[llvm-branch-commits] [llvm] [LV] Vectorize down-counting floating-point argmin/argmax reductions (PR #209827)

[Florian Hahn via llvm-branch-commits]

================
@@ -1626,6 +1626,36 @@ static VPInstruction *findFindIVSelect(VPValue *BackedgeVal) {
       }));
 }
 
+/// Recover the wide induction underlying the value \p IVOp that a FindIV select
+/// stores. \p IVOp is either the induction itself (then \p Offset is left null)
----------------
fhahn wrote:

I think the use of stored here is a bit confusing, there are no stores of the loop.

https://github.com/llvm/llvm-project/pull/209827