2 files changed, 39 insertions, 34 deletions

diff --git a/lib/Transforms/Scalar/LoopIdiomRecognize.cpp b/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
index be3ff9258f..d7fa149492 100644
--- a/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
+++ b/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
@@ -388,43 +388,24 @@ static Constant *getMemSetPatternValue(Value *V, const TargetData &TD) {
   if (Size == 0 || (Size & 7) || (Size & (Size-1)))
     return 0;
   
-  // Convert the constant to an integer type of the appropriate size so we can
-  // start hacking on it.
-  if (isa<PointerType>(V->getType()))
-    C = ConstantExpr::getPtrToInt(C, IntegerType::get(C->getContext(), Size));
-  else if (isa<VectorType>(V->getType()) || V->getType()->isFloatingPointTy())
-    C = ConstantExpr::getBitCast(C, IntegerType::get(C->getContext(), Size));
-  else if (!isa<IntegerType>(V->getType()))
-    return 0;  // Unhandled type.
+  // Don't care enough about darwin/ppc to implement this.
+  if (TD.isBigEndian())
+    return 0;
 
   // Convert to size in bytes.
   Size /= 8;
-  
-  // If we couldn't fold this to an integer, we fail.  We don't bother to handle
-  // relocatable expressions like the address of a global yet.
-  // FIXME!
-  ConstantInt *CI = dyn_cast<ConstantInt>(C);
-  if (CI == 0) return 0;
 
-  APInt CVal = CI->getValue();
-  
   // TODO: If CI is larger than 16-bytes, we can try slicing it in half to see
-  // if the top and bottom are the same.
+  // if the top and bottom are the same (e.g. for vectors and large integers).
   if (Size > 16) return 0;
-
-  // If this is a big endian target (PPC) then we need to bswap.
-  if (TD.isBigEndian())
-    CVal = CVal.byteSwap();
-  
-  // Determine what each byte of the pattern value should be.
-  char Value[16];
-  for (unsigned i = 0; i != 16; ++i) {
-    // Get the byte value we're indexing into.
-    unsigned CByte = i % Size;
-    Value[i] = (unsigned char)(CVal.getZExtValue() >> CByte);
-  }
   
-  return ConstantArray::get(V->getContext(), StringRef(Value, 16), false);
+  // If the constant is exactly 16 bytes, just use it.
+  if (Size == 16) return C;
+
+  // Otherwise, we'll use an array of the constants.
+  unsigned ArraySize = 16/Size;
+  ArrayType *AT = ArrayType::get(V->getType(), ArraySize);
+  return ConstantArray::get(AT, std::vector<Constant*>(ArraySize, C));
 }
 
 
@@ -518,8 +499,7 @@ processLoopStridedStore(Value *DestPtr, unsigned StoreSize,
                                             PatternValue, ".memset_pattern");
     GV->setUnnamedAddr(true); // Ok to merge these.
     GV->setAlignment(16);
-    Value *PatternPtr = Builder.CreateConstInBoundsGEP2_32(GV, 0, 0, "pattern");
-    
+    Value *PatternPtr = ConstantExpr::getBitCast(GV, Builder.getInt8PtrTy());
     NewCall = Builder.CreateCall3(MSP, BasePtr, PatternPtr, NumBytes);
   }
   
diff --git a/test/Transforms/LoopIdiom/basic.ll b/test/Transforms/LoopIdiom/basic.ll
index 122d25a22a..485114c8d4 100644
--- a/test/Transforms/LoopIdiom/basic.ll
+++ b/test/Transforms/LoopIdiom/basic.ll
@@ -277,7 +277,7 @@ for.end13:                                        ; preds = %for.inc10
 ; On darwin10 (which is the triple in this .ll file) this loop can be turned
 ; into a memset_pattern call.
 ; rdar://9009151
-define void @test11(i32* nocapture %P) nounwind ssp {
+define void @test11_pattern(i32* nocapture %P) nounwind ssp {
 entry:
   br label %for.body
 
@@ -291,7 +291,7 @@ for.body:                                         ; preds = %entry, %for.body
 
 for.end:                                          ; preds = %for.body
   ret void
-; CHECK: @test11
+; CHECK: @test11_pattern
 ; CHECK-NEXT: entry:
 ; CHECK-NEXT: bitcast
 ; CHECK-NEXT: memset_pattern
@@ -322,3 +322,28 @@ for.end:                                          ; preds = %for.body
 ; CHECK: ret void
 }
 
+@G = global i32 5
+
+; This store-of-address loop can be turned into a memset_pattern call.
+; rdar://9009151
+define void @test13_pattern(i32** nocapture %P) nounwind ssp {
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %entry, %for.body
+  %indvar = phi i64 [ 0, %entry ], [ %indvar.next, %for.body ]
+  %arrayidx = getelementptr i32** %P, i64 %indvar
+  store i32* @G, i32** %arrayidx, align 4
+  %indvar.next = add i64 %indvar, 1
+  %exitcond = icmp eq i64 %indvar.next, 10000
+  br i1 %exitcond, label %for.end, label %for.body
+
+for.end:                                          ; preds = %for.body
+  ret void
+; CHECK: @test13_pattern
+; CHECK-NEXT: entry:
+; CHECK-NEXT: bitcast
+; CHECK-NEXT: memset_pattern
+; CHECK-NOT: store
+; CHECK: ret void
+}