--- /dev/null
+; RUN: llvm-as < %s | llc -march=x86 -mattr=sse | grep movaps
+; Test that the load is NOT folded into the intrinsic, which would zero the top
+; elts of the loaded vector.
+
+target endian = little
+target pointersize = 32
+target triple = "i686-apple-darwin8.7.2"
+
+implementation ; Functions:
+
+<4 x float> %test(<4 x float> %A, <4 x float>* %B) {
+ %BV = load <4 x float>* %B
+ %tmp28 = tail call <4 x float> %llvm.x86.sse.sub.ss( <4 x float> %A, <4 x float> %BV)
+ ret <4 x float> %tmp28
+}
+
+declare <4 x float> %llvm.x86.sse.sub.ss(<4 x float>, <4 x float>)