snarkmaster wrote: @NewSigma, thanks! Two things: - I'm moving some code around, but the `await_suspend` path is 100% unchanged. I effectively just added a branch where a bunch of that stuff isn't needed. - I quickly ran my `folly::result` benchmark from the description under clang-17 rather than clang-19, and ... its performance is even worse on 17 than 19. So, I don't think that issue will help here. The fundamental problem I'm addressing isn't lack of devirtualization, but lack of heap alloc elision.
https://github.com/llvm/llvm-project/pull/152623 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
