[llvm-commits] [llvm] r136250 - in /llvm/trunk: lib/Transforms/IPO/ConstantMerge.cpp test/Transforms/ConstantMerge/merge-both.ll

[Nick Lewycky]

Author: nicholas
Date: Wed Jul 27 14:47:34 2011
New Revision: 136250

URL: http://llvm.org/viewvc/llvm-project?rev=136250&view=rev
Log:
Teach the ConstantMerge pass about alignment. Fixes PR10514!

Modified:
    llvm/trunk/lib/Transforms/IPO/ConstantMerge.cpp
    llvm/trunk/test/Transforms/ConstantMerge/merge-both.ll

Modified: llvm/trunk/lib/Transforms/IPO/ConstantMerge.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/ConstantMerge.cpp?rev=136250&r1=136249&r2=136250&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/ConstantMerge.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/ConstantMerge.cpp Wed Jul 27 14:47:34 2011
@@ -23,7 +23,9 @@
 #include "llvm/DerivedTypes.h"
 #include "llvm/Module.h"
 #include "llvm/Pass.h"
+#include "llvm/Target/TargetData.h"
 #include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/PointerIntPair.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/Statistic.h"
 using namespace llvm;
@@ -37,10 +39,18 @@
       initializeConstantMergePass(*PassRegistry::getPassRegistry());
     }
 
-    // run - For this pass, process all of the globals in the module,
-    // eliminating duplicate constants.
-    //
+    // For this pass, process all of the globals in the module, eliminating
+    // duplicate constants.
     bool runOnModule(Module &M);
+
+    // Return true iff we can determine the alignment of this global variable.
+    bool hasKnownAlignment(GlobalVariable *GV) const;
+
+    // Return the alignment of the global, including converting the default
+    // alignment to a concrete value.
+    unsigned getAlignment(GlobalVariable *GV) const;
+
+    const TargetData *TD;
   };
 }
 
@@ -77,15 +87,28 @@
   return A.hasUnnamedAddr();
 }
 
+bool ConstantMerge::hasKnownAlignment(GlobalVariable *GV) const {
+  return TD || GV->getAlignment() != 0;
+}
+
+unsigned ConstantMerge::getAlignment(GlobalVariable *GV) const {
+  if (TD)
+    return TD->getPreferredAlignment(GV);
+  return GV->getAlignment();
+}
+
 bool ConstantMerge::runOnModule(Module &M) {
+  TD = getAnalysisIfAvailable<TargetData>();
+
   // Find all the globals that are marked "used".  These cannot be merged.
   SmallPtrSet<const GlobalValue*, 8> UsedGlobals;
   FindUsedValues(M.getGlobalVariable("llvm.used"), UsedGlobals);
   FindUsedValues(M.getGlobalVariable("llvm.compiler.used"), UsedGlobals);
   
-  // Map unique constant/section pairs to globals.  We don't want to merge
-  // globals in different sections.
-  DenseMap<Constant*, GlobalVariable*> CMap;
+  // Map unique <constants, has-unknown-alignment> pairs to globals.  We don't
+  // want to merge globals of unknown alignment with those of explicit
+  // alignment.  If we have TargetData, we always know the alignment.
+  DenseMap<PointerIntPair<Constant*, 1, bool>, GlobalVariable*> CMap;
 
   // Replacements - This vector contains a list of replacements to perform.
   SmallVector<std::pair<GlobalVariable*, GlobalVariable*>, 32> Replacements;
@@ -120,7 +143,8 @@
       Constant *Init = GV->getInitializer();
 
       // Check to see if the initializer is already known.
-      GlobalVariable *&Slot = CMap[Init];
+      PointerIntPair<Constant*, 1, bool> Pair(Init, hasKnownAlignment(GV));
+      GlobalVariable *&Slot = CMap[Pair];
 
       // If this is the first constant we find or if the old on is local,
       // replace with the current one. It the current is externally visible
@@ -152,7 +176,8 @@
       Constant *Init = GV->getInitializer();
 
       // Check to see if the initializer is already known.
-      GlobalVariable *Slot = CMap[Init];
+      PointerIntPair<Constant*, 1, bool> Pair(Init, hasKnownAlignment(GV));
+      GlobalVariable *Slot = CMap[Pair];
 
       if (!Slot || Slot == GV)
         continue;
@@ -175,6 +200,14 @@
     // now.  This avoid invalidating the pointers in CMap, which are unneeded
     // now.
     for (unsigned i = 0, e = Replacements.size(); i != e; ++i) {
+      // Bump the alignment if necessary.
+      if (Replacements[i].first->getAlignment() ||
+          Replacements[i].second->getAlignment()) {
+        Replacements[i].second->setAlignment(std::max(
+            Replacements[i].first->getAlignment(),
+            Replacements[i].second->getAlignment()));
+      }
+
       // Eliminate any uses of the dead global.
       Replacements[i].first->replaceAllUsesWith(Replacements[i].second);
 

Modified: llvm/trunk/test/Transforms/ConstantMerge/merge-both.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/ConstantMerge/merge-both.ll?rev=136250&r1=136249&r2=136250&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/ConstantMerge/merge-both.ll (original)
+++ llvm/trunk/test/Transforms/ConstantMerge/merge-both.ll Wed Jul 27 14:47:34 2011
@@ -1,5 +1,7 @@
 ; RUN: opt -constmerge %s -S -o - | FileCheck %s
 ; Test that in one run var3 is merged into var2 and var1 into var4.
+; Test that we merge @var5 and @var6 into one with the higher alignment, and
+; don't merge var7/var8 into var5/var6.
 
 declare void @zed(%struct.foobar*, %struct.foobar*)
 
@@ -14,13 +16,24 @@
 ; CHECK-NOT: @
 ; CHECK: @var2 = constant %struct.foobar { i32 2 }
 ; CHECK-NEXT: @var4 = constant %struct.foobar { i32 2 }
-; CHECK-NOT: @
-; CHECK: declare void @zed(%struct.foobar*, %struct.foobar*)
+
+declare void @helper([16 x i8]*)
+ at var5 = internal constant [16 x i8] c"foo1bar2foo3bar\00", align 16
+ at var6 = private unnamed_addr constant [16 x i8] c"foo1bar2foo3bar\00", align 1
+ at var7 = internal constant [16 x i8] c"foo1bar2foo3bar\00"
+ at var8 = private unnamed_addr constant [16 x i8] c"foo1bar2foo3bar\00"
+
+; CHECK-NEXT: @var6 = private constant [16 x i8] c"foo1bar2foo3bar\00", align 16
+; CHECK-NEXT: @var8 = private constant [16 x i8] c"foo1bar2foo3bar\00"
 
 define i32 @main() {
 entry:
   call void @zed(%struct.foobar* @var1, %struct.foobar* @var2)
   call void @zed(%struct.foobar* @var3, %struct.foobar* @var4)
+  call void @helper([16 x i8]* @var5)
+  call void @helper([16 x i8]* @var6)
+  call void @helper([16 x i8]* @var7)
+  call void @helper([16 x i8]* @var8)
   ret i32 0
 }