[PATCH] D111272: [InlineCost] model calls to llvm.is.constant* more carefully

[Nick Desaulniers via Phabricator via llvm-commits]

nickdesaulniers added a comment.

Thanks for the review; I will post a fix for your latest feedback by EOD.

Right now I'm playing with use-def chains; I suspect that if the operand to `@llvm.is.constant.*` is not `Constant`, we can build the list of `Arguments` from `F` that the operand depends on, then scan the use-def chains of the corresponding arguments in `F` to see if they would be `Constant` (ie. inlining would be beneficial; or more precisely at least we can evaluate `@llvm.is.constant.*` precisely).


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D111272/new/

https://reviews.llvm.org/D111272