erichkeane wrote: > > 1- How many instantiations these have > > -Thus: How much space taken, and saved by switch > > For the ones outside clang-tools-extra, none are in headers so they just > replace one instantiation each. > > 2- Which are hot-path enough to affect benchmarks, and which are irrelevant > > The conclusion we arrived at last time is that the performance is the same, > and the regressions are beacuse more vtables means more relocations; last > time I tested that by migrating only the unexpanded parameter packs visitor > and observed no difference in performance, so I would assume this is a > similar situation, but I’ll revert this change for some of the visitors in > codegen and see if this changes anything.
Yeah, sorry, I'm not caught up on this (and missed it last time through). Aaron is out another week or two, so it might be better to wait and see what he has to say/his thoughts here. https://github.com/llvm/llvm-project/pull/160290 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
