1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s -basicaa -slp-vectorizer -S -mtriple=i386-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
3
4define i32 @foo(i32* nocapture readonly %diff) #0 {
5; CHECK-LABEL: @foo(
6; CHECK: load <4 x i32>
7; CHECK: load <4 x i32>
8; CHECK: [[S1:%.+]] = add nsw <4 x i32>
9; CHECK: store <4 x i32> [[S1]],
10; CHECK:         [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[S1]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
11; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[S1]], [[RDX_SHUF]]
12; CHECK-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x i32> [[BIN_RDX]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
13; CHECK-NEXT:    [[BIN_RDX2:%.*]] = add <4 x i32> [[BIN_RDX]], [[RDX_SHUF1]]
14; CHECK-NEXT:    [[TMP15:%.*]] = extractelement <4 x i32> [[BIN_RDX2]], i32 0
15; CHECK-NEXT:    [[ADD52:%.*]] = add nsw i32 [[TMP15]],
16; CHECK:          ret i32 [[ADD52]]
17;
18entry:
19  %m2 = alloca [8 x [8 x i32]], align 16
20  %0 = bitcast [8 x [8 x i32]]* %m2 to i8*
21  br label %for.body
22
23for.body:                                         ; preds = %for.body, %entry
24  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
25  %a.088 = phi i32 [ 0, %entry ], [ %add52, %for.body ]
26  %1 = shl i64 %indvars.iv, 3
27  %arrayidx = getelementptr inbounds i32, i32* %diff, i64 %1
28  %2 = load i32, i32* %arrayidx, align 4
29  %3 = or i64 %1, 4
30  %arrayidx2 = getelementptr inbounds i32, i32* %diff, i64 %3
31  %4 = load i32, i32* %arrayidx2, align 4
32  %add3 = add nsw i32 %4, %2
33  %arrayidx6 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 %indvars.iv, i64 0
34  store i32 %add3, i32* %arrayidx6, align 16
35  %add10 = add nsw i32 %add3, %a.088
36  %5 = or i64 %1, 1
37  %arrayidx13 = getelementptr inbounds i32, i32* %diff, i64 %5
38  %6 = load i32, i32* %arrayidx13, align 4
39  %7 = or i64 %1, 5
40  %arrayidx16 = getelementptr inbounds i32, i32* %diff, i64 %7
41  %8 = load i32, i32* %arrayidx16, align 4
42  %add17 = add nsw i32 %8, %6
43  %arrayidx20 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 %indvars.iv, i64 1
44  store i32 %add17, i32* %arrayidx20, align 4
45  %add24 = add nsw i32 %add10, %add17
46  %9 = or i64 %1, 2
47  %arrayidx27 = getelementptr inbounds i32, i32* %diff, i64 %9
48  %10 = load i32, i32* %arrayidx27, align 4
49  %11 = or i64 %1, 6
50  %arrayidx30 = getelementptr inbounds i32, i32* %diff, i64 %11
51  %12 = load i32, i32* %arrayidx30, align 4
52  %add31 = add nsw i32 %12, %10
53  %arrayidx34 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 %indvars.iv, i64 2
54  store i32 %add31, i32* %arrayidx34, align 8
55  %add38 = add nsw i32 %add24, %add31
56  %13 = or i64 %1, 3
57  %arrayidx41 = getelementptr inbounds i32, i32* %diff, i64 %13
58  %14 = load i32, i32* %arrayidx41, align 4
59  %15 = or i64 %1, 7
60  %arrayidx44 = getelementptr inbounds i32, i32* %diff, i64 %15
61  %16 = load i32, i32* %arrayidx44, align 4
62  %add45 = add nsw i32 %16, %14
63  %arrayidx48 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 %indvars.iv, i64 3
64  store i32 %add45, i32* %arrayidx48, align 4
65  %add52 = add nsw i32 %add38, %add45
66  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
67  %exitcond = icmp eq i64 %indvars.iv.next, 8
68  br i1 %exitcond, label %for.end, label %for.body
69
70for.end:                                          ; preds = %for.body
71  %arraydecay = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 0
72  call void @ff([8 x i32]* %arraydecay) #1
73  ret i32 %add52
74}
75
76declare void @ff([8 x i32]*) #2
77
78
79