[PATCH] D66216: [ValueTracking] Fix recurrence detection to check both PHI operands.

[Florian Hahn via Phabricator via llvm-commits]

fhahn marked 3 inline comments as done.
fhahn added inline comments.


================
Comment at: llvm/test/Transforms/InstCombine/phi-known-bits-operand-order.ll:40
   %start = add nsw i32 %cell.0, 1
   br i1 %cond.v, label %for.cond11, label %for.cond26
 
----------------
lebedev.ri wrote:
> I think this should/can be unconditional branch to `for.cond11`?
> It branches on the same condition that brought us here.
> Does the fold still happen then?
If this branch is unconditional, we get the `ult` through some other code path (I did not track it down though). My guess is that we evaluate things in a different order.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D66216/new/

https://reviews.llvm.org/D66216