[Bug target/105602] New: [OpenMP][gcn] — Support multiple arch in gcc/config/gcn/t-omp-device? Add 'amdgcn' (additionally to/instead of 'amd')

[burnus at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105602

            Bug ID: 105602
           Summary: [OpenMP][gcn] — Support multiple arch in
                    gcc/config/gcn/t-omp-device? Add 'amdgcn'
                    (additionally to/instead of 'amd')
           Product: gcc
           Version: 13.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: burnus at gcc dot gnu.org
                CC: ams at gcc dot gnu.org, jakub at gcc dot gnu.org
  Target Milestone: ---

I note that GCC has in gcc/config/gcn/t-omp-device:
        echo kind: gpu > $@
        echo arch: gcn >> $@
        echo isa: fiji gfx900 gfx906 gfx908 >> $@

That is, GCC supports as context selector:
   match(construct={target},device={arch(gcn)})

In contrast, llvm has, e.g. in clang/lib/Headers/openmp_wrappers/cmath:
   #pragma omp begin declare variant match(device = {arch(amdgcn)})

Thus: LLVM uses 'amdgcn'.

Expected: arch = 'amdgcn' (also) works in GCC for better compatibility. I did
not find a documentation and I think either variant is fine – albeit 'amdgcn'
seems to be more natural.