1; RUN: opt < %s -basicaa -slp-vectorizer -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
2
3target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
4target triple = "x86_64-apple-macosx10.8.0"
5
6; int foo(int * restrict B,  int * restrict A, int n, int m) {
7;   B[0] = n * A[0] + m * A[0];
8;   B[1] = n * A[1] + m * A[1];
9;   B[2] = n * A[2] + m * A[2];
10;   B[3] = n * A[3] + m * A[3];
11;   return 0;
12; }
13
14; CHECK-LABEL: @foo(
15; CHECK: load <4 x i32>
16; CHECK: mul <4 x i32>
17; CHECK: store <4 x i32>
18; CHECK: ret
19define i32 @foo(i32* noalias nocapture %B, i32* noalias nocapture %A, i32 %n, i32 %m) #0 {
20entry:
21  %0 = load i32, i32* %A, align 4
22  %mul238 = add i32 %m, %n
23  %add = mul i32 %0, %mul238
24  store i32 %add, i32* %B, align 4
25  %arrayidx4 = getelementptr inbounds i32, i32* %A, i64 1
26  %1 = load i32, i32* %arrayidx4, align 4
27  %add8 = mul i32 %1, %mul238
28  %arrayidx9 = getelementptr inbounds i32, i32* %B, i64 1
29  store i32 %add8, i32* %arrayidx9, align 4
30  %arrayidx10 = getelementptr inbounds i32, i32* %A, i64 2
31  %2 = load i32, i32* %arrayidx10, align 4
32  %add14 = mul i32 %2, %mul238
33  %arrayidx15 = getelementptr inbounds i32, i32* %B, i64 2
34  store i32 %add14, i32* %arrayidx15, align 4
35  %arrayidx16 = getelementptr inbounds i32, i32* %A, i64 3
36  %3 = load i32, i32* %arrayidx16, align 4
37  %add20 = mul i32 %3, %mul238
38  %arrayidx21 = getelementptr inbounds i32, i32* %B, i64 3
39  store i32 %add20, i32* %arrayidx21, align 4
40  ret i32 0
41}
42
43
44; int extr_user(int * restrict B,  int * restrict A, int n, int m) {
45;   B[0] = n * A[0] + m * A[0];
46;   B[1] = n * A[1] + m * A[1];
47;   B[2] = n * A[2] + m * A[2];
48;   B[3] = n * A[3] + m * A[3];
49;   return A[0];
50; }
51
52; CHECK-LABEL: @extr_user(
53; CHECK: load <4 x i32>
54; CHECK: store <4 x i32>
55; CHECK: extractelement <4 x i32>
56; CHECK-NEXT: ret
57define i32 @extr_user(i32* noalias nocapture %B, i32* noalias nocapture %A, i32 %n, i32 %m) {
58entry:
59  %0 = load i32, i32* %A, align 4
60  %mul238 = add i32 %m, %n
61  %add = mul i32 %0, %mul238
62  store i32 %add, i32* %B, align 4
63  %arrayidx4 = getelementptr inbounds i32, i32* %A, i64 1
64  %1 = load i32, i32* %arrayidx4, align 4
65  %add8 = mul i32 %1, %mul238
66  %arrayidx9 = getelementptr inbounds i32, i32* %B, i64 1
67  store i32 %add8, i32* %arrayidx9, align 4
68  %arrayidx10 = getelementptr inbounds i32, i32* %A, i64 2
69  %2 = load i32, i32* %arrayidx10, align 4
70  %add14 = mul i32 %2, %mul238
71  %arrayidx15 = getelementptr inbounds i32, i32* %B, i64 2
72  store i32 %add14, i32* %arrayidx15, align 4
73  %arrayidx16 = getelementptr inbounds i32, i32* %A, i64 3
74  %3 = load i32, i32* %arrayidx16, align 4
75  %add20 = mul i32 %3, %mul238
76  %arrayidx21 = getelementptr inbounds i32, i32* %B, i64 3
77  store i32 %add20, i32* %arrayidx21, align 4
78  ret i32 %0  ;<--------- This value has multiple users
79}
80
81; In this example we have an external user that is not the first element in the vector.
82; CHECK-LABEL: @extr_user1(
83; CHECK: load <4 x i32>
84; CHECK: store <4 x i32>
85; CHECK: extractelement <4 x i32>
86; CHECK-NEXT: ret
87define i32 @extr_user1(i32* noalias nocapture %B, i32* noalias nocapture %A, i32 %n, i32 %m) {
88entry:
89  %0 = load i32, i32* %A, align 4
90  %mul238 = add i32 %m, %n
91  %add = mul i32 %0, %mul238
92  store i32 %add, i32* %B, align 4
93  %arrayidx4 = getelementptr inbounds i32, i32* %A, i64 1
94  %1 = load i32, i32* %arrayidx4, align 4
95  %add8 = mul i32 %1, %mul238
96  %arrayidx9 = getelementptr inbounds i32, i32* %B, i64 1
97  store i32 %add8, i32* %arrayidx9, align 4
98  %arrayidx10 = getelementptr inbounds i32, i32* %A, i64 2
99  %2 = load i32, i32* %arrayidx10, align 4
100  %add14 = mul i32 %2, %mul238
101  %arrayidx15 = getelementptr inbounds i32, i32* %B, i64 2
102  store i32 %add14, i32* %arrayidx15, align 4
103  %arrayidx16 = getelementptr inbounds i32, i32* %A, i64 3
104  %3 = load i32, i32* %arrayidx16, align 4
105  %add20 = mul i32 %3, %mul238
106  %arrayidx21 = getelementptr inbounds i32, i32* %B, i64 3
107  store i32 %add20, i32* %arrayidx21, align 4
108  ret i32 %1  ;<--------- This value has multiple users
109}
110