I am interesting to know whether it will be possible to extend the Mojo compiler to lower to Loom IR instead of LLVM IR → AMDGPU assembly.
The AMDGPU backend in upstream LLVM has historically had, let’s say, some issues.
AMD is developing (in their OSS repo here: hrx-system/loom at main · ROCm/hrx-system · GitHub) a new non-LLVM-based compiler to go from their custom Loom IR to AMDGPU assembly directly.
It would be great if it were possible to add a final MLIR pass to emit Loom IR. Is this something that will be possible to do once Mojo is open-sourced?