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
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme -mattr=+sme2 -mattr=+faminmax -force-streaming -verify-machineinstrs < %s | FileCheck %s
; FAMAX (Multi, x2)
define { <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_max_multi_x2_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2) {
; CHECK-LABEL: multi_vec_max_multi_x2_f16:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z7.d, z4.d
; CHECK-NEXT: mov z5.d, z2.d
; CHECK-NEXT: mov z6.d, z3.d
; CHECK-NEXT: mov z4.d, z1.d
; CHECK-NEXT: famax { z4.h, z5.h }, { z4.h, z5.h }, { z6.h, z7.h }
; CHECK-NEXT: mov z0.d, z4.d
; CHECK-NEXT: mov z1.d, z5.d
; CHECK-NEXT: ret
%res = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x2.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2)
ret { <vscale x 8 x half>, <vscale x 8 x half> } %res
}
define { <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_max_multi_x2_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2) {
; CHECK-LABEL: multi_vec_max_multi_x2_f32:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z7.d, z4.d
; CHECK-NEXT: mov z5.d, z2.d
; CHECK-NEXT: mov z6.d, z3.d
; CHECK-NEXT: mov z4.d, z1.d
; CHECK-NEXT: famax { z4.s, z5.s }, { z4.s, z5.s }, { z6.s, z7.s }
; CHECK-NEXT: mov z0.d, z4.d
; CHECK-NEXT: mov z1.d, z5.d
; CHECK-NEXT: ret
%res = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x2.nxv4f32(<vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2)
ret { <vscale x 4 x float>, <vscale x 4 x float> } %res
}
define { <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_max_multi_x2_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2) {
; CHECK-LABEL: multi_vec_max_multi_x2_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z7.d, z4.d
; CHECK-NEXT: mov z5.d, z2.d
; CHECK-NEXT: mov z6.d, z3.d
; CHECK-NEXT: mov z4.d, z1.d
; CHECK-NEXT: famax { z4.d, z5.d }, { z4.d, z5.d }, { z6.d, z7.d }
; CHECK-NEXT: mov z0.d, z4.d
; CHECK-NEXT: mov z1.d, z5.d
; CHECK-NEXT: ret
%res = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famax.x2.nxv2f64(<vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2)
ret { <vscale x 2 x double>, <vscale x 2 x double> } %res
}
; FAMAX (Multi, x4)
define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }@multi_vec_max_multi_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
; CHECK-LABEL: multi_vec_max_multi_x4_f16:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z30.d, z7.d
; CHECK-NEXT: mov z27.d, z4.d
; CHECK-NEXT: ptrue p0.h
; CHECK-NEXT: mov z29.d, z6.d
; CHECK-NEXT: mov z26.d, z3.d
; CHECK-NEXT: mov z28.d, z5.d
; CHECK-NEXT: mov z25.d, z2.d
; CHECK-NEXT: ld1h { z31.h }, p0/z, [x0]
; CHECK-NEXT: mov z24.d, z1.d
; CHECK-NEXT: famax { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
; CHECK-NEXT: mov z0.d, z24.d
; CHECK-NEXT: mov z1.d, z25.d
; CHECK-NEXT: mov z2.d, z26.d
; CHECK-NEXT: mov z3.d, z27.d
; CHECK-NEXT: ret
%res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famax.x4.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4)
ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
}
define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_max_multi_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
; CHECK-LABEL: multi_vec_max_multi_x4_f32:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z30.d, z7.d
; CHECK-NEXT: mov z27.d, z4.d
; CHECK-NEXT: ptrue p0.s
; CHECK-NEXT: mov z29.d, z6.d
; CHECK-NEXT: mov z26.d, z3.d
; CHECK-NEXT: mov z28.d, z5.d
; CHECK-NEXT: mov z25.d, z2.d
