[PATCH] D127443: [InstCombine] add zeroinitializer handling to memchr folders

[Martin Sebor via Phabricator via llvm-commits]

msebor added a comment.

Neither API is ideal: both require the client to handle the `zeroinitializer` specially to some extent.  The difference is that one approach is optimally efficient at the cost of more code, while the other allows for more uniform handling at the cost of inefficiency (memory allocation and linear traversal over all zeros).  Both are trade-offs, and although I don't feel strongly enough to insist on it, in D125114 <https://reviews.llvm.org/D125114> we discussed a similar concern and agreed to work toward a more efficient solution there in the future.  I have a similar preference for this change.  If it turns out that the uniform handling is more important than optimal efficiency the API can easily be changed.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D127443/new/

https://reviews.llvm.org/D127443