[turbofan] Add Int32AddWithOverflow machine operator.

src/compiler/machine-node-factory.h

Index: src/compiler/machine-node-factory.h
diff --git a/src/compiler/machine-node-factory.h b/src/compiler/machine-node-factory.h
index b0e153181feb8d143c2e6341b54dc6319c7c54f1..f3815d15fdcdab12aca3b531795d6c1980bbbe03 100644
--- a/src/compiler/machine-node-factory.h
+++ b/src/compiler/machine-node-factory.h
@@ -85,6 +85,11 @@ class MachineNodeFactory {
     return NEW_NODE_0(COMMON()->HeapConstant(val));
   }
 
+  // Projections.
+  Node* Projection(int index, Node* tuple) {
+    return NEW_NODE_1(COMMON()->Projection(index), tuple);
+  }
+
   // Memory Operations.
   Node* Load(MachineRepresentation rep, Node* base) {
     return Load(rep, base, Int32Constant(0));
@@ -195,6 +200,9 @@ class MachineNodeFactory {
   Node* Int32Add(Node* a, Node* b) {
     return NEW_NODE_2(MACHINE()->Int32Add(), a, b);
   }
+  Node* Int32AddWithOverflow(Node* a, Node* b) {

[titzer, 2014/07/31 15:16:34]

Maybe we want to return the two projection nodes from this (i.e. accept
{Node**add} {Node** ovf})? Otherwise it's easy to screw it up and use the add
directly, which is a violation.

[Benedikt Meurer, 2014/07/31 15:18:19]

I'd leave it this way for consistency with calls.

[Benedikt Meurer, 2014/08/01 05:01:18]

Done.

+    return NEW_NODE_2(MACHINE()->Int32AddWithOverflow(), a, b);
+  }
   Node* Int32Sub(Node* a, Node* b) {
     return NEW_NODE_2(MACHINE()->Int32Sub(), a, b);
   }