rgreenblatt added a comment. In D94337#2496329 <https://reviews.llvm.org/D94337#2496329>, @tra wrote:
> We were talking about the case where clang can't expand CUDA-specific macros. Somehow I totally forgot about or missed this context. > I think this kind of compilation pipeline restructuring is a bit premature. Yeah for sure. I was trying to make `clang++ file.cuh ...` do something same - this isn't at all needed for tooling. Also I was mostly doing this to improve my understanding of what is going on in Driver.cpp wrt. cuda compilation. Sorry about missing or misunderstanding a lot of what you're saying and saying obvious things. >> However, the fact that the preprocessor includes depend on arch >> means that the __clang_cuda_standalone_defs.h approach won't always be >> perfectly >> correct. > > `...` The goal of `__clang_cuda_standalone_defs.h` is to make it possible to > parse CUDA sources at all w/o having to rely on CUDA SDK. `...` Should `__clang_cuda_standalone_defs.h` depend on the arch? For example, `__match_all_sync` doesn't exist in sm_35 but does exist in sm_75. For tooling this might not matter because the host cuda includes don't depend on the arch. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D94337/new/ https://reviews.llvm.org/D94337 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
