1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s -basic-aa -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-NEXT:  entry:
7; CHECK-NEXT:    [[M2:%.*]] = alloca [8 x [8 x i32]], align 16
8; CHECK-NEXT:    [[TMP0:%.*]] = bitcast [8 x [8 x i32]]* [[M2]] to i8*
9; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
10; CHECK:       for.body:
11; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
12; CHECK-NEXT:    [[A_088:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[OP_EXTRA:%.*]], [[FOR_BODY]] ]
13; CHECK-NEXT:    [[TMP1:%.*]] = shl i64 [[INDVARS_IV]], 3
14; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[DIFF:%.*]], i64 [[TMP1]]
15; CHECK-NEXT:    [[TMP2:%.*]] = or i64 [[TMP1]], 4
16; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP2]]
17; CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* [[M2]], i64 0, i64 [[INDVARS_IV]], i64 0
18; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[TMP1]], 1
19; CHECK-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP3]]
20; CHECK-NEXT:    [[TMP4:%.*]] = or i64 [[TMP1]], 5
21; CHECK-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP4]]
22; CHECK-NEXT:    [[TMP5:%.*]] = or i64 [[TMP1]], 2
23; CHECK-NEXT:    [[ARRAYIDX27:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP5]]
24; CHECK-NEXT:    [[TMP6:%.*]] = or i64 [[TMP1]], 6
25; CHECK-NEXT:    [[ARRAYIDX30:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP6]]
26; CHECK-NEXT:    [[TMP7:%.*]] = or i64 [[TMP1]], 3
27; CHECK-NEXT:    [[ARRAYIDX41:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP7]]
28; CHECK-NEXT:    [[TMP8:%.*]] = bitcast i32* [[ARRAYIDX]] to <4 x i32>*
29; CHECK-NEXT:    [[TMP9:%.*]] = load <4 x i32>, <4 x i32>* [[TMP8]], align 4
30; CHECK-NEXT:    [[TMP10:%.*]] = or i64 [[TMP1]], 7
31; CHECK-NEXT:    [[ARRAYIDX44:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP10]]
32; CHECK-NEXT:    [[TMP11:%.*]] = bitcast i32* [[ARRAYIDX2]] to <4 x i32>*
33; CHECK-NEXT:    [[TMP12:%.*]] = load <4 x i32>, <4 x i32>* [[TMP11]], align 4
34; CHECK-NEXT:    [[TMP13:%.*]] = add nsw <4 x i32> [[TMP12]], [[TMP9]]
35; CHECK-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* [[M2]], i64 0, i64 [[INDVARS_IV]], i64 1
36; CHECK-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* [[M2]], i64 0, i64 [[INDVARS_IV]], i64 2
37; CHECK-NEXT:    [[ARRAYIDX48:%.*]] = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* [[M2]], i64 0, i64 [[INDVARS_IV]], i64 3
38; CHECK-NEXT:    [[TMP14:%.*]] = bitcast i32* [[ARRAYIDX6]] to <4 x i32>*
39; CHECK-NEXT:    store <4 x i32> [[TMP13]], <4 x i32>* [[TMP14]], align 16
40; CHECK-NEXT:    [[TMP15:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP13]])
41; CHECK-NEXT:    [[OP_EXTRA]] = add nsw i32 [[TMP15]], [[A_088]]
42; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
43; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 8
44; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
45; CHECK:       for.end:
46; CHECK-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* [[M2]], i64 0, i64 0
47; CHECK-NEXT:    call void @ff([8 x i32]* [[ARRAYDECAY]])
48; CHECK-NEXT:    ret i32 [[OP_EXTRA]]
49;
50entry:
51  %m2 = alloca [8 x [8 x i32]], align 16
52  %0 = bitcast [8 x [8 x i32]]* %m2 to i8*
53  br label %for.body
54
55for.body:                                         ; preds = %for.body, %entry
56  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
57  %a.088 = phi i32 [ 0, %entry ], [ %add52, %for.body ]
58  %1 = shl i64 %indvars.iv, 3
59  %arrayidx = getelementptr inbounds i32, i32* %diff, i64 %1
60  %2 = load i32, i32* %arrayidx, align 4
61  %3 = or i64 %1, 4
62  %arrayidx2 = getelementptr inbounds i32, i32* %diff, i64 %3
63  %4 = load i32, i32* %arrayidx2, align 4
64  %add3 = add nsw i32 %4, %2
65  %arrayidx6 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 %indvars.iv, i64 0
66  store i32 %add3, i32* %arrayidx6, align 16
67  %add10 = add nsw i32 %add3, %a.088
68  %5 = or i64 %1, 1
69  %arrayidx13 = getelementptr inbounds i32, i32* %diff, i64 %5
70  %6 = load i32, i32* %arrayidx13, align 4
71  %7 = or i64 %1, 5
72  %arrayidx16 = getelementptr inbounds i32, i32* %diff, i64 %7
73  %8 = load i32, i32* %arrayidx16, align 4
74  %add17 = add nsw i32 %8, %6
75  %arrayidx20 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 %indvars.iv, i64 1
76  store i32 %add17, i32* %arrayidx20, align 4
77  %add24 = add nsw i32 %add10, %add17
78  %9 = or i64 %1, 2
79  %arrayidx27 = getelementptr inbounds i32, i32* %diff, i64 %9
80  %10 = load i32, i32* %arrayidx27, align 4
81  %11 = or i64 %1, 6
82  %arrayidx30 = getelementptr inbounds i32, i32* %diff, i64 %11
83  %12 = load i32, i32* %arrayidx30, align 4
84  %add31 = add nsw i32 %12, %10
85  %arrayidx34 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 %indvars.iv, i64 2
86  store i32 %add31, i32* %arrayidx34, align 8
87  %add38 = add nsw i32 %add24, %add31
88  %13 = or i64 %1, 3
89  %arrayidx41 = getelementptr inbounds i32, i32* %diff, i64 %13
90  %14 = load i32, i32* %arrayidx41, align 4
91  %15 = or i64 %1, 7
92  %arrayidx44 = getelementptr inbounds i32, i32* %diff, i64 %15
93  %16 = load i32, i32* %arrayidx44, align 4
94  %add45 = add nsw i32 %16, %14
95  %arrayidx48 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 %indvars.iv, i64 3
96  store i32 %add45, i32* %arrayidx48, align 4
97  %add52 = add nsw i32 %add38, %add45
98  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
99  %exitcond = icmp eq i64 %indvars.iv.next, 8
100  br i1 %exitcond, label %for.end, label %for.body
101
102for.end:                                          ; preds = %for.body
103  %arraydecay = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 0
104  call void @ff([8 x i32]* %arraydecay) #1
105  ret i32 %add52
106}
107
108declare void @ff([8 x i32]*) #2
109
110
111