JonChesterfield wrote: This seems legitimate to me. The IR pass is enabled on nvptx and turns LLVM variadics into functions which match the PTX variadic calling convention as far as I know. I _think_ the guard is still there on CUDA because Joseph and I were mostly thinking in terms of freestanding C++ when we wrote this.
https://github.com/llvm/llvm-project/pull/161350 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
