cor3ntin wrote: > As far as the changes go, I prefer the original code because it was more > clear what was going on. So unless this measurably improves performance, I'm > not certain it's a direction we should head (though @cor3ntin is the code > owner there, so maybe he feels differently).
I do agree with you - trying to improve performance of `CharInfo` is tantalizing because it directly impacts lexing performance. However, I was never able to convince myself that changes such as the one proposed here offer improvements that are not in the noise - I fully admit that this is in part because I'm not an expert in profiling. But I do expect the table to always stay in cache during translation (exactly because how frequently it's used), so intuitively i would expect all of these things to be somewhat equivalent. But yeah maybe @nikic or other backend people would have a better feel for what is the most efficient pattern here. https://github.com/llvm/llvm-project/pull/171052 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
