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

test/compiler-unittests/compiler-unittests.gyp

Index: test/compiler-unittests/compiler-unittests.gyp
diff --git a/test/compiler-unittests/compiler-unittests.gyp b/test/compiler-unittests/compiler-unittests.gyp
index 13bc0c7c3aa63617dddbc136683662cbe5c8a484..e486fe32d6534a7848e7b9aeb650541525fe2bf2 100644
--- a/test/compiler-unittests/compiler-unittests.gyp
+++ b/test/compiler-unittests/compiler-unittests.gyp
@@ -28,6 +28,7 @@
         'graph-unittest.h',
         'instruction-selector-unittest.cc',
         'machine-operator-reducer-unittest.cc',
+        'machine-operator-unittest.cc',
       ],
       'conditions': [
         ['v8_target_arch=="arm"', {
@@ -45,6 +46,11 @@
             'ia32/instruction-selector-ia32-unittest.cc',
           ],
         }],
+        ['v8_target_arch=="x64"', {
+          'sources': [  ### gcmole(arch:x64) ###
+            'x64/instruction-selector-x64-unittest.cc',
+          ],
+        }],
         ['component=="shared_library"', {
           # compiler-unittests can't be built against a shared library, so we
           # need to depend on the underlying static target in that case.