================ @@ -5,7 +5,7 @@ // RUN: %clang_cc1 -triple riscv64 -verify=expected,riscv64 -fsyntax-only %s // RUN: %clang_cc1 -triple x86_64 -verify=expected,x86_64 -fsyntax-only %s ---------------- Artem-B wrote:
nvptx does not *use* the code models, but CUDA compilation will *see* those attribute in the host code in the same TU. The main goal here is to make sure the GPU-side compilation will not be broken. I guess it may need a separate test under test/SemaCUDA/ where we have things set up to run a CUDA compilation. You could use RUN lines in ["test/SemaCUDA/function-target.cu"](https://github.com/llvm/llvm-project/blob/d3161defd6b717241a85b6ca528754c747060735/clang/test/SemaCUDA/function-target.cu#L1) as the base and the same content as you have in clang/test/CodeGen/X86/codemodel.cpp https://github.com/llvm/llvm-project/pull/124834 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
