BBVectorize: Fix a small bug introduced in r167042.

We need to make sure that we take the correct load/store alignment
when the inputs are flipped.

llvm-svn: 167044

2 files changed, 28 insertions, 1 deletions

diff --git a/llvm/lib/Transforms/Vectorize/BBVectorize.cpp b/llvm/lib/Transforms/Vectorize/BBVectorize.cpp
index b29a45e9e81..32a18f21085 100644
--- a/llvm/lib/Transforms/Vectorize/BBVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/BBVectorize.cpp
@@ -2472,7 +2472,6 @@ namespace {
       Instruction *L = I, *H = J;
       if (FlipMemInputs)
         std::swap(H, L);
-      FlipMemInputs = false;
 
       unsigned NumOperands = I->getNumOperands();
       SmallVector<Value *, 3> ReplacedOperands(NumOperands);
diff --git a/llvm/test/Transforms/BBVectorize/simple-ldstr.ll b/llvm/test/Transforms/BBVectorize/simple-ldstr.ll
index c1e6a09eea2..6883e844bef 100644
--- a/llvm/test/Transforms/BBVectorize/simple-ldstr.ll
+++ b/llvm/test/Transforms/BBVectorize/simple-ldstr.ll
@@ -140,3 +140,31 @@ if.end:
 ; CHECK-AO-NOT: <2 x double>
 }
 
+; Simple 3-pair chain with loads and stores
+define void @test5(double* %a, double* %b, double* %c) nounwind uwtable readonly {
+entry:
+  %i0 = load double* %a, align 8
+  %i1 = load double* %b, align 8
+  %mul = fmul double %i0, %i1
+  %arrayidx3 = getelementptr inbounds double* %a, i64 1
+  %i3 = load double* %arrayidx3, align 8
+  %arrayidx4 = getelementptr inbounds double* %b, i64 1
+  %i4 = load double* %arrayidx4, align 8
+  %mul5 = fmul double %i3, %i4
+  %arrayidx5 = getelementptr inbounds double* %c, i64 1
+  store double %mul5, double* %arrayidx5, align 8
+  store double %mul, double* %c, align 4
+  ret void
+; CHECK: @test5
+; CHECK: %i0.v.i0 = bitcast double* %a to <2 x double>*
+; CHECK: %i1.v.i0 = bitcast double* %b to <2 x double>*
+; CHECK: %i0 = load <2 x double>* %i0.v.i0, align 8
+; CHECK: %i1 = load <2 x double>* %i1.v.i0, align 8
+; CHECK: %mul = fmul <2 x double> %i0, %i1
+; CHECK: %0 = bitcast double* %c to <2 x double>*
+; CHECK: store <2 x double> %mul, <2 x double>* %0, align 4
+; CHECK: ret void
+; CHECK-AO: @test5
+; CHECK-AO-NOT: <2 x double>
+}
+