1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -S -slp-vectorizer %s -slp-threshold=-10 | FileCheck %s
3target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
4target triple = "aarch64--linux-gnu"
5
6%structA = type { [2 x float] }
7
8define void @test1(%structA* nocapture readonly %J, i32 %xmin, i32 %ymin) {
9; CHECK-LABEL: @test1(
10; CHECK-NEXT:  entry:
11; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> poison, i32 [[XMIN:%.*]], i32 0
12; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32> [[TMP0]], i32 [[YMIN:%.*]], i32 1
13; CHECK-NEXT:    br label [[FOR_BODY3_LR_PH:%.*]]
14; CHECK:       for.body3.lr.ph:
15; CHECK-NEXT:    [[TMP2:%.*]] = sitofp <2 x i32> [[TMP1]] to <2 x float>
16; CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [[STRUCTA:%.*]], %structA* [[J:%.*]], i64 0, i32 0, i64 0
17; CHECK-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [[STRUCTA]], %structA* [[J]], i64 0, i32 0, i64 1
18; CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[ARRAYIDX4]] to <2 x float>*
19; CHECK-NEXT:    [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[TMP3]], align 4
20; CHECK-NEXT:    [[TMP5:%.*]] = fsub fast <2 x float> [[TMP2]], [[TMP4]]
21; CHECK-NEXT:    [[TMP6:%.*]] = fmul fast <2 x float> [[TMP5]], [[TMP5]]
22; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x float> [[TMP6]], i32 0
23; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x float> [[TMP6]], i32 1
24; CHECK-NEXT:    [[ADD:%.*]] = fadd fast float [[TMP7]], [[TMP8]]
25; CHECK-NEXT:    [[CMP:%.*]] = fcmp oeq float [[ADD]], 0.000000e+00
26; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY3_LR_PH]], label [[FOR_END27:%.*]]
27; CHECK:       for.end27:
28; CHECK-NEXT:    ret void
29;
30
31entry:
32  br label %for.body3.lr.ph
33
34for.body3.lr.ph:
35  %conv5 = sitofp i32 %ymin to float
36  %conv = sitofp i32 %xmin to float
37  %arrayidx4 = getelementptr inbounds %structA, %structA* %J, i64 0, i32 0, i64 0
38  %0 = load float, float* %arrayidx4, align 4
39  %sub = fsub fast float %conv, %0
40  %arrayidx9 = getelementptr inbounds %structA, %structA* %J, i64 0, i32 0, i64 1
41  %1 = load float, float* %arrayidx9, align 4
42  %sub10 = fsub fast float %conv5, %1
43  %mul11 = fmul fast float %sub, %sub
44  %mul12 = fmul fast float %sub10, %sub10
45  %add = fadd fast float %mul11, %mul12
46  %cmp = fcmp oeq float %add, 0.000000e+00
47  br i1 %cmp, label %for.body3.lr.ph, label %for.end27
48
49for.end27:
50  ret void
51}
52
53define void @test2(%structA* nocapture readonly %J, i32 %xmin, i32 %ymin) {
54; CHECK-LABEL: @test2(
55; CHECK-NEXT:  entry:
56; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> poison, i32 [[XMIN:%.*]], i32 0
57; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32> [[TMP0]], i32 [[YMIN:%.*]], i32 1
58; CHECK-NEXT:    br label [[FOR_BODY3_LR_PH:%.*]]
59; CHECK:       for.body3.lr.ph:
60; CHECK-NEXT:    [[TMP2:%.*]] = sitofp <2 x i32> [[TMP1]] to <2 x float>
61; CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [[STRUCTA:%.*]], %structA* [[J:%.*]], i64 0, i32 0, i64 0
62; CHECK-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [[STRUCTA]], %structA* [[J]], i64 0, i32 0, i64 1
63; CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[ARRAYIDX4]] to <2 x float>*
64; CHECK-NEXT:    [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[TMP3]], align 4
65; CHECK-NEXT:    [[TMP5:%.*]] = fsub fast <2 x float> [[TMP2]], [[TMP4]]
66; CHECK-NEXT:    [[TMP6:%.*]] = fmul fast <2 x float> [[TMP5]], [[TMP5]]
67; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x float> [[TMP6]], i32 0
68; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x float> [[TMP6]], i32 1
69; CHECK-NEXT:    [[ADD:%.*]] = fadd fast float [[TMP8]], [[TMP7]]
70; CHECK-NEXT:    [[CMP:%.*]] = fcmp oeq float [[ADD]], 0.000000e+00
71; CHECK-NEXT:    br i1 [[CMP]], label [[FOR_BODY3_LR_PH]], label [[FOR_END27:%.*]]
72; CHECK:       for.end27:
73; CHECK-NEXT:    ret void
74;
75
76entry:
77  br label %for.body3.lr.ph
78
79for.body3.lr.ph:
80  %conv5 = sitofp i32 %ymin to float
81  %conv = sitofp i32 %xmin to float
82  %arrayidx4 = getelementptr inbounds %structA, %structA* %J, i64 0, i32 0, i64 0
83  %0 = load float, float* %arrayidx4, align 4
84  %sub = fsub fast float %conv, %0
85  %arrayidx9 = getelementptr inbounds %structA, %structA* %J, i64 0, i32 0, i64 1
86  %1 = load float, float* %arrayidx9, align 4
87  %sub10 = fsub fast float %conv5, %1
88  %mul11 = fmul fast float %sub, %sub
89  %mul12 = fmul fast float %sub10, %sub10
90  %add = fadd fast float %mul12, %mul11         ;;;<---- Operands commuted!!
91  %cmp = fcmp oeq float %add, 0.000000e+00
92  br i1 %cmp, label %for.body3.lr.ph, label %for.end27
93
94for.end27:
95  ret void
96}
97