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
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.2,+gfni | FileCheck %s --check-prefixes=GFNISSE
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+gfni | FileCheck %s --check-prefixes=GFNIAVX,GFNIAVX1OR2,GFNIAVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+gfni | FileCheck %s --check-prefixes=GFNIAVX,GFNIAVX1OR2,GFNIAVX2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl,+gfni | FileCheck %s --check-prefixes=GFNIAVX,GFNIAVX512
;
; 128 Bit Vector Shifts
;
define <16 x i8> @splatconstant_shl_v16i8(<16 x i8> %a) nounwind {
; GFNISSE-LABEL: splatconstant_shl_v16i8:
; GFNISSE: # %bb.0:
; GFNISSE-NEXT: psllw $3, %xmm0
; GFNISSE-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; GFNISSE-NEXT: retq
;
; GFNIAVX-LABEL: splatconstant_shl_v16i8:
; GFNIAVX: # %bb.0:
; GFNIAVX-NEXT: vpsllw $3, %xmm0, %xmm0
; GFNIAVX-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; GFNIAVX-NEXT: retq
%shift = shl <16 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
ret <16 x i8> %shift
}
define <16 x i8> @splatconstant_lshr_v16i8(<16 x i8> %a) nounwind {
; GFNISSE-LABEL: splatconstant_lshr_v16i8:
; GFNISSE: # %bb.0:
; GFNISSE-NEXT: psrlw $7, %xmm0
; GFNISSE-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; GFNISSE-NEXT: retq
;
; GFNIAVX-LABEL: splatconstant_lshr_v16i8:
; GFNIAVX: # %bb.0:
; GFNIAVX-NEXT: vpsrlw $7, %xmm0, %xmm0
; GFNIAVX-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; GFNIAVX-NEXT: retq
%shift = lshr <16 x i8> %a, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
ret <16 x i8> %shift
}
define <16 x i8> @splatconstant_ashr_v16i8(<16 x i8> %a) nounwind {
; GFNISSE-LABEL: splatconstant_ashr_v16i8:
; GFNISSE: # %bb.0:
; GFNISSE-NEXT: psrlw $4, %xmm0
; GFNISSE-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; GFNISSE-NEXT: movdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
; GFNISSE-NEXT: pxor %xmm1, %xmm0
; GFNISSE-NEXT: psubb %xmm1, %xmm0
; GFNISSE-NEXT: retq
;
; GFNIAVX1OR2-LABEL: splatconstant_ashr_v16i8:
; GFNIAVX1OR2: # %bb.0:
; GFNIAVX1OR2-NEXT: vpsrlw $4, %xmm0, %xmm0
; GFNIAVX1OR2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; GFNIAVX1OR2-NEXT: vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
; GFNIAVX1OR2-NEXT: vpxor %xmm1, %xmm0, %xmm0
; GFNIAVX1OR2-NEXT: vpsubb %xmm1, %xmm0, %xmm0
; GFNIAVX1OR2-NEXT: retq
;
; GFNIAVX512-LABEL: splatconstant_ashr_v16i8:
; GFNIAVX512: # %bb.0:
; GFNIAVX512-NEXT: vpsrlw $4, %xmm0, %xmm0
; GFNIAVX512-NEXT: vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
; GFNIAVX512-NEXT: vpternlogq $108, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm0
; GFNIAVX512-NEXT: vpsubb %xmm1, %xmm0, %xmm0
; GFNIAVX512-NEXT: retq
%shift = ashr <16 x i8> %a, <i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4>
ret <16 x i8> %shift
}
;
; 256 Bit Vector Shifts
;
define <32 x i8> @splatconstant_shl_v32i8(<32 x i8> %a) nounwind {
; GFNISSE-LABEL: splatconstant_shl_v32i8:
; GFNISSE: # %bb.0:
; GFNISSE-NEXT: psllw $6, %xmm0
; GFNISSE-NEXT: movdqa {{.*#+}} xmm2 = [192,192,192,192,192,192,192,192,192,192,192,192,192,192,192,192]
; GFNISSE-NEXT: pand %xmm2, %xmm0
; GFNISSE-NEXT: psllw $6, %xmm1
; GFNISSE-NEXT: pand %xmm2, %xmm1
; GFNISSE-NEXT: retq
;
; GFNIAVX1-LABEL: splatconstant_shl_v32i8:
; GFNIAVX1: # %bb.0:
; GFNIAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; GFNIAVX1-NEXT: vpsllw $6, %xmm1, %xmm1
; GFNIAVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [192,192,192,192,192,192,192,192,192,192,192,192,192,192,192,192]
