PR9561: A store with a negative offset (via GEP) could erroniously say that it

completely overlaps a previous store, thus mistakenly deleting that store. Check
for this condition.

llvm-svn: 128319

2 files changed, 24 insertions, 0 deletions

diff --git a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
index 867a06ad202..d07cf0727a8 100644
--- a/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
+++ b/llvm/lib/Transforms/Scalar/DeadStoreElimination.cpp
@@ -354,8 +354,10 @@ static bool isCompleteOverwrite(const AliasAnalysis::Location &Later,
   // In this case, we see if the later store completely overlaps all bytes
   // stored by the previous store.
   if (Off1 < Off2 ||                       // Earlier starts before Later.
+      Off2 < 0 ||                          // Later is -.
       Off1+Earlier.Size > Off2+Later.Size) // Earlier goes beyond Later.
     return false;
+
   // Otherwise, we have complete overlap.
   return true;
 }
diff --git a/llvm/test/Transforms/DeadStoreElimination/2011-03-25-DSEMiscompile.ll b/llvm/test/Transforms/DeadStoreElimination/2011-03-25-DSEMiscompile.ll
new file mode 100644
index 00000000000..e268dd54140
--- /dev/null
+++ b/llvm/test/Transforms/DeadStoreElimination/2011-03-25-DSEMiscompile.ll
@@ -0,0 +1,22 @@
+; RUN: opt < %s -basicaa -dse -S | FileCheck %s
+; PR9561
+target triple = "i386-apple-darwin9.8"
+
+@A = external global [0 x i32]
+
+declare cc10 void @Func2(i32*, i32*, i32*, i32)
+
+define cc10 void @Func1(i32* noalias %Arg1, i32* noalias %Arg2, i32* %Arg3, i32 %Arg4) {
+entry:
+  store i32 add (i32 ptrtoint ([0 x i32]* @A to i32), i32 1), i32* %Arg2
+; CHECK: store i32 add (i32 ptrtoint ([0 x i32]* @A to i32), i32 1), i32* %Arg2
+  %ln2gz = getelementptr i32* %Arg1, i32 14
+  %ln2gA = bitcast i32* %ln2gz to double*
+  %ln2gB = load double* %ln2gA
+  %ln2gD = getelementptr i32* %Arg2, i32 -3
+  %ln2gE = bitcast i32* %ln2gD to double*
+  store double %ln2gB, double* %ln2gE
+; CHECK: store double %ln2gB, double* %ln2gE
+  tail call cc10 void @Func2(i32* %Arg1, i32* %Arg2, i32* %Arg3, i32 %Arg4) nounwind
+  ret void
+}