[PATCH] D69747: [IPConstantProp][NFCI] Improve and modernize tests

[Davide Italiano via Phabricator via llvm-commits]

davide accepted this revision.
davide added a comment.
This revision is now accepted and ready to land.

This patch looks good to me.

Now, on the more general discussion. There are some cases that are caught by IPCP [but not IPSCCP]. We can craft examples, if we want, but it's unclear how often they happen in practice. 
Having IPCP in tree doesn't seem to be a huge maintenance burden. OTOH, the algorithm is extremely naive and the pass isn't used anywhere, so if somebody wants to remove it, I wouldn't have any objections.

Also note that the cases that are caught by IPCP and not by IPSCCP are due to the fact that IPCP isn't really an analysis, and replaces stuffs as it goes. Having a context-sensitive IPSCCP pass could alleviate some of the differences, but it's unclear whether it's worth the complexity.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69747/new/

https://reviews.llvm.org/D69747