tra added a comment. In D74571#1876647 <https://reviews.llvm.org/D74571#1876647>, @jdoerfert wrote:
> That sounds like the right approach for OpenMP. We require a minimal CUDA > version, based on what we use internally, but no maximal version if possible > since we don't use new features anyway. SGTM, with a caveat. Word of caution -- it's not just the features. New CUDA versions also come with sometimes substantially different header files. Clang pre-includes a lot of them, so if the new headers have something funky, clang may need extra changes before it can compile anything, even if you don't use the new features. If you silence compiler with the option set during build config, it may be prudent to tell the user that CUDA compilation *may* fail if the CUDA version if not one of the known "not quite supported, but usable on par with the oldest supported version." CHANGES SINCE LAST ACTION https://reviews.llvm.org/D74571/new/ https://reviews.llvm.org/D74571 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
