[PATCH] [AtomicExpand] Set branch weights when new basic blocks are created

[Philip Reames]

For the performance question, even a simple micro benchmark would be sufficient.  What is the difference in a contented vs uncontented access to a single word value with and without your change?  Is the difference material enough to worry about getting it wrong when the access is contented?


================
Comment at: include/llvm/Analysis/BranchProbabilityInfo.h:118
@@ -117,1 +117,3 @@
 
+  static MDNode *getBranchWeightsAtomicExpand(bool Reverse, LLVMContext &Ctx);
+
----------------
I know you're copying the StackProtector code here, but I'm not sure I see the need for this function to be exposed on BPI.  I'd just make it a static helper in the transform file.

http://reviews.llvm.org/D7804

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/