1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 | FileCheck %s
3
4define float @dotf(<4 x float> %x, <4 x float> %y) {
5; CHECK-LABEL: @dotf(
6; CHECK-NEXT:  entry:
7; CHECK-NEXT:    [[TMP0:%.*]] = fmul fast <4 x float> [[X:%.*]], [[Y:%.*]]
8; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
9; CHECK-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
10; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x float> [[TMP1]], [[RDX_SHUF]]
11; CHECK-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x float> [[BIN_RDX]], <4 x float> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
12; CHECK-NEXT:    [[BIN_RDX2:%.*]] = fadd fast <4 x float> [[BIN_RDX]], [[RDX_SHUF1]]
13; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[BIN_RDX2]], i32 0
14; CHECK-NEXT:    ret float [[TMP2]]
15;
16entry:
17  %vecext = extractelement <4 x float> %x, i32 0
18  %vecext1 = extractelement <4 x float> %y, i32 0
19  %mul = fmul fast float %vecext, %vecext1
20  %vecext.1 = extractelement <4 x float> %x, i32 1
21  %vecext1.1 = extractelement <4 x float> %y, i32 1
22  %mul.1 = fmul fast float %vecext.1, %vecext1.1
23  %add.1 = fadd fast float %mul.1, %mul
24  %vecext.2 = extractelement <4 x float> %x, i32 2
25  %vecext1.2 = extractelement <4 x float> %y, i32 2
26  %mul.2 = fmul fast float %vecext.2, %vecext1.2
27  %add.2 = fadd fast float %mul.2, %add.1
28  %vecext.3 = extractelement <4 x float> %x, i32 3
29  %vecext1.3 = extractelement <4 x float> %y, i32 3
30  %mul.3 = fmul fast float %vecext.3, %vecext1.3
31  %add.3 = fadd fast float %mul.3, %add.2
32  ret float %add.3
33}
34
35define double @dotd(<4 x double>* byval nocapture readonly align 32, <4 x double>* byval nocapture readonly align 32) {
36; CHECK-LABEL: @dotd(
37; CHECK-NEXT:  entry:
38; CHECK-NEXT:    [[X:%.*]] = load <4 x double>, <4 x double>* [[TMP0:%.*]], align 32
39; CHECK-NEXT:    [[Y:%.*]] = load <4 x double>, <4 x double>* [[TMP1:%.*]], align 32
40; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast <4 x double> [[X]], [[Y]]
41; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x double> [[TMP2]], <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
42; CHECK-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x double> [[TMP3]], <4 x double> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
43; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x double> [[TMP3]], [[RDX_SHUF]]
44; CHECK-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x double> [[BIN_RDX]], <4 x double> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
45; CHECK-NEXT:    [[BIN_RDX2:%.*]] = fadd fast <4 x double> [[BIN_RDX]], [[RDX_SHUF1]]
46; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x double> [[BIN_RDX2]], i32 0
47; CHECK-NEXT:    ret double [[TMP4]]
48;
49entry:
50  %x = load <4 x double>, <4 x double>* %0, align 32
51  %y = load <4 x double>, <4 x double>* %1, align 32
52  %vecext = extractelement <4 x double> %x, i32 0
53  %vecext1 = extractelement <4 x double> %y, i32 0
54  %mul = fmul fast double %vecext, %vecext1
55  %vecext.1 = extractelement <4 x double> %x, i32 1
56  %vecext1.1 = extractelement <4 x double> %y, i32 1
57  %mul.1 = fmul fast double %vecext.1, %vecext1.1
58  %add.1 = fadd fast double %mul.1, %mul
59  %vecext.2 = extractelement <4 x double> %x, i32 2
60  %vecext1.2 = extractelement <4 x double> %y, i32 2
61  %mul.2 = fmul fast double %vecext.2, %vecext1.2
62  %add.2 = fadd fast double %mul.2, %add.1
63  %vecext.3 = extractelement <4 x double> %x, i32 3
64  %vecext1.3 = extractelement <4 x double> %y, i32 3
65  %mul.3 = fmul fast double %vecext.3, %vecext1.3
66  %add.3 = fadd fast double %mul.3, %add.2
67  ret double %add.3
68}
69
70define float @dotfq(<4 x float>* nocapture readonly %x, <4 x float>* nocapture readonly %y) {
71; CHECK-LABEL: @dotfq(
72; CHECK-NEXT:  entry:
73; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x float>, <4 x float>* [[X:%.*]], align 16
74; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, <4 x float>* [[Y:%.*]], align 16
75; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast <4 x float> [[TMP1]], [[TMP0]]
76; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x float> [[TMP2]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
77; CHECK-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
78; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x float> [[TMP3]], [[RDX_SHUF]]
79; CHECK-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x float> [[BIN_RDX]], <4 x float> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
80; CHECK-NEXT:    [[BIN_RDX2:%.*]] = fadd fast <4 x float> [[BIN_RDX]], [[RDX_SHUF1]]
81; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x float> [[BIN_RDX2]], i32 0
82; CHECK-NEXT:    ret float [[TMP4]]
83;
84entry:
85  %0 = load <4 x float>, <4 x float>* %x, align 16
86  %1 = load <4 x float>, <4 x float>* %y, align 16
87  %vecext = extractelement <4 x float> %0, i32 0
88  %vecext1 = extractelement <4 x float> %1, i32 0
89  %mul = fmul fast float %vecext1, %vecext
90  %vecext.1 = extractelement <4 x float> %0, i32 1
91  %vecext1.1 = extractelement <4 x float> %1, i32 1
92  %mul.1 = fmul fast float %vecext1.1, %vecext.1
93  %add.1 = fadd fast float %mul.1, %mul
94  %vecext.2 = extractelement <4 x float> %0, i32 2
95  %vecext1.2 = extractelement <4 x float> %1, i32 2
96  %mul.2 = fmul fast float %vecext1.2, %vecext.2
97  %add.2 = fadd fast float %mul.2, %add.1
98  %vecext.3 = extractelement <4 x float> %0, i32 3
99  %vecext1.3 = extractelement <4 x float> %1, i32 3
100  %mul.3 = fmul fast float %vecext1.3, %vecext.3
101  %add.3 = fadd fast float %mul.3, %add.2
102  ret float %add.3
103}
104
105define double @dotdq(<4 x double>* nocapture readonly %x, <4 x double>* nocapture readonly %y) {
106; CHECK-LABEL: @dotdq(
107; CHECK-NEXT:  entry:
108; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x double>, <4 x double>* [[X:%.*]], align 32
109; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x double>, <4 x double>* [[Y:%.*]], align 32
110; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast <4 x double> [[TMP1]], [[TMP0]]
111; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x double> [[TMP2]], <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
112; CHECK-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <4 x double> [[TMP3]], <4 x double> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
113; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x double> [[TMP3]], [[RDX_SHUF]]
114; CHECK-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <4 x double> [[BIN_RDX]], <4 x double> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
115; CHECK-NEXT:    [[BIN_RDX2:%.*]] = fadd fast <4 x double> [[BIN_RDX]], [[RDX_SHUF1]]
116; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x double> [[BIN_RDX2]], i32 0
117; CHECK-NEXT:    ret double [[TMP4]]
118;
119entry:
120  %0 = load <4 x double>, <4 x double>* %x, align 32
121  %1 = load <4 x double>, <4 x double>* %y, align 32
122  %vecext = extractelement <4 x double> %0, i32 0
123  %vecext1 = extractelement <4 x double> %1, i32 0
124  %mul = fmul fast double %vecext1, %vecext
125  %vecext.1 = extractelement <4 x double> %0, i32 1
126  %vecext1.1 = extractelement <4 x double> %1, i32 1
127  %mul.1 = fmul fast double %vecext1.1, %vecext.1
128  %add.1 = fadd fast double %mul.1, %mul
129  %vecext.2 = extractelement <4 x double> %0, i32 2
130  %vecext1.2 = extractelement <4 x double> %1, i32 2
131  %mul.2 = fmul fast double %vecext1.2, %vecext.2
132  %add.2 = fadd fast double %mul.2, %add.1
133  %vecext.3 = extractelement <4 x double> %0, i32 3
134  %vecext1.3 = extractelement <4 x double> %1, i32 3
135  %mul.3 = fmul fast double %vecext1.3, %vecext.3
136  %add.3 = fadd fast double %mul.3, %add.2
137  ret double %add.3
138}
139