[turbofan] Add proper conversion operators for int32<->int64.

src/compiler/x64/instruction-selector-x64.cc

Index: src/compiler/x64/instruction-selector-x64.cc
diff --git a/src/compiler/x64/instruction-selector-x64.cc b/src/compiler/x64/instruction-selector-x64.cc
index 55c35da5027be85b7510e5f9c6723bba057cb157..bb4c03bd7e96c65bca7336d070b095e759fbcccb 100644
--- a/src/compiler/x64/instruction-selector-x64.cc
+++ b/src/compiler/x64/instruction-selector-x64.cc
@@ -513,6 +513,24 @@ void InstructionSelector::VisitChangeFloat64ToUint32(Node* node) {
 }
 
 
+void InstructionSelector::VisitChangeInt32ToInt64(Node* node) {
+  X64OperandGenerator g(this);
+  Emit(kX64Movsxlq, g.DefineAsRegister(node), g.Use(node->InputAt(0)));
+}
+
+
+void InstructionSelector::VisitChangeUint32ToUint64(Node* node) {
+  X64OperandGenerator g(this);
+  Emit(kX64Movl, g.DefineAsRegister(node), g.Use(node->InputAt(0)));
+}
+
+
+void InstructionSelector::VisitTruncateInt64ToInt32(Node* node) {
+  X64OperandGenerator g(this);
+  Emit(kX64Movl, g.DefineAsRegister(node), g.Use(node->InputAt(0)));
+}
+
+
 void InstructionSelector::VisitFloat64Add(Node* node) {
   X64OperandGenerator g(this);
   Emit(kSSEFloat64Add, g.DefineSameAsFirst(node),
@@ -554,22 +572,6 @@ void InstructionSelector::VisitFloat64Mod(Node* node) {
 }
 
 
-void InstructionSelector::VisitConvertInt64ToInt32(Node* node) {
-  X64OperandGenerator g(this);
-  // TODO(dcarney): other modes
-  Emit(kX64Int64ToInt32, g.DefineAsRegister(node),
-       g.UseRegister(node->InputAt(0)));
-}
-
-
-void InstructionSelector::VisitConvertInt32ToInt64(Node* node) {
-  X64OperandGenerator g(this);
-  // TODO(dcarney): other modes
-  Emit(kX64Int32ToInt64, g.DefineAsRegister(node),
-       g.UseRegister(node->InputAt(0)));
-}
-
-
 void InstructionSelector::VisitInt32AddWithOverflow(Node* node,
                                                     FlagsContinuation* cont) {
   VisitBinop(this, node, kX64Add32, cont);