1 ; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
3 target 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"
4 target triple = "x86_64-apple-macosx10.8.0"
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];
15 ; CHECK: load <4 x i32>
16 ; CHECK: mul <4 x i32>
17 ; CHECK: store <4 x i32>
19 define i32 @foo(i32* noalias nocapture %B, i32* noalias nocapture %A, i32 %n, i32 %m) #0 {
21 %0 = load 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* %A, i64 1
26 %1 = load i32* %arrayidx4, align 4
27 %add8 = mul i32 %1, %mul238
28 %arrayidx9 = getelementptr inbounds i32* %B, i64 1
29 store i32 %add8, i32* %arrayidx9, align 4
30 %arrayidx10 = getelementptr inbounds i32* %A, i64 2
31 %2 = load i32* %arrayidx10, align 4
32 %add14 = mul i32 %2, %mul238
33 %arrayidx15 = getelementptr inbounds i32* %B, i64 2
34 store i32 %add14, i32* %arrayidx15, align 4
35 %arrayidx16 = getelementptr inbounds i32* %A, i64 3
36 %3 = load i32* %arrayidx16, align 4
37 %add20 = mul i32 %3, %mul238
38 %arrayidx21 = getelementptr inbounds i32* %B, i64 3
39 store i32 %add20, i32* %arrayidx21, align 4
44 ; int foo_fail(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];
53 ; CHECK-NOT: load <4 x i32>
55 define i32 @foo_fail(i32* noalias nocapture %B, i32* noalias nocapture %A, i32 %n, i32 %m) {
57 %0 = load i32* %A, align 4
58 %mul238 = add i32 %m, %n
59 %add = mul i32 %0, %mul238
60 store i32 %add, i32* %B, align 4
61 %arrayidx4 = getelementptr inbounds i32* %A, i64 1
62 %1 = load i32* %arrayidx4, align 4
63 %add8 = mul i32 %1, %mul238
64 %arrayidx9 = getelementptr inbounds i32* %B, i64 1
65 store i32 %add8, i32* %arrayidx9, align 4
66 %arrayidx10 = getelementptr inbounds i32* %A, i64 2
67 %2 = load i32* %arrayidx10, align 4
68 %add14 = mul i32 %2, %mul238
69 %arrayidx15 = getelementptr inbounds i32* %B, i64 2
70 store i32 %add14, i32* %arrayidx15, align 4
71 %arrayidx16 = getelementptr inbounds i32* %A, i64 3
72 %3 = load i32* %arrayidx16, align 4
73 %add20 = mul i32 %3, %mul238
74 %arrayidx21 = getelementptr inbounds i32* %B, i64 3
75 store i32 %add20, i32* %arrayidx21, align 4
76 ret i32 %0 ;<--------- This value has multiple users and can't be vectorized.