[turbofan] Don't eliminate FinishRegion nodes.

The way we use FinishRegion for transitioning stores makes them eligible
for elimination by TypedOptimization, which is unintended and removes
the atomicity of the transitioning stores. This is a quickfix to ensure
that we don't remove the FinishRegion nodes during TypedOptimization;
the real fix is probably to have separate region operators for value
(producing) regions (i.e. allocations) and for effect-only regions (i.e.
transitioning stores).

R=jarin@chromium.org
BUG=v8:5303

Review-Url: https://codereview.chromium.org/2293023003
Cr-Commit-Position: refs/heads/master@{#39033}

src/compiler/typed-optimization.cc

@@ -33,6 +33,10 @@ Reduction TypedOptimization::Reduce(Node* node) {
   // result value and can simply replace the node if it's eliminable.
   if (!NodeProperties::IsConstant(node) && NodeProperties::IsTyped(node) &&
       node->op()->HasProperty(Operator::kEliminatable)) {
+    // TODO(v8:5303): We must not eliminate FinishRegion here. This special
+    // case can be removed once we have separate operators for value and
+    // effect regions.
+    if (node->opcode() == IrOpcode::kFinishRegion) return NoChange();
     // We can only constant-fold nodes here, that are known to not cause any
     // side-effect, may it be a JavaScript observable side-effect or a possible
     // eager deoptimization exit (i.e. {node} has an operator that doesn't have