[PATCH] D21672: [LICM] Avoid repeating expensive call while promoting loads. NFC

[Phabricator via llvm-commits]

This revision was automatically updated to reflect the committed changes.
Closed by commit rL273671: [LICM] Avoid repeating expensive call while promoting loads. NFC (authored by annat).

Changed prior to commit:
  http://reviews.llvm.org/D21672?vs=61751&id=61781#toc

Repository:
  rL LLVM

http://reviews.llvm.org/D21672

Files:
  llvm/trunk/lib/Transforms/Scalar/LICM.cpp

Index: llvm/trunk/lib/Transforms/Scalar/LICM.cpp
===================================================================
--- llvm/trunk/lib/Transforms/Scalar/LICM.cpp
+++ llvm/trunk/lib/Transforms/Scalar/LICM.cpp
@@ -945,15 +945,16 @@
         // instruction will be executed, update the alignment.
         // Larger is better, with the exception of 0 being the best alignment.
         unsigned InstAlignment = Store->getAlignment();
-        if ((InstAlignment > Alignment || InstAlignment == 0) && Alignment != 0)
+        if ((InstAlignment > Alignment || InstAlignment == 0) &&
+            Alignment != 0) {
           if (isGuaranteedToExecute(*UI, DT, CurLoop, SafetyInfo)) {
             GuaranteedToExecute = true;
             Alignment = InstAlignment;
           }
-
-        if (!GuaranteedToExecute)
+        } else if (!GuaranteedToExecute) {
           GuaranteedToExecute =
               isGuaranteedToExecute(*UI, DT, CurLoop, SafetyInfo);
+        }
 
         if (!GuaranteedToExecute && !CanSpeculateLoad) {
           CanSpeculateLoad = isDereferenceableAndAlignedPointer(


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D21672.61781.patch
Type: text/x-patch
Size: 1094 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160624/1fff73ed/attachment.bin>