[PATCH] D103009: [DSE] Transform memset + malloc --> calloc (PR25892)

[Fangrui Song via Phabricator via llvm-commits]

MaskRay added a comment.

I agree with alexfh that this is questionable.
55 lines of complexity and compile time hurt in return of an "optimization naturally performed by humans, deoptimization/surprise at times, with other hidden costs"?

Other hidden costs: new references to msan/asan/hwasan exclusion. I think you missed memtag. When a new sanitizer is introduced, developers may go over the list of reference sites and adjust accordingly.

  we either lose test coverage or add some low-value tests.

Justification in real applications will help. I am not sure the benefits can be justified (we will need an additional cost: a post-dominance test and its accompanying test).


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D103009/new/

https://reviews.llvm.org/D103009