[LV] Consistently apply address def scalarization across loop.

Consistently scalarize loads used as part of address computations across
all uses in the loop. This aligns the VPlan and legacy cost model and
fixes a divergence crash. It doesn't matter if the load and address
users are in different blocks, as long as they are in the same loop, the
scalar value can be used. This removes a number of insert/extracts.

1 files changed, 1 insertions, 1 deletions

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 3f16b03055a9..e62d57e6920b 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -5696,7 +5696,7 @@ void LoopVectorizationCostModel::setCostBasedWideningDecision(ElementCount VF) {
     Instruction *I = Worklist.pop_back_val();
     for (auto &Op : I->operands())
       if (auto *InstOp = dyn_cast<Instruction>(Op))
-        if ((InstOp->getParent() == I->getParent()) && !isa<PHINode>(InstOp) &&
+        if (TheLoop->contains(InstOp) && !isa<PHINode>(InstOp) &&
             AddrDefs.insert(InstOp).second)
           Worklist.push_back(InstOp);
   }