[VPlan] Limit stride replacement to vector region and middle VPBB (NFC).

At the moment this in NFC, but ensures we only replace uses that are
dominated by runtime checks as we model more of the skeleton in VPlan.

Author: fhahn

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

@@ -9189,6 +9189,14 @@ LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(VFRange &Range) {
 
   // Replace VPValues for known constant strides guaranteed by predicate scalar
   // evolution.
+  auto CanUseVersionedStride = [&Plan](VPUser &U, unsigned) {
+    auto *R = dyn_cast<VPRecipeBase>(&U);
+    if (!R)
+      return false;
+    return R->getParent()->getParent() ||
+           R->getParent() ==
+               Plan->getVectorLoopRegion()->getSinglePredecessor();
+  };
   for (auto [_, Stride] : Legal->getLAI()->getSymbolicStrides()) {
     auto *StrideV = cast<SCEVUnknown>(Stride)->getValue();
     auto *ScevStride = dyn_cast<SCEVConstant>(PSE.getSCEV(StrideV));
@@ -9199,7 +9207,7 @@ LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(VFRange &Range) {
     auto *CI = Plan->getOrAddLiveIn(
         ConstantInt::get(Stride->getType(), ScevStride->getAPInt()));
     if (VPValue *StrideVPV = Plan->getLiveIn(StrideV))
-      StrideVPV->replaceAllUsesWith(CI);
+      StrideVPV->replaceUsesWithIf(CI, CanUseVersionedStride);
 
     // The versioned value may not be used in the loop directly but through a
     // sext/zext. Add new live-ins in those cases.
@@ -9213,7 +9221,7 @@ LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(VFRange &Range) {
       APInt C = isa<SExtInst>(U) ? ScevStride->getAPInt().sext(BW)
                                  : ScevStride->getAPInt().zext(BW);
       VPValue *CI = Plan->getOrAddLiveIn(ConstantInt::get(U->getType(), C));
-      StrideVPV->replaceAllUsesWith(CI);
+      StrideVPV->replaceUsesWithIf(CI, CanUseVersionedStride);
     }
   }