JonChesterfield added a comment. I like the current behaviour of bypass semantic checking and emit IR with the same number on it as my primary use case for feeding freestanding C++ to clang is as a convenient way to emit specific IR and I'm not that bothered about clang detecting mistakes along the way.
Funny that you mention addrspace(5) global as that's a pretty good approximation to what we'd like for errno. The semantics that would have been convenient is put it at the bottom of the stack in every kernel that can reach the variable and wire accesses to it through some base-of-stack built-in, which is presently not a thing but could be implemented. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D151087/new/ https://reviews.llvm.org/D151087 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits