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
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=slp-vectorizer -S | FileCheck %s
target triple = "aarch64--linux-gnu"
; This checks that we we prefer splats rather than load vectors + shuffles.
; A load + broadcast can be done efficiently with a single `ld1r` instruction.
define void @splat_loads_double(ptr %array1, ptr %array2, ptr %ptrA, ptr %ptrB) {
; CHECK-LABEL: @splat_loads_double(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[GEP_2_1:%.*]] = getelementptr inbounds double, ptr [[ARRAY2:%.*]], i64 1
; CHECK-NEXT: [[LD_2_0:%.*]] = load double, ptr [[ARRAY2]], align 8
; CHECK-NEXT: [[LD_2_1:%.*]] = load double, ptr [[GEP_2_1]], align 8
; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, ptr [[ARRAY1:%.*]], align 8
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x double> poison, double [[LD_2_0]], i32 0
; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <2 x double> [[TMP2]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP3:%.*]] = fmul <2 x double> [[TMP1]], [[SHUFFLE]]
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x double> poison, double [[LD_2_1]], i32 0
; CHECK-NEXT: [[SHUFFLE1:%.*]] = shufflevector <2 x double> [[TMP4]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP5:%.*]] = fmul <2 x double> [[TMP1]], [[SHUFFLE1]]
; CHECK-NEXT: [[TMP6:%.*]] = fadd <2 x double> [[TMP3]], [[TMP5]]
; CHECK-NEXT: store <2 x double> [[TMP6]], ptr [[ARRAY1]], align 8
; CHECK-NEXT: ret void
;
entry:
%gep_1_1 = getelementptr inbounds double, ptr %array1, i64 1
%ld_1_0 = load double, ptr %array1, align 8
%ld_1_1 = load double, ptr %gep_1_1, align 8
%gep_2_1 = getelementptr inbounds double, ptr %array2, i64 1
%ld_2_0 = load double, ptr %array2, align 8
%ld_2_1 = load double, ptr %gep_2_1, align 8
%mul0 = fmul double %ld_1_0, %ld_2_0
%mul1 = fmul double %ld_1_1, %ld_2_0
%mul2 = fmul double %ld_1_0, %ld_2_1
%mul3 = fmul double %ld_1_1, %ld_2_1
%add0 = fadd double %mul0, %mul2
%add1 = fadd double %mul1, %mul3
store double %add0, ptr %array1
store double %add1, ptr %gep_1_1
ret void
}
; Same but with float instead of double
define void @splat_loads_float(ptr %array1, ptr %array2, ptr %ptrA, ptr %ptrB) {
; CHECK-LABEL: @splat_loads_float(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[GEP_2_1:%.*]] = getelementptr inbounds float, ptr [[ARRAY2:%.*]], i64 1
; CHECK-NEXT: [[LD_2_0:%.*]] = load float, ptr [[ARRAY2]], align 8
; CHECK-NEXT: [[LD_2_1:%.*]] = load float, ptr [[GEP_2_1]], align 8
; CHECK-NEXT: [[TMP1:%.*]] = load <2 x float>, ptr [[ARRAY1:%.*]], align 8
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x float> poison, float [[LD_2_0]], i32 0
; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP3:%.*]] = fmul <2 x float> [[TMP1]], [[SHUFFLE]]
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x float> poison, float [[LD_2_1]], i32 0
; CHECK-NEXT: [[SHUFFLE1:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP5:%.*]] = fmul <2 x float> [[TMP1]], [[SHUFFLE1]]
; CHECK-NEXT: [[TMP6:%.*]] = fadd <2 x float> [[TMP3]], [[TMP5]]
; CHECK-NEXT: store <2 x float> [[TMP6]], ptr [[ARRAY1]], align 4
; CHECK-NEXT: ret void
;
entry:
%gep_1_1 = getelementptr inbounds float, ptr %array1, i64 1
%ld_1_0 = load float, ptr %array1, align 8
%ld_1_1 = load float, ptr %gep_1_1, align 8
%gep_2_1 = getelementptr inbounds float, ptr %array2, i64 1
%ld_2_0 = load float, ptr %array2, align 8
%ld_2_1 = load float, ptr %gep_2_1, align 8
%mul0 = fmul float %ld_1_0, %ld_2_0
%mul1 = fmul float %ld_1_1, %ld_2_0
%mul2 = fmul float %ld_1_0, %ld_2_1
%mul3 = fmul float %ld_1_1, %ld_2_1
%add0 = fadd float %mul0, %mul2
%add1 = fadd float %mul1, %mul3
store float %add0, ptr %array1
store float %add1, ptr %gep_1_1
ret void
}
; Same but with i64
define void @splat_loads_i64(ptr %array1, ptr %array2, ptr %ptrA, ptr %ptrB) {
; CHECK-LABEL: @splat_loads_i64(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[GEP_2_1:%.*]] = getelementptr inbounds i64, ptr [[ARRAY2:%.*]], i64 1
; CHECK-NEXT: [[LD_2_0:%.*]] = load i64, ptr [[ARRAY2]], align 8
; CHECK-NEXT: [[LD_2_1:%.*]] = load i64, ptr [[GEP_2_1]], align 8
; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr [[ARRAY1:%.*]], align 8
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x i64> poison, i64 [[LD_2_0]], i32 0
; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <2 x i64> [[TMP2]], <2 x i64> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP3:%.*]] = or <2 x i64> [[TMP1]], [[SHUFFLE]]
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x i64> poison, i64 [[LD_2_1]], i32 0
; CHECK-NEXT: [[SHUFFLE1:%.*]] = shufflevector <2 x i64> [[TMP4]], <2 x i64> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP5:%.*]] = or <2 x i64> [[TMP1]], [[SHUFFLE1]]
; CHECK-NEXT: [[TMP6:%.*]] = add <2 x i64> [[TMP3]], [[TMP5]]
; CHECK-NEXT: store <2 x i64> [[TMP6]], ptr [[ARRAY1]], align 4
; CHECK-NEXT: ret void
;
entry:
%gep_1_1 = getelementptr inbounds i64, ptr %array1, i64 1
%ld_1_0 = load i64, ptr %array1, align 8
%ld_1_1 = load i64, ptr %gep_1_1, align 8
%gep_2_1 = getelementptr inbounds i64, ptr %array2, i64 1
%ld_2_0 = load i64, ptr %array2, align 8
%ld_2_1 = load i64, ptr %gep_2_1, align 8
%or0 = or i64 %ld_1_0, %ld_2_0
%or1 = or i64 %ld_1_1, %ld_2_0
%or2 = or i64 %ld_1_0, %ld_2_1
%or3 = or i64 %ld_1_1, %ld_2_1
%add0 = add i64 %or0, %or2
%add1 = add i64 %or1, %or3
store i64 %add0, ptr %array1
store i64 %add1, ptr %gep_1_1
ret void
}
; Same but with i32
define void @splat_loads_i32(ptr %array1, ptr %array2, ptr %ptrA, ptr %ptrB) {
; CHECK-LABEL: @splat_loads_i32(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[GEP_2_1:%.*]] = getelementptr inbounds i32, ptr [[ARRAY2:%.*]], i64 1
; CHECK-NEXT: [[LD_2_0:%.*]] = load i32, ptr [[ARRAY2]], align 8
; CHECK-NEXT: [[LD_2_1:%.*]] = load i32, ptr [[GEP_2_1]], align 8
; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i32>, ptr [[ARRAY1:%.*]], align 8
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x i32> poison, i32 [[LD_2_0]], i32 0
; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <2 x i32> [[TMP2]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP3:%.*]] = or <2 x i32> [[TMP1]], [[SHUFFLE]]
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x i32> poison, i32 [[LD_2_1]], i32 0
; CHECK-NEXT: [[SHUFFLE1:%.*]] = shufflevector <2 x i32> [[TMP4]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP5:%.*]] = or <2 x i32> [[TMP1]], [[SHUFFLE1]]
; CHECK-NEXT: [[TMP6:%.*]] = add <2 x i32> [[TMP3]], [[TMP5]]
; CHECK-NEXT: store <2 x i32> [[TMP6]], ptr [[ARRAY1]], align 4
; CHECK-NEXT: ret void
;
entry:
%gep_1_1 = getelementptr inbounds i32, ptr %array1, i64 1
%ld_1_0 = load i32, ptr %array1, align 8
%ld_1_1 = load i32, ptr %gep_1_1, align 8
%gep_2_1 = getelementptr inbounds i32, ptr %array2, i64 1
%ld_2_0 = load i32, ptr %array2, align 8
%ld_2_1 = load i32, ptr %gep_2_1, align 8
%or0 = or i32 %ld_1_0, %ld_2_0
%or1 = or i32 %ld_1_1, %ld_2_0
%or2 = or i32 %ld_1_0, %ld_2_1
%or3 = or i32 %ld_1_1, %ld_2_1
%add0 = add i32 %or0, %or2
%add1 = add i32 %or1, %or3
store i32 %add0, ptr %array1
store i32 %add1, ptr %gep_1_1
ret void
}
|
