Avoid fusing bitcasts with dynamic allocas if the amount-to-allocate

might overflow.  Re-typing the alloca to a larger type (e.g. double)
hoists a shift into the alloca, potentially exposing overflow in the
expression.  rdar://problem/9265821

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132926 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/InstCombine/InstCombineCasts.cpp b/lib/Transforms/InstCombine/InstCombineCasts.cpp
index 199902aa41f8431e93655724cbfdb07d57a493b9..601d9b42f31454c4157c0344a75b6342bde41730 100644 (file)
--- a/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -71,6 +71,11 @@ Instruction *InstCombiner::PromoteCastOfAllocation(BitCastInst &CI,
   // This requires TargetData to get the alloca alignment and size information.
   if (!TD) return 0;
 
+  // Insist that the amount-to-allocate not overflow.
+  OverflowingBinaryOperator *OBI =
+    dyn_cast<OverflowingBinaryOperator>(AI.getOperand(0));
+  if (OBI && !(OBI->hasNoSignedWrap() || OBI->hasNoUnsignedWrap())) return 0;
+
   const PointerType *PTy = cast<PointerType>(CI.getType());
   
   BuilderTy AllocaBuilder(*Builder);

diff --git a/test/Transforms/InstCombine/2011-06-13-nsw-alloca.ll b/test/Transforms/InstCombine/2011-06-13-nsw-alloca.ll
new file mode 100644 (file)
index 0000000..2f72b73
--- /dev/null
+++ b/test/Transforms/InstCombine/2011-06-13-nsw-alloca.ll
@@ -0,0 +1,60 @@
+; RUN: opt -S -instcombine < %s | FileCheck %s
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32"
+target triple = "i386-apple-darwin10.0.0"
+
+define void @fu1(i32 %parm) nounwind ssp {
+  %1 = alloca i32, align 4
+  %ptr = alloca double*, align 4
+  store i32 %parm, i32* %1, align 4
+  store double* null, double** %ptr, align 4
+  %2 = load i32* %1, align 4
+  %3 = icmp ne i32 %2, 0
+  br i1 %3, label %4, label %10
+
+; <label>:4                                       ; preds = %0
+  %5 = load i32* %1, align 4
+  %6 = mul nsw i32 %5, 8
+; With "nsw", the alloca and its bitcast can be fused:
+  %7 = add nsw i32 %6, 2048
+; CHECK: alloca double*
+  %8 = alloca i8, i32 %7
+  %9 = bitcast i8* %8 to double*
+  store double* %9, double** %ptr, align 4
+  br label %10
+
+; <label>:10                                      ; preds = %4, %0
+  %11 = load double** %ptr, align 4
+  call void @bar(double* %11)
+; CHECK: ret
+  ret void
+}
+
+declare void @bar(double*)
+
+define void @fu2(i32 %parm) nounwind ssp {
+  %1 = alloca i32, align 4
+  %ptr = alloca double*, align 4
+  store i32 %parm, i32* %1, align 4
+  store double* null, double** %ptr, align 4
+  %2 = load i32* %1, align 4
+  %3 = icmp ne i32 %2, 0
+  br i1 %3, label %4, label %10
+
+; <label>:4                                       ; preds = %0
+  %5 = load i32* %1, align 4
+  %6 = mul nsw i32 %5, 8
+; Without "nsw", the alloca and its bitcast cannot be fused:
+  %7 = add  i32 %6, 2048
+; CHECK: alloca i8
+  %8 = alloca i8, i32 %7
+; CHECK-NEXT: bitcast i8*
+  %9 = bitcast i8* %8 to double*
+  store double* %9, double** %ptr, align 4
+  br label %10
+
+; <label>:10                                      ; preds = %4, %0
+  %11 = load double** %ptr, align 4
+  call void @bar(double* %11)
+  ret void
+}
+