[PATCH] Optimize sext/zext insertion algorithm in back-end

[James Molloy]

Hi Jiangning,

This looks good to me, with the two minor comments addressed.

Cheers,

James

================
Comment at: include/llvm/CodeGen/FunctionLoweringInfo.h:112
@@ +111,3 @@
+  /// for a value.
+  DenseMap<const Value *, ISD::NodeType> PerferredExtendType;
+
----------------
s/Perferred/Preferred/

================
Comment at: test/CodeGen/AArch64/atomic-ops.ll:512
@@ -511,3 +511,3 @@
 
-; CHECK: mov {{[xw]}}0, {{[xw]}}[[OLD]]
+; CHECK: mov {{[xw]}}0, {{[xw]}}[[OLD_EXT]]
    ret i8 %old
----------------
Why the changes to this file?

http://reviews.llvm.org/D5257