; GFNIAVX1-NEXT: vpand %xmm2, %xmm1, %xmm1
; GFNIAVX1-NEXT: vpsllw $6, %xmm0, %xmm0
; GFNIAVX1-NEXT: vpand %xmm2, %xmm0, %xmm0
; GFNIAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; GFNIAVX1-NEXT: retq
;
; GFNIAVX2-LABEL: splatconstant_shl_v32i8:
; GFNIAVX2: # %bb.0:
; GFNIAVX2-NEXT: vpsllw $6, %ymm0, %ymm0
; GFNIAVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; GFNIAVX2-NEXT: retq
;
; GFNIAVX512-LABEL: splatconstant_shl_v32i8:
; GFNIAVX512: # %bb.0:
; GFNIAVX512-NEXT: vpsllw $6, %ymm0, %ymm0
; GFNIAVX512-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; GFNIAVX512-NEXT: retq
%shift = shl <32 x i8> %a, <i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6>
ret <32 x i8> %shift
}
define <32 x i8> @splatconstant_lshr_v32i8(<32 x i8> %a) nounwind {
; GFNISSE-LABEL: splatconstant_lshr_v32i8:
; GFNISSE: # %bb.0:
; GFNISSE-NEXT: psrlw $1, %xmm0
; GFNISSE-NEXT: movdqa {{.*#+}} xmm2 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
; GFNISSE-NEXT: pand %xmm2, %xmm0
; GFNISSE-NEXT: psrlw $1, %xmm1
; GFNISSE-NEXT: pand %xmm2, %xmm1
; GFNISSE-NEXT: retq
;
; GFNIAVX1-LABEL: splatconstant_lshr_v32i8:
; GFNIAVX1: # %bb.0:
; GFNIAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; GFNIAVX1-NEXT: vpsrlw $1, %xmm1, %xmm1
; GFNIAVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
; GFNIAVX1-NEXT: vpand %xmm2, %xmm1, %xmm1
; GFNIAVX1-NEXT: vpsrlw $1, %xmm0, %xmm0
; GFNIAVX1-NEXT: vpand %xmm2, %xmm0, %xmm0
; GFNIAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; GFNIAVX1-NEXT: retq
;
; GFNIAVX2-LABEL: splatconstant_lshr_v32i8:
; GFNIAVX2: # %bb.0:
; GFNIAVX2-NEXT: vpsrlw $1, %ymm0, %ymm0
; GFNIAVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; GFNIAVX2-NEXT: retq
;
; GFNIAVX512-LABEL: splatconstant_lshr_v32i8:
; GFNIAVX512: # %bb.0:
; GFNIAVX512-NEXT: vpsrlw $1, %ymm0, %ymm0
; GFNIAVX512-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; GFNIAVX512-NEXT: retq
%shift = lshr <32 x i8> %a, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
ret <32 x i8> %shift
}
define <32 x i8> @splatconstant_ashr_v32i8(<32 x i8> %a) nounwind {
; GFNISSE-LABEL: splatconstant_ashr_v32i8:
; GFNISSE: # %bb.0:
; GFNISSE-NEXT: psrlw $2, %xmm0
; GFNISSE-NEXT: movdqa {{.*#+}} xmm2 = [63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63]
; GFNISSE-NEXT: pand %xmm2, %xmm0
; GFNISSE-NEXT: movdqa {{.*#+}} xmm3 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
; GFNISSE-NEXT: pxor %xmm3, %xmm0
; GFNISSE-NEXT: psubb %xmm3, %xmm0
; GFNISSE-NEXT: psrlw $2, %xmm1
; GFNISSE-NEXT: pand %xmm2, %xmm1
; GFNISSE-NEXT: pxor %xmm3, %xmm1
; GFNISSE-NEXT: psubb %xmm3, %xmm1
; GFNISSE-NEXT: retq
;
; GFNIAVX1-LABEL: splatconstant_ashr_v32i8:
; GFNIAVX1: # %bb.0:
; GFNIAVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; GFNIAVX1-NEXT: vpsrlw $2, %xmm1, %xmm1
; GFNIAVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [63,63,63,63,63,63,63,63,63,63,63,63,63,63,63,63]
; GFNIAVX1-NEXT: vpand %xmm2, %xmm1, %xmm1
; GFNIAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
; GFNIAVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
; GFNIAVX1-NEXT: vpsubb %xmm3, %xmm1, %xmm1
; GFNIAVX1-NEXT: vpsrlw $2, %xmm0, %xmm0
; GFNIAVX1-NEXT: vpand %xmm2, %xmm0, %xmm0
; GFNIAVX1-NEXT: vpxor %xmm3, %xmm0, %xmm0
; GFNIAVX1-NEXT: vpsubb %xmm3, %xmm0, %xmm0
; GFNIAVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; GFNIAVX1-NEXT: retq
;
; GFNIAVX2-LABEL: splatconstant_ashr_v32i8:
; GFNIAVX2: # %bb.0:
; GFNIAVX2-NEXT: vpsrlw $2, %ymm0, %ymm0
; GFNIAVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; GFNIAVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
; GFNIAVX2-NEXT: vpxor %ymm1, %ymm0, %ymm0
; GFNIAVX2-NEXT: vpsubb %ymm1, %ymm0, %ymm0
; GFNIAVX2-NEXT: retq
;
; GFNIAVX512-LABEL: splatconstant_ashr_v32i8:
; GFNIAVX512: # %bb.0:
; GFNIAVX512-NEXT: vpsrlw $2, %ymm0, %ymm0
; GFNIAVX512-NEXT: vmovdqa {{.*#+}} ymm1 = [32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32,32]
; GFNIAVX512-NEXT: vpternlogq $108, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm1, %ymm0
; GFNIAVX512-NEXT: vpsubb %ymm1, %ymm0, %ymm0
; GFNIAVX512-NEXT: retq
%shift = ashr <32 x i8> %a, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
ret <32 x i8> %shift
}
;
; 512 Bit Vector Shifts
;
define <64 x i8> @splatconstant_shl_v64i8(<64 x i8> %a) nounwind {
; GFNISSE-LABEL: splatconstant_shl_v64i8:
; GFNISSE: # %bb.0:
; GFNISSE-NEXT: psllw $5, %xmm0
; GFNISSE-NEXT: movdqa {{.*#+}} xmm4 = [224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224]
; GFNISSE-NEXT: pand %xmm4, %xmm0
; GFNISSE-NEXT: psllw $5, %xmm1
; GFNISSE-NEXT: pand %xmm4, %xmm1
; GFNISSE-NEXT: psllw $5, %xmm2
; GFNISSE-NEXT: pand %xmm4, %xmm2
; GFNISSE-NEXT: psllw $5, %xmm3
; GFNISSE-NEXT: pand %xmm4, %xmm3
; GFNISSE-NEXT: retq
;
; GFNIAVX1-LABEL: splatconstant_shl_v64i8:
; GFNIAVX1: # %bb.0:
; GFNIAVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
; GFNIAVX1-NEXT: vpsllw $5, %xmm2, %xmm2
; GFNIAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224]
; GFNIAVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpsllw $5, %xmm0, %xmm0
; GFNIAVX1-NEXT: vpand %xmm3, %xmm0, %xmm0
; GFNIAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; GFNIAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; GFNIAVX1-NEXT: vpsllw $5, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpsllw $5, %xmm1, %xmm1
; GFNIAVX1-NEXT: vpand %xmm3, %xmm1, %xmm1
; GFNIAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; GFNIAVX1-NEXT: retq
;
; GFNIAVX2-LABEL: splatconstant_shl_v64i8:
; GFNIAVX2: # %bb.0:
; GFNIAVX2-NEXT: vpsllw $5, %ymm0, %ymm0
; GFNIAVX2-NEXT: vmovdqa {{.*#+}} ymm2 = [224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224,224]
; GFNIAVX2-NEXT: vpand %ymm2, %ymm0, %ymm0
; GFNIAVX2-NEXT: vpsllw $5, %ymm1, %ymm1
; GFNIAVX2-NEXT: vpand %ymm2, %ymm1, %ymm1
; GFNIAVX2-NEXT: retq
;
; GFNIAVX512-LABEL: splatconstant_shl_v64i8:
; GFNIAVX512: # %bb.0:
; GFNIAVX512-NEXT: vpsllw $5, %zmm0, %zmm0
; GFNIAVX512-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
; GFNIAVX512-NEXT: retq
%shift = shl <64 x i8> %a, <i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5>
ret <64 x i8> %shift
}
define <64 x i8> @splatconstant_lshr_v64i8(<64 x i8> %a) nounwind {
; GFNISSE-LABEL: splatconstant_lshr_v64i8:
; GFNISSE: # %bb.0:
; GFNISSE-NEXT: psrlw $7, %xmm0
; GFNISSE-NEXT: movdqa {{.*#+}} xmm4 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
; GFNISSE-NEXT: pand %xmm4, %xmm0
; GFNISSE-NEXT: psrlw $7, %xmm1
; GFNISSE-NEXT: pand %xmm4, %xmm1
; GFNISSE-NEXT: psrlw $7, %xmm2
; GFNISSE-NEXT: pand %xmm4, %xmm2
; GFNISSE-NEXT: psrlw $7, %xmm3
; GFNISSE-NEXT: pand %xmm4, %xmm3
; GFNISSE-NEXT: retq
;
; GFNIAVX1-LABEL: splatconstant_lshr_v64i8:
; GFNIAVX1: # %bb.0:
; GFNIAVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
; GFNIAVX1-NEXT: vpsrlw $7, %xmm2, %xmm2
; GFNIAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
; GFNIAVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpsrlw $7, %xmm0, %xmm0
