for allowing OpenMP and OpenACC code, among other possible use-cases, to be offloaded to Radeon GPUs for compute. GCC has long had the NVIDIA NVPTX back-end as well as support for Xeon Phi and friends, but not this proper AMD GCN back-end that has been in development for a while. Granted, there is the AMDGPU back-end within LLVM that is used by the open-source OpenGL and Vulkan drivers, among other use-cases.

While GCC9 is under its documentation and regression fixing mode, the new code will still end up landing as it doesn't risk regressing existing targets. Red Hat's Jeff Law who is also a member of the GCC steering committee green lit this Radeon GPU code for merging:

And I think the V3 patch is reasonable enough to go in now. There's some concerns that have been raised with the implementation, but I'm comfortable with Andrew faulting in fixes if those concerns turn into real issues.

Andrew, you're green-lighted for the trunk.

GCC 9.1 is due for release around April and should now feature this preliminary Radeon GPU back-end.