[PATCH] D92674: [TableGen] Cache two things to improve speed.

[Craig Topper via Phabricator via llvm-commits]

craig.topper added a comment.

Can you split this up? The namespace caching change seems quite a bit simpler and easy to approve. The other change makes me think we should return an ArrayRef instead of a vector from getAllDerivedDefinitions. Otherwise we're still making a copy on every call. I also wonder if it should use StringMap instead of std::unordered_map.

Which backends are requesting Instructions multiple times? Don't most backends go through getInstructionsByEnumValue that already caches?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D92674/new/

https://reviews.llvm.org/D92674