; GFNIAVX1-NEXT: vpand %xmm3, %xmm0, %xmm0
; GFNIAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; GFNIAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; GFNIAVX1-NEXT: vpsrlw $7, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpsrlw $7, %xmm1, %xmm1
; GFNIAVX1-NEXT: vpand %xmm3, %xmm1, %xmm1
; GFNIAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; GFNIAVX1-NEXT: retq
;
; GFNIAVX2-LABEL: splatconstant_lshr_v64i8:
; GFNIAVX2: # %bb.0:
; GFNIAVX2-NEXT: vpsrlw $7, %ymm0, %ymm0
; GFNIAVX2-NEXT: vmovdqa {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
; GFNIAVX2-NEXT: vpand %ymm2, %ymm0, %ymm0
; GFNIAVX2-NEXT: vpsrlw $7, %ymm1, %ymm1
; GFNIAVX2-NEXT: vpand %ymm2, %ymm1, %ymm1
; GFNIAVX2-NEXT: retq
;
; GFNIAVX512-LABEL: splatconstant_lshr_v64i8:
; GFNIAVX512: # %bb.0:
; GFNIAVX512-NEXT: vpsrlw $7, %zmm0, %zmm0
; GFNIAVX512-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
; GFNIAVX512-NEXT: retq
%shift = lshr <64 x i8> %a, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
ret <64 x i8> %shift
}
define <64 x i8> @splatconstant_ashr_v64i8(<64 x i8> %a) nounwind {
; GFNISSE-LABEL: splatconstant_ashr_v64i8:
; GFNISSE: # %bb.0:
; GFNISSE-NEXT: psrlw $1, %xmm0
; GFNISSE-NEXT: movdqa {{.*#+}} xmm4 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
; GFNISSE-NEXT: pand %xmm4, %xmm0
; GFNISSE-NEXT: movdqa {{.*#+}} xmm5 = [64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64]
; GFNISSE-NEXT: pxor %xmm5, %xmm0
; GFNISSE-NEXT: psubb %xmm5, %xmm0
; GFNISSE-NEXT: psrlw $1, %xmm1
; GFNISSE-NEXT: pand %xmm4, %xmm1
; GFNISSE-NEXT: pxor %xmm5, %xmm1
; GFNISSE-NEXT: psubb %xmm5, %xmm1
; GFNISSE-NEXT: psrlw $1, %xmm2
; GFNISSE-NEXT: pand %xmm4, %xmm2
; GFNISSE-NEXT: pxor %xmm5, %xmm2
; GFNISSE-NEXT: psubb %xmm5, %xmm2
; GFNISSE-NEXT: psrlw $1, %xmm3
; GFNISSE-NEXT: pand %xmm4, %xmm3
; GFNISSE-NEXT: pxor %xmm5, %xmm3
; GFNISSE-NEXT: psubb %xmm5, %xmm3
; GFNISSE-NEXT: retq
;
; GFNIAVX1-LABEL: splatconstant_ashr_v64i8:
; GFNIAVX1: # %bb.0:
; GFNIAVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
; GFNIAVX1-NEXT: vpsrlw $1, %xmm2, %xmm2
; GFNIAVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
; GFNIAVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; GFNIAVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64]
; GFNIAVX1-NEXT: vpxor %xmm4, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpsubb %xmm4, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpsrlw $1, %xmm0, %xmm0
; GFNIAVX1-NEXT: vpand %xmm3, %xmm0, %xmm0
; GFNIAVX1-NEXT: vpxor %xmm4, %xmm0, %xmm0
; GFNIAVX1-NEXT: vpsubb %xmm4, %xmm0, %xmm0
; GFNIAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; GFNIAVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; GFNIAVX1-NEXT: vpsrlw $1, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpand %xmm3, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpxor %xmm4, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpsubb %xmm4, %xmm2, %xmm2
; GFNIAVX1-NEXT: vpsrlw $1, %xmm1, %xmm1
; GFNIAVX1-NEXT: vpand %xmm3, %xmm1, %xmm1
; GFNIAVX1-NEXT: vpxor %xmm4, %xmm1, %xmm1
; GFNIAVX1-NEXT: vpsubb %xmm4, %xmm1, %xmm1
; GFNIAVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; GFNIAVX1-NEXT: retq
;
; GFNIAVX2-LABEL: splatconstant_ashr_v64i8:
; GFNIAVX2: # %bb.0:
; GFNIAVX2-NEXT: vpsrlw $1, %ymm0, %ymm0
; GFNIAVX2-NEXT: vmovdqa {{.*#+}} ymm2 = [127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127,127]
; GFNIAVX2-NEXT: vpand %ymm2, %ymm0, %ymm0
; GFNIAVