Sirraide wrote: > My suspicion is thus that this actually marginally regresses binary loading > time.
Huh, that’s interesting; I didn’t even consider that as a possibility, but that would definitely explain why reverting some of the most commonly used visitors back to using CRTP didn’t do anything at all (except increase Clang’s binary size and make compiling it slower again...). > In terms of landing this, I'd split this up in at least three parts, which is > 1. the implementation (which is the part that will need detailed review), 2. > converting unit tests (that should be a no-brainer) and 3. migrating > production visitors, possibly further split down (e.g. something like > ARCMigrate is self-contained). Yeah, that stgm. This pr was mostly just to show ‘this is what it’s like right now’. I’ll start by splitting the `DynamicRecursiveASTVisitor` implementation into a separate pr. https://github.com/llvm/llvm-project/pull/105195 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
