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
|
; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
; PR12281
; Test generataion of code for vmull instruction when multiplying 128-bit
; vectors that were created by sign-extending smaller vector sizes.
;
; The vmull operation requires 64-bit vectors, so we must extend the original
; vector size to 64 bits for vmull operation.
; Previously failed with an assertion because the <4 x i8> vector was too small
; for vmull.
; Vector x Constant
; v4i8
;
define void @sextload_v4i8_c(<4 x i8>* %v) nounwind {
;CHECK: sextload_v4i8_c:
entry:
%0 = load <4 x i8>* %v, align 8
%v0 = sext <4 x i8> %0 to <4 x i32>
;CHECK: vmull
%v1 = mul <4 x i32> %v0, <i32 3, i32 3, i32 3, i32 3>
store <4 x i32> %v1, <4 x i32>* undef, align 8
ret void;
}
; v2i8
;
define void @sextload_v2i8_c(<2 x i8>* %v) nounwind {
;CHECK: sextload_v2i8_c:
entry:
%0 = load <2 x i8>* %v, align 8
%v0 = sext <2 x i8> %0 to <2 x i64>
;CHECK: vmull
%v1 = mul <2 x i64> %v0, <i64 3, i64 3>
store <2 x i64> %v1, <2 x i64>* undef, align 8
ret void;
}
; v2i16
;
define void @sextload_v2i16_c(<2 x i16>* %v) nounwind {
;CHECK: sextload_v2i16_c:
entry:
%0 = load <2 x i16>* %v, align 8
%v0 = sext <2 x i16> %0 to <2 x i64>
;CHECK: vmull
%v1 = mul <2 x i64> %v0, <i64 3, i64 3>
store <2 x i64> %v1, <2 x i64>* undef, align 8
ret void;
}
; Vector x Vector
; v4i8
;
define void @sextload_v4i8_v(<4 x i8>* %v, <4 x i8>* %p) nounwind {
;CHECK: sextload_v4i8_v:
entry:
%0 = load <4 x i8>* %v, align 8
%v0 = sext <4 x i8> %0 to <4 x i32>
%1 = load <4 x i8>* %p, align 8
%v2 = sext <4 x i8> %1 to <4 x i32>
;CHECK: vmull
%v1 = mul <4 x i32> %v0, %v2
store <4 x i32> %v1, <4 x i32>* undef, align 8
ret void;
}
; v2i8
;
define void @sextload_v2i8_v(<2 x i8>* %v, <2 x i8>* %p) nounwind {
;CHECK: sextload_v2i8_v:
entry:
%0 = load <2 x i8>* %v, align 8
%v0 = sext <2 x i8> %0 to <2 x i64>
%1 = load <2 x i8>* %p, align 8
%v2 = sext <2 x i8> %1 to <2 x i64>
;CHECK: vmull
%v1 = mul <2 x i64> %v0, %v2
store <2 x i64> %v1, <2 x i64>* undef, align 8
ret void;
}
; v2i16
;
define void @sextload_v2i16_v(<2 x i16>* %v, <2 x i16>* %p) nounwind {
;CHECK: sextload_v2i16_v:
entry:
%0 = load <2 x i16>* %v, align 8
%v0 = sext <2 x i16> %0 to <2 x i64>
%1 = load <2 x i16>* %p, align 8
%v2 = sext <2 x i16> %1 to <2 x i64>
;CHECK: vmull
%v1 = mul <2 x i64> %v0, %v2
store <2 x i64> %v1, <2 x i64>* undef, align 8
ret void;
}
; Vector(small) x Vector(big)
; v4i8 x v4i16
;
define void @sextload_v4i8_vs(<4 x i8>* %v, <4 x i16>* %p) nounwind {
;CHECK: sextload_v4i8_vs:
entry:
%0 = load <4 x i8>* %v, align 8
%v0 = sext <4 x i8> %0 to <4 x i32>
%1 = load <4 x i16>* %p, align 8
%v2 = sext <4 x i16> %1 to <4 x i32>
;CHECK: vmull
%v1 = mul <4 x i32> %v0, %v2
store <4 x i32> %v1, <4 x i32>* undef, align 8
ret void;
}
; v2i8
; v2i8 x v2i16
define void @sextload_v2i8_vs(<2 x i8>* %v, <2 x i16>* %p) nounwind {
;CHECK: sextload_v2i8_vs:
entry:
%0 = load <2 x i8>* %v, align 8
%v0 = sext <2 x i8> %0 to <2 x i64>
%1 = load <2 x i16>* %p, align 8
%v2 = sext <2 x i16> %1 to <2 x i64>
;CHECK: vmull
%v1 = mul <2 x i64> %v0, %v2
store <2 x i64> %v1, <2 x i64>* undef, align 8
ret void;
}
; v2i16
; v2i16 x v2i32
define void @sextload_v2i16_vs(<2 x i16>* %v, <2 x i32>* %p) nounwind {
;CHECK: sextload_v2i16_vs:
entry:
%0 = load <2 x i16>* %v, align 8
%v0 = sext <2 x i16> %0 to <2 x i64>
%1 = load <2 x i32>* %p, align 8
%v2 = sext <2 x i32> %1 to <2 x i64>
;CHECK: vmull
%v1 = mul <2 x i64> %v0, %v2
store <2 x i64> %v1, <2 x i64>* undef, align 8
ret void;
}
|
