Optimize loads and stores to Int32Array and Uint32Array.

runtime/vm/flow_graph_compiler.cc

Index: runtime/vm/flow_graph_compiler.cc
===================================================================
--- runtime/vm/flow_graph_compiler.cc	(revision 17337)
+++ runtime/vm/flow_graph_compiler.cc	(working copy)
@@ -890,6 +890,10 @@
       return Int16Array::kBytesPerElement;
     case kUint16ArrayCid:
       return Uint16Array::kBytesPerElement;
+    case kInt32ArrayCid:

[Kevin Millikin (Google), 2013/01/21 15:16:01]

There are several places with boilerplate.  Keep in mind a way to generate them
with macros or templates.

Another alternative is to have a separate enum for the Array subclasses, a
function to map cid to array class ID, and switch over that everywhere --- that
would give a compiler warning for nonexhaustiveness.

[Florian Schneider, 2013/01/21 16:03:25]

I'll keep that in mind for my next change where I want to add all loads/stores
with External* array variants. It would become even more boilerplate.

+      return Int32Array::kBytesPerElement;
+    case kUint32ArrayCid:
+      return Uint32Array::kBytesPerElement;
     case kOneByteStringCid:
       return OneByteString::kBytesPerElement;
     case kTwoByteStringCid:
@@ -922,6 +926,10 @@
       return Int16Array::data_offset();
     case kUint16ArrayCid:
       return Uint16Array::data_offset();
+    case kInt32ArrayCid:
+      return Int32Array::data_offset();
+    case kUint32ArrayCid:
+      return Uint32Array::data_offset();
     case kOneByteStringCid:
       return OneByteString::data_offset();
     case kTwoByteStringCid: