1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=slp-vectorizer,instcombine -S -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 | FileCheck %s
define float @dotf(<4 x float> %x, <4 x float> %y) {
; CHECK-LABEL: @dotf(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[TMP0:%.*]] = fmul fast <4 x float> [[X:%.*]], [[Y:%.*]]
; CHECK-NEXT: [[TMP1:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float -0.000000e+00, <4 x float> [[TMP0]])
; CHECK-NEXT: ret float [[TMP1]]
;
entry:
%vecext = extractelement <4 x float> %x, i32 0
%vecext1 = extractelement <4 x float> %y, i32 0
%mul = fmul fast float %vecext, %vecext1
%vecext.1 = extractelement <4 x float> %x, i32 1
%vecext1.1 = extractelement <4 x float> %y, i32 1
%mul.1 = fmul fast float %vecext.1, %vecext1.1
%add.1 = fadd fast float %mul.1, %mul
%vecext.2 = extractelement <4 x float> %x, i32 2
%vecext1.2 = extractelement <4 x float> %y, i32 2
%mul.2 = fmul fast float %vecext.2, %vecext1.2
%add.2 = fadd fast float %mul.2, %add.1
%vecext.3 = extractelement <4 x float> %x, i32 3
%vecext1.3 = extractelement <4 x float> %y, i32 3
%mul.3 = fmul fast float %vecext.3, %vecext1.3
%add.3 = fadd fast float %mul.3, %add.2
ret float %add.3
}
define double @dotd(ptr byval(<4 x double>) nocapture readonly align 32, ptr byval(<4 x double>) nocapture readonly align 32) {
; CHECK-LABEL: @dotd(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[X:%.*]] = load <4 x double>, ptr [[TMP0:%.*]], align 32
; CHECK-NEXT: [[Y:%.*]] = load <4 x double>, ptr [[TMP1:%.*]], align 32
; CHECK-NEXT: [[TMP2:%.*]] = fmul fast <4 x double> [[X]], [[Y]]
; CHECK-NEXT: [[TMP3:%.*]] = call fast double @llvm.vector.reduce.fadd.v4f64(double -0.000000e+00, <4 x double> [[TMP2]])
; CHECK-NEXT: ret double [[TMP3]]
;
entry:
%x = load <4 x double>, ptr %0, align 32
%y = load <4 x double>, ptr %1, align 32
%vecext = extractelement <4 x double> %x, i32 0
%vecext1 = extractelement <4 x double> %y, i32 0
%mul = fmul fast double %vecext, %vecext1
%vecext.1 = extractelement <4 x double> %x, i32 1
%vecext1.1 = extractelement <4 x double> %y, i32 1
%mul.1 = fmul fast double %vecext.1, %vecext1.1
%add.1 = fadd fast double %mul.1, %mul
%vecext.2 = extractelement <4 x double> %x, i32 2
%vecext1.2 = extractelement <4 x double> %y, i32 2
%mul.2 = fmul fast double %vecext.2, %vecext1.2
%add.2 = fadd fast double %mul.2, %add.1
%vecext.3 = extractelement <4 x double> %x, i32 3
%vecext1.3 = extractelement <4 x double> %y, i32 3
%mul.3 = fmul fast double %vecext.3, %vecext1.3
%add.3 = fadd fast double %mul.3, %add.2
ret double %add.3
}
define float @dotfq(ptr nocapture readonly %x, ptr nocapture readonly %y) {
; CHECK-LABEL: @dotfq(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[TMP0:%.*]] = load <4 x float>, ptr [[X:%.*]], align 16
; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, ptr [[Y:%.*]], align 16
; CHECK-NEXT: [[TMP2:%.*]] = fmul fast <4 x float> [[TMP1]], [[TMP0]]
; CHECK-NEXT: [[TMP3:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float -0.000000e+00, <4 x float> [[TMP2]])
; CHECK-NEXT: ret float [[TMP3]]
;
entry:
%0 = load <4 x float>, ptr %x, align 16
%1 = load <4 x float>, ptr %y, align 16
%vecext = extractelement <4 x float> %0, i32 0
%vecext1 = extractelement <4 x float> %1, i32 0
%mul = fmul fast float %vecext1, %vecext
%vecext.1 = extractelement <4 x float> %0, i32 1
%vecext1.1 = extractelement <4 x float> %1, i32 1
%mul.1 = fmul fast float %vecext1.1, %vecext.1
%add.1 = fadd fast float %mul.1, %mul
%vecext.2 = extractelement <4 x float> %0, i32 2
%vecext1.2 = extractelement <4 x float> %1, i32 2
%mul.2 = fmul fast float %vecext1.2, %vecext.2
%add.2 = fadd fast float %mul.2, %add.1
%vecext.3 = extractelement <4 x float> %0, i32 3
%vecext1.3 = extractelement <4 x float> %1, i32 3
%mul.3 = fmul fast float %vecext1.3, %vecext.3
%add.3 = fadd fast float %mul.3, %add.2
ret float %add.3
}
define double @dotdq(ptr nocapture readonly %x, ptr nocapture readonly %y) {
; CHECK-LABEL: @dotdq(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[TMP0:%.*]] = load <4 x double>, ptr [[X:%.*]], align 32
; CHECK-NEXT: [[TMP1:%.*]] = load <4 x double>, ptr [[Y:%.*]], align 32
; CHECK-NEXT: [[TMP2:%.*]] = fmul fast <4 x double> [[TMP1]], [[TMP0]]
; CHECK-NEXT: [[TMP3:%.*]] = call fast double @llvm.vector.reduce.fadd.v4f64(double -0.000000e+00, <4 x double> [[TMP2]])
; CHECK-NEXT: ret double [[TMP3]]
;
entry:
%0 = load <4 x double>, ptr %x, align 32
%1 = load <4 x double>, ptr %y, align 32
%vecext = extractelement <4 x double> %0, i32 0
%vecext1 = extractelement <4 x double> %1, i32 0
%mul = fmul fast double %vecext1, %vecext
%vecext.1 = extractelement <4 x double> %0, i32 1
%vecext1.1 = extractelement <4 x double> %1, i32 1
%mul.1 = fmul fast double %vecext1.1, %vecext.1
%add.1 = fadd fast double %mul.1, %mul
%vecext.2 = extractelement <4 x double> %0, i32 2
%vecext1.2 = extractelement <4 x double> %1, i32 2
%mul.2 = fmul fast double %vecext1.2, %vecext.2
%add.2 = fadd fast double %mul.2, %add.1
%vecext.3 = extractelement <4 x double> %0, i32 3
%vecext1.3 = extractelement <4 x double> %1, i32 3
%mul.3 = fmul fast double %vecext1.3, %vecext.3
%add.3 = fadd fast double %mul.3, %add.2
ret double %add.3
}
|
