1; RUN: llc < %s -march=x86-64 -mcpu=corei7 -mattr=+sse4.1 | FileCheck %s
2
3; rdar://12721174
4; We should not fold movss into pshufd since pshufd expects m128 while movss
5; loads from m32.
6define void @sample_test(<4 x float>* %source, <2 x float>* %dest) nounwind {
7; CHECK: sample_test
8; CHECK-NOT: movaps
9; CHECK: insertps
10entry:
11  %source.addr = alloca <4 x float>*, align 8
12  %dest.addr = alloca <2 x float>*, align 8
13  %tmp = alloca <2 x float>, align 8
14  store <4 x float>* %source, <4 x float>** %source.addr, align 8
15  store <2 x float>* %dest, <2 x float>** %dest.addr, align 8
16  store <2 x float> zeroinitializer, <2 x float>* %tmp, align 8
17  %0 = load <4 x float>** %source.addr, align 8
18  %arrayidx = getelementptr inbounds <4 x float>* %0, i64 0
19  %1 = load <4 x float>* %arrayidx, align 16
20  %2 = extractelement <4 x float> %1, i32 0
21  %3 = load <2 x float>* %tmp, align 8
22  %4 = insertelement <2 x float> %3, float %2, i32 1
23  store <2 x float> %4, <2 x float>* %tmp, align 8
24  %5 = load <2 x float>* %tmp, align 8
25  %6 = load <2 x float>** %dest.addr, align 8
26  %arrayidx1 = getelementptr inbounds <2 x float>* %6, i64 0
27  store <2 x float> %5, <2 x float>* %arrayidx1, align 8
28  %7 = load <2 x float>** %dest.addr, align 8
29  %arrayidx2 = getelementptr inbounds <2 x float>* %7, i64 0
30  %8 = load <2 x float>* %arrayidx2, align 8
31  %vecext = extractelement <2 x float> %8, i32 0
32  %9 = load <2 x float>** %dest.addr, align 8
33  %arrayidx3 = getelementptr inbounds <2 x float>* %9, i64 0
34  %10 = load <2 x float>* %arrayidx3, align 8
35  %vecext4 = extractelement <2 x float> %10, i32 1
36  call void @ext(float %vecext, float %vecext4)
37  ret void
38}
39declare void @ext(float, float)
40