jhuber6 wrote: Updated this to be simpler. This logic is *only* run when compiled from `clang --target=amdgcn-amd-amdhsa`, which means it will only show up for OpenCL and direct C++ targeting, HIP uses a different method to get the Toolchain. For regular C++ I do not want to depend on the ROCm tools, it should be freestanding compiler fare without the extra platform support.
This technically does have the effect that if someone does `clang -x cl foo.c -x c bar.c --target=amdgcn-amd-amdhsa` it will include the device libs for both, but I do not think that's an issue, since by using OpenCL you're already opting into it. Any chance of this landing before the merge? This primarily allows letting CMake use the AMDGPU as a freestanding target without EVERY single compile option check failing with "Could not find ROCm toolchain". https://github.com/llvm/llvm-project/pull/99687 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
