1; RUN: llc < %s -O0
2target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:32-f32:32:32-f64:32:32-v64:64:64-v128:128:128-a0:0:64-n32"
3target triple = "armv7-apple-darwin"
4
5define i32 @main() nounwind ssp {
6entry:
7  %retval = alloca i32, align 4
8  %X = alloca <4 x i32>, align 16
9  %Y = alloca <4 x float>, align 16
10  store i32 0, i32* %retval
11  %tmp = load <4 x i32>* %X, align 16
12  call void @__aa(<4 x i32> %tmp, i8* null, i32 3, <4 x float>* %Y)
13  %0 = load i32* %retval
14  ret i32 %0
15}
16
17define internal void @__aa(<4 x i32> %v, i8* %p, i32 %offset, <4 x float>* %constants) nounwind inlinehint ssp {
18entry:
19  %__a.addr.i = alloca <4 x i32>, align 16
20  %v.addr = alloca <4 x i32>, align 16
21  %p.addr = alloca i8*, align 4
22  %offset.addr = alloca i32, align 4
23  %constants.addr = alloca <4 x float>*, align 4
24  store <4 x i32> %v, <4 x i32>* %v.addr, align 16
25  store i8* %p, i8** %p.addr, align 4
26  store i32 %offset, i32* %offset.addr, align 4
27  store <4 x float>* %constants, <4 x float>** %constants.addr, align 4
28  %tmp = load <4 x i32>* %v.addr, align 16
29  store <4 x i32> %tmp, <4 x i32>* %__a.addr.i, align 16
30  %tmp.i = load <4 x i32>* %__a.addr.i, align 16
31  %0 = bitcast <4 x i32> %tmp.i to <16 x i8>
32  %1 = bitcast <16 x i8> %0 to <4 x i32>
33  %vcvt.i = sitofp <4 x i32> %1 to <4 x float>
34  %tmp1 = load i8** %p.addr, align 4
35  %tmp2 = load i32* %offset.addr, align 4
36  %tmp3 = load <4 x float>** %constants.addr, align 4
37  call void @__bb(<4 x float> %vcvt.i, i8* %tmp1, i32 %tmp2, <4 x float>* %tmp3)
38  ret void
39}
40
41define internal void @__bb(<4 x float> %v, i8* %p, i32 %offset, <4 x float>* %constants) nounwind inlinehint ssp {
42entry:
43  %v.addr = alloca <4 x float>, align 16
44  %p.addr = alloca i8*, align 4
45  %offset.addr = alloca i32, align 4
46  %constants.addr = alloca <4 x float>*, align 4
47  %data = alloca i64, align 4
48  store <4 x float> %v, <4 x float>* %v.addr, align 16
49  store i8* %p, i8** %p.addr, align 4
50  store i32 %offset, i32* %offset.addr, align 4
51  store <4 x float>* %constants, <4 x float>** %constants.addr, align 4
52  %tmp = load i64* %data, align 4
53  %tmp1 = load i8** %p.addr, align 4
54  %tmp2 = load i32* %offset.addr, align 4
55  %add.ptr = getelementptr i8* %tmp1, i32 %tmp2
56  %0 = bitcast i8* %add.ptr to i64*
57  %arrayidx = getelementptr inbounds i64* %0, i32 0
58  store i64 %tmp, i64* %arrayidx
59  ret void
60}
61