[PATCH] Using Masked Load / Store intrinsics in Loop Vectorizer

[Michael Zolotukhin]

Some more minor notes from me.


================
Comment at: lib/Analysis/TargetTransformInfo.cpp:104-105
@@ -103,4 +103,4 @@
 
-bool TargetTransformInfo::isLegalPredicatedLoad(Type *DataType,
+bool TargetTransformInfo::isLegalMaskedLoad(Type *DataType,
                                                 int Consecutive) const {
   return false;
----------------
Please fix indentation here.

================
Comment at: lib/Analysis/TargetTransformInfo.cpp:109-110
@@ -108,4 +108,4 @@
 
-bool TargetTransformInfo::isLegalPredicatedStore(Type *DataType,
+bool TargetTransformInfo::isLegalMaskedStore(Type *DataType,
                                                  int Consecutive) const {
   return false;
----------------
And here.

================
Comment at: lib/Transforms/Vectorize/LoopVectorize.cpp:1914
@@ -1873,3 +1913,3 @@
       // If the address is consecutive but reversed, then the
-      // wide store needs to start at the last vector element.
+      // wide load needs to start at the last vector element.
       PartPtr = Builder.CreateGEP(Ptr, Builder.getInt32(-Part * VF));
----------------
This also might be committed separately.

http://reviews.llvm.org/D6527

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