; CHECK-NEXT: ld1w { z31.s }, p0/z, [x0]
; CHECK-NEXT: mov z24.d, z1.d
; CHECK-NEXT: famax { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
; CHECK-NEXT: mov z0.d, z24.d
; CHECK-NEXT: mov z1.d, z25.d
; CHECK-NEXT: mov z2.d, z26.d
; CHECK-NEXT: mov z3.d, z27.d
; CHECK-NEXT: ret
%res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famax.x4.nxv4f32(<vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4)
ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
}
define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_max_multi_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
; CHECK-LABEL: multi_vec_max_multi_x4_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z30.d, z7.d
; CHECK-NEXT: mov z27.d, z4.d
; CHECK-NEXT: ptrue p0.d
; CHECK-NEXT: mov z29.d, z6.d
; CHECK-NEXT: mov z26.d, z3.d
; CHECK-NEXT: mov z28.d, z5.d
; CHECK-NEXT: mov z25.d, z2.d
; CHECK-NEXT: ld1d { z31.d }, p0/z, [x0]
; CHECK-NEXT: mov z24.d, z1.d
; CHECK-NEXT: famax { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
; CHECK-NEXT: mov z0.d, z24.d
; CHECK-NEXT: mov z1.d, z25.d
; CHECK-NEXT: mov z2.d, z26.d
; CHECK-NEXT: mov z3.d, z27.d
; CHECK-NEXT: ret
%res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
@llvm.aarch64.sme.famax.x4.nxv2f64(<vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4,
<vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4)
ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %res
}
; FAMIN (Multi, x2)
define { <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_min_multi_x2_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2) {
; CHECK-LABEL: multi_vec_min_multi_x2_f16:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z7.d, z4.d
; CHECK-NEXT: mov z5.d, z2.d
; CHECK-NEXT: mov z6.d, z3.d
; CHECK-NEXT: mov z4.d, z1.d
; CHECK-NEXT: famin { z4.h, z5.h }, { z4.h, z5.h }, { z6.h, z7.h }
; CHECK-NEXT: mov z0.d, z4.d
; CHECK-NEXT: mov z1.d, z5.d
; CHECK-NEXT: ret
%res = call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sme.famin.x2.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2)
ret { <vscale x 8 x half>, <vscale x 8 x half> } %res
}
define { <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_min_multi_x2_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2) {
; CHECK-LABEL: multi_vec_min_multi_x2_f32:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z7.d, z4.d
; CHECK-NEXT: mov z5.d, z2.d
; CHECK-NEXT: mov z6.d, z3.d
; CHECK-NEXT: mov z4.d, z1.d
; CHECK-NEXT: famin { z4.s, z5.s }, { z4.s, z5.s }, { z6.s, z7.s }
; CHECK-NEXT: mov z0.d, z4.d
; CHECK-NEXT: mov z1.d, z5.d
; CHECK-NEXT: ret
%res = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sme.famin.x2.nxv4f32(<vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2)
ret { <vscale x 4 x float>, <vscale x 4 x float> } %res
}
define { <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_main_multi_x2_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2) {
; CHECK-LABEL: multi_vec_main_multi_x2_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z7.d, z4.d
; CHECK-NEXT: mov z5.d, z2.d
; CHECK-NEXT: mov z6.d, z3.d
; CHECK-NEXT: mov z4.d, z1.d
; CHECK-NEXT: famin { z4.d, z5.d }, { z4.d, z5.d }, { z6.d, z7.d }
; CHECK-NEXT: mov z0.d, z4.d
; CHECK-NEXT: mov z1.d, z5.d
; CHECK-NEXT: ret
%res = call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sme.famin.x2.nxv2f64(<vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2)
ret { <vscale x 2 x double>, <vscale x 2 x double> } %res
}
; FAMIN (Multi, x4)
define { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @multi_vec_min_multi_x4_f16(<vscale x 8 x half> %unused, <vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4, <vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4) {
; CHECK-LABEL: multi_vec_min_multi_x4_f16:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z30.d, z7.d
; CHECK-NEXT: mov z27.d, z4.d
; CHECK-NEXT: ptrue p0.h
; CHECK-NEXT: mov z29.d, z6.d
; CHECK-NEXT: mov z26.d, z3.d
; CHECK-NEXT: mov z28.d, z5.d
; CHECK-NEXT: mov z25.d, z2.d
; CHECK-NEXT: ld1h { z31.h }, p0/z, [x0]
; CHECK-NEXT: mov z24.d, z1.d
; CHECK-NEXT: famin { z24.h - z27.h }, { z24.h - z27.h }, { z28.h - z31.h }
; CHECK-NEXT: mov z0.d, z24.d
; CHECK-NEXT: mov z1.d, z25.d
; CHECK-NEXT: mov z2.d, z26.d
; CHECK-NEXT: mov z3.d, z27.d
; CHECK-NEXT: ret
%res = call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> }
@llvm.aarch64.sme.famin.x4.nxv8f16(<vscale x 8 x half> %zdn1, <vscale x 8 x half> %zdn2, <vscale x 8 x half> %zdn3, <vscale x 8 x half> %zdn4,
<vscale x 8 x half> %zm1, <vscale x 8 x half> %zm2, <vscale x 8 x half> %zm3, <vscale x 8 x half> %zm4)
ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } %res
}
define { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @multi_vec_min_multi_x4_f32(<vscale x 4 x float> %unused, <vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4, <vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4) {
; CHECK-LABEL: multi_vec_min_multi_x4_f32:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z30.d, z7.d
; CHECK-NEXT: mov z27.d, z4.d
; CHECK-NEXT: ptrue p0.s
; CHECK-NEXT: mov z29.d, z6.d
; CHECK-NEXT: mov z26.d, z3.d
; CHECK-NEXT: mov z28.d, z5.d
; CHECK-NEXT: mov z25.d, z2.d
; CHECK-NEXT: ld1w { z31.s }, p0/z, [x0]
; CHECK-NEXT: mov z24.d, z1.d
; CHECK-NEXT: famin { z24.s - z27.s }, { z24.s - z27.s }, { z28.s - z31.s }
; CHECK-NEXT: mov z0.d, z24.d
; CHECK-NEXT: mov z1.d, z25.d
; CHECK-NEXT: mov z2.d, z26.d
; CHECK-NEXT: mov z3.d, z27.d
; CHECK-NEXT: ret
%res = call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> }
@llvm.aarch64.sme.famin.x4.nxv4f32(<vscale x 4 x float> %zdn1, <vscale x 4 x float> %zdn2, <vscale x 4 x float> %zdn3, <vscale x 4 x float> %zdn4,
<vscale x 4 x float> %zm1, <vscale x 4 x float> %zm2, <vscale x 4 x float> %zm3, <vscale x 4 x float> %zm4)
ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } %res
}
define { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @multi_vec_min_multi_x4_f64(<vscale x 2 x double> %unused, <vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4, <vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4) {
; CHECK-LABEL: multi_vec_min_multi_x4_f64:
; CHECK: // %bb.0:
; CHECK-NEXT: mov z30.d, z7.d
; CHECK-NEXT: mov z27.d, z4.d
; CHECK-NEXT: ptrue p0.d
; CHECK-NEXT: mov z29.d, z6.d
; CHECK-NEXT: mov z26.d, z3.d
; CHECK-NEXT: mov z28.d, z5.d
; CHECK-NEXT: mov z25.d, z2.d
; CHECK-NEXT: ld1d { z31.d }, p0/z, [x0]
; CHECK-NEXT: mov z24.d, z1.d
; CHECK-NEXT: famin { z24.d - z27.d }, { z24.d - z27.d }, { z28.d - z31.d }
; CHECK-NEXT: mov z0.d, z24.d
; CHECK-NEXT: mov z1.d, z25.d
; CHECK-NEXT: mov z2.d, z26.d
; CHECK-NEXT: mov z3.d, z27.d
; CHECK-NEXT: ret
%res = call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> }
@llvm.aarch64.sme.famin.x4.nxv2f64(<vscale x 2 x double> %zdn1, <vscale x 2 x double> %zdn2, <vscale x 2 x double> %zdn3, <vscale x 2 x double> %zdn4,
<vscale x 2 x double> %zm1, <vscale x 2 x double> %zm2, <vscale x 2 x double> %zm3, <vscale x 2 x double> %zm4)
ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } %res
}
|
