AMD Radeon GPU Offloading For GCC Still Maturing In 2020

[phoronix]

Phoronix: AMD Radeon GPU Offloading For GCC Still Maturing In 2020

When it comes to Radeon GPU offloading for compute, most of the emphasis placed by AMD has been on Radeon Open Compute (ROCm) and now as part of that as well is the AOMP compiler for OpenMP targeting the AMD GPUs. Both ROCm and AOMP along with their other graphics driver components like RadeonSI and AMDVLK target an LLVM-based compiler stack as their principal focus, but they do continue engaging with Mentor Graphics who leads the development work on a Radeon GPU back-end for GCC...

http://www.phoronix.com/scan.php?page=news_item&px=Radeon-GCC-2020-Status

[FireBurn]

Wasn't there an article recently saying it was being removed, or am I misremembering

[bridgman]

Wasn't there an article recently saying it was being removed, or am I misremembering

I think that was just the HSAIL back end.

The ROCm stack has switched to generating native ISA rather than generating an intermediate IR (HSAIL) and JIT-compiling the HSAIL to native ISA at runtime via the "finalizer".

[torsionbar28]

Personally I'm waiting for GCC offloading support for the Matrox G200 card.

[oleid]

Thus GCC 10 is more usable now with AMD Radeon GPUs, but even with this current version of the GCC compiler only Fiji and Vega 10/20 are supported -- not the newer Navi or older GCN GPUs yet.

Is this because GCC's GCN-backend doesn't generate code compatible with those GPUs or is it because ROCm doesn't support those, yet?