chandlerc wrote: > You mention the performance tradeoff of pascal strings v null terminated ones > doesn't seem too important - I guess that's based on some judgement about > where/how these are used/where the strlens end up happening that you've > looked into? Could you summarize that in a bit more detail (like they only > happen during diagnostic processing, or something?)
I've not dug deeply, but the only code using it is updated in this PR. The biggest use is right before copying the string contents into a `std::string` (potentially including allocation), so the `strlen` seems definitively no problem there. The other one is a binary search for a string which is already somewhat unoptimized, but doesn't seem like a critical path of anything that needs to be optimized heavily. The pascal size there doesn't really help either as the search looks at the string content first for order. I'm moderately confident this isn't a problem, just mentioned it in case there was something really weird I missed that means I need to take the other approach. https://github.com/llvm/llvm-project/pull/123302 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
