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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
|
; RUN: llc < %s -mattr=+neon | FileCheck %s
target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:32-f32:32:32-f64:32:32-v64:64:64-v128:128:128-a0:0:32"
target triple = "thumbv7-elf"
define i32 @vget_lanes8(<8 x i8>* %A) nounwind {
;CHECK: vget_lanes8:
;CHECK: vmov.s8
%tmp1 = load <8 x i8>* %A
%tmp2 = extractelement <8 x i8> %tmp1, i32 1
%tmp3 = sext i8 %tmp2 to i32
ret i32 %tmp3
}
define i32 @vget_lanes16(<4 x i16>* %A) nounwind {
;CHECK: vget_lanes16:
;CHECK: vmov.s16
%tmp1 = load <4 x i16>* %A
%tmp2 = extractelement <4 x i16> %tmp1, i32 1
%tmp3 = sext i16 %tmp2 to i32
ret i32 %tmp3
}
define i32 @vget_laneu8(<8 x i8>* %A) nounwind {
;CHECK: vget_laneu8:
;CHECK: vmov.u8
%tmp1 = load <8 x i8>* %A
%tmp2 = extractelement <8 x i8> %tmp1, i32 1
%tmp3 = zext i8 %tmp2 to i32
ret i32 %tmp3
}
define i32 @vget_laneu16(<4 x i16>* %A) nounwind {
;CHECK: vget_laneu16:
;CHECK: vmov.u16
%tmp1 = load <4 x i16>* %A
%tmp2 = extractelement <4 x i16> %tmp1, i32 1
%tmp3 = zext i16 %tmp2 to i32
ret i32 %tmp3
}
; Do a vector add to keep the extraction from being done directly from memory.
define i32 @vget_lanei32(<2 x i32>* %A) nounwind {
;CHECK: vget_lanei32:
;CHECK: vmov.32
%tmp1 = load <2 x i32>* %A
%tmp2 = add <2 x i32> %tmp1, %tmp1
%tmp3 = extractelement <2 x i32> %tmp2, i32 1
ret i32 %tmp3
}
define i32 @vgetQ_lanes8(<16 x i8>* %A) nounwind {
;CHECK: vgetQ_lanes8:
;CHECK: vmov.s8
%tmp1 = load <16 x i8>* %A
%tmp2 = extractelement <16 x i8> %tmp1, i32 1
%tmp3 = sext i8 %tmp2 to i32
ret i32 %tmp3
}
define i32 @vgetQ_lanes16(<8 x i16>* %A) nounwind {
;CHECK: vgetQ_lanes16:
;CHECK: vmov.s16
%tmp1 = load <8 x i16>* %A
%tmp2 = extractelement <8 x i16> %tmp1, i32 1
%tmp3 = sext i16 %tmp2 to i32
ret i32 %tmp3
}
define i32 @vgetQ_laneu8(<16 x i8>* %A) nounwind {
;CHECK: vgetQ_laneu8:
;CHECK: vmov.u8
%tmp1 = load <16 x i8>* %A
%tmp2 = extractelement <16 x i8> %tmp1, i32 1
%tmp3 = zext i8 %tmp2 to i32
ret i32 %tmp3
}
define i32 @vgetQ_laneu16(<8 x i16>* %A) nounwind {
;CHECK: vgetQ_laneu16:
;CHECK: vmov.u16
%tmp1 = load <8 x i16>* %A
%tmp2 = extractelement <8 x i16> %tmp1, i32 1
%tmp3 = zext i16 %tmp2 to i32
ret i32 %tmp3
}
; Do a vector add to keep the extraction from being done directly from memory.
define i32 @vgetQ_lanei32(<4 x i32>* %A) nounwind {
;CHECK: vgetQ_lanei32:
;CHECK: vmov.32
%tmp1 = load <4 x i32>* %A
%tmp2 = add <4 x i32> %tmp1, %tmp1
%tmp3 = extractelement <4 x i32> %tmp2, i32 1
ret i32 %tmp3
}
define arm_aapcs_vfpcc void @test_vget_laneu16() nounwind {
entry:
; CHECK: vmov.u16 r0, d0[1]
%arg0_uint16x4_t = alloca <4 x i16> ; <<4 x i16>*> [#uses=1]
%out_uint16_t = alloca i16 ; <i16*> [#uses=1]
%"alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0]
%0 = load <4 x i16>* %arg0_uint16x4_t, align 8 ; <<4 x i16>> [#uses=1]
%1 = extractelement <4 x i16> %0, i32 1 ; <i16> [#uses=1]
store i16 %1, i16* %out_uint16_t, align 2
br label %return
return: ; preds = %entry
ret void
}
define arm_aapcs_vfpcc void @test_vget_laneu8() nounwind {
entry:
; CHECK: vmov.u8 r0, d0[1]
%arg0_uint8x8_t = alloca <8 x i8> ; <<8 x i8>*> [#uses=1]
%out_uint8_t = alloca i8 ; <i8*> [#uses=1]
%"alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0]
%0 = load <8 x i8>* %arg0_uint8x8_t, align 8 ; <<8 x i8>> [#uses=1]
%1 = extractelement <8 x i8> %0, i32 1 ; <i8> [#uses=1]
store i8 %1, i8* %out_uint8_t, align 1
br label %return
return: ; preds = %entry
ret void
}
define arm_aapcs_vfpcc void @test_vgetQ_laneu16() nounwind {
entry:
; CHECK: vmov.u16 r0, d0[1]
%arg0_uint16x8_t = alloca <8 x i16> ; <<8 x i16>*> [#uses=1]
%out_uint16_t = alloca i16 ; <i16*> [#uses=1]
%"alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0]
%0 = load <8 x i16>* %arg0_uint16x8_t, align 16 ; <<8 x i16>> [#uses=1]
%1 = extractelement <8 x i16> %0, i32 1 ; <i16> [#uses=1]
store i16 %1, i16* %out_uint16_t, align 2
br label %return
return: ; preds = %entry
ret void
}
define arm_aapcs_vfpcc void @test_vgetQ_laneu8() nounwind {
entry:
; CHECK: vmov.u8 r0, d0[1]
%arg0_uint8x16_t = alloca <16 x i8> ; <<16 x i8>*> [#uses=1]
%out_uint8_t = alloca i8 ; <i8*> [#uses=1]
%"alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0]
%0 = load <16 x i8>* %arg0_uint8x16_t, align 16 ; <<16 x i8>> [#uses=1]
%1 = extractelement <16 x i8> %0, i32 1 ; <i8> [#uses=1]
store i8 %1, i8* %out_uint8_t, align 1
br label %return
return: ; preds = %entry
ret void
}
define <8 x i8> @vset_lane8(<8 x i8>* %A, i8 %B) nounwind {
;CHECK: vset_lane8:
;CHECK: vmov.8
%tmp1 = load <8 x i8>* %A
%tmp2 = insertelement <8 x i8> %tmp1, i8 %B, i32 1
ret <8 x i8> %tmp2
}
define <4 x i16> @vset_lane16(<4 x i16>* %A, i16 %B) nounwind {
;CHECK: vset_lane16:
;CHECK: vmov.16
%tmp1 = load <4 x i16>* %A
%tmp2 = insertelement <4 x i16> %tmp1, i16 %B, i32 1
ret <4 x i16> %tmp2
}
define <2 x i32> @vset_lane32(<2 x i32>* %A, i32 %B) nounwind {
;CHECK: vset_lane32:
;CHECK: vmov.32
%tmp1 = load <2 x i32>* %A
%tmp2 = insertelement <2 x i32> %tmp1, i32 %B, i32 1
ret <2 x i32> %tmp2
}
define <16 x i8> @vsetQ_lane8(<16 x i8>* %A, i8 %B) nounwind {
;CHECK: vsetQ_lane8:
;CHECK: vmov.8
%tmp1 = load <16 x i8>* %A
%tmp2 = insertelement <16 x i8> %tmp1, i8 %B, i32 1
ret <16 x i8> %tmp2
}
define <8 x i16> @vsetQ_lane16(<8 x i16>* %A, i16 %B) nounwind {
;CHECK: vsetQ_lane16:
;CHECK: vmov.16
%tmp1 = load <8 x i16>* %A
%tmp2 = insertelement <8 x i16> %tmp1, i16 %B, i32 1
ret <8 x i16> %tmp2
}
define <4 x i32> @vsetQ_lane32(<4 x i32>* %A, i32 %B) nounwind {
;CHECK: vsetQ_lane32:
;CHECK: vmov.32
%tmp1 = load <4 x i32>* %A
%tmp2 = insertelement <4 x i32> %tmp1, i32 %B, i32 1
ret <4 x i32> %tmp2
}
define arm_aapcs_vfpcc <2 x float> @test_vset_lanef32(float %arg0_float32_t, <2 x float> %arg1_float32x2_t) nounwind {
;CHECK: test_vset_lanef32:
;CHECK: fcpys
;CHECK: fcpys
entry:
%0 = insertelement <2 x float> %arg1_float32x2_t, float %arg0_float32_t, i32 1 ; <<2 x float>> [#uses=1]
ret <2 x float> %0
}
|
