jhuber6 wrote: > It's how it works today, I believe: `--offload-arch` unambiguously > establishes the toolchain
It's pretty ambiguous, right now it mostly works through a combination of the file type and good guessing because the targets people care about now have distinct names. For OpenMP, you can offload to x64 CPUs or whatever and then it doesn't work via `--offload-arch` because there's not a distinct CPU name we can infer the triple based on. For HIP we just infer the triple that it's AMD, but this SPIR-V thing throws the same wrench that OpenMP has. Maybe in the future someone will want `clang++ -x cuda foo.cpp --offload-arch=gfx1030` as well, who knows. I don't think this is a blocker, but I think it bears deep consideration on how we expect these toolchains to actually work. https://github.com/llvm/llvm-project/pull/113509 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
