1; RUN: opt -S -slp-vectorizer %s | FileCheck %s
2target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
3target triple = "aarch64--linux-gnu"
4
5%structA = type { [2 x float] }
6
7define void @test1(%structA* nocapture readonly %J, i32 %xmin, i32 %ymin) {
8; CHECK-LABEL: test1
9; CHECK: %arrayidx4 = getelementptr inbounds %structA* %J, i64 0, i32 0, i64 0
10; CHECK: %arrayidx9 = getelementptr inbounds %structA* %J, i64 0, i32 0, i64 1
11; CHECK: %3 = bitcast float* %arrayidx4 to <2 x float>*
12; CHECK: %4 = load <2 x float>* %3, align 4
13; CHECK: %5 = fsub fast <2 x float> %2, %4
14; CHECK: %6 = fmul fast <2 x float> %5, %5
15; CHECK: %7 = extractelement <2 x float> %6, i32 0
16; CHECK: %8 = extractelement <2 x float> %6, i32 1
17; CHECK: %add = fadd fast float %7, %8
18; CHECK: %cmp = fcmp oeq float %add, 0.000000e+00
19
20entry:
21  br label %for.body3.lr.ph
22
23for.body3.lr.ph:
24  %conv5 = sitofp i32 %ymin to float
25  %conv = sitofp i32 %xmin to float
26  %arrayidx4 = getelementptr inbounds %structA* %J, i64 0, i32 0, i64 0
27  %0 = load float* %arrayidx4, align 4
28  %sub = fsub fast float %conv, %0
29  %arrayidx9 = getelementptr inbounds %structA* %J, i64 0, i32 0, i64 1
30  %1 = load float* %arrayidx9, align 4
31  %sub10 = fsub fast float %conv5, %1
32  %mul11 = fmul fast float %sub, %sub
33  %mul12 = fmul fast float %sub10, %sub10
34  %add = fadd fast float %mul11, %mul12
35  %cmp = fcmp oeq float %add, 0.000000e+00
36  br i1 %cmp, label %for.body3.lr.ph, label %for.end27
37
38for.end27:
39  ret void
40}
41
42define void @test2(%structA* nocapture readonly %J, i32 %xmin, i32 %ymin) {
43; CHECK-LABEL: test2
44; CHECK: %arrayidx4 = getelementptr inbounds %structA* %J, i64 0, i32 0, i64 0
45; CHECK: %arrayidx9 = getelementptr inbounds %structA* %J, i64 0, i32 0, i64 1
46; CHECK: %3 = bitcast float* %arrayidx4 to <2 x float>*
47; CHECK: %4 = load <2 x float>* %3, align 4
48; CHECK: %5 = fsub fast <2 x float> %2, %4
49; CHECK: %6 = fmul fast <2 x float> %5, %5
50; CHECK: %7 = extractelement <2 x float> %6, i32 0
51; CHECK: %8 = extractelement <2 x float> %6, i32 1
52; CHECK: %add = fadd fast float %8, %7
53; CHECK: %cmp = fcmp oeq float %add, 0.000000e+00
54
55entry:
56  br label %for.body3.lr.ph
57
58for.body3.lr.ph:
59  %conv5 = sitofp i32 %ymin to float
60  %conv = sitofp i32 %xmin to float
61  %arrayidx4 = getelementptr inbounds %structA* %J, i64 0, i32 0, i64 0
62  %0 = load float* %arrayidx4, align 4
63  %sub = fsub fast float %conv, %0
64  %arrayidx9 = getelementptr inbounds %structA* %J, i64 0, i32 0, i64 1
65  %1 = load float* %arrayidx9, align 4
66  %sub10 = fsub fast float %conv5, %1
67  %mul11 = fmul fast float %sub, %sub
68  %mul12 = fmul fast float %sub10, %sub10
69  %add = fadd fast float %mul12, %mul11         ;;;<---- Operands commuted!!
70  %cmp = fcmp oeq float %add, 0.000000e+00
71  br i1 %cmp, label %for.body3.lr.ph, label %for.end27
72
73for.end27:
74  ret void
75}
76