[PATCH] D95288: [ValueTracking] Don't assume readonly function will return

[Nikita Popov via Phabricator via llvm-commits]

nikic added inline comments.


================
Comment at: lib/Analysis/ValueTracking.cpp:5059
-    // is guaranteed to return.
-    return CB->onlyReadsMemory() || CB->onlyAccessesArgMemory();
   }
----------------
fhahn wrote:
> I guess we could incorporate the `argmemonly` reasoning in 'FunctionAttrs`, using something like `argmemonly + nosync => willreturn`, assuming that IO can only happen through inaccessible memory? Just `argmemonly` would not be enough I think, because we still need to rule out volatile accesses/synchronization via. the arguments. 
Yeah, I agree that mustprogress, argmemonly and nosync should imply willreturn. However, FuncAttrs currently infers neither nosync nor argmemonly, so it would need those first. I think that the argmemonly check here was there to cover some common intrinsics like `llvm.memcpy`, but those are explicitly willreturn now.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D95288/new/

https://reviews.llvm.org/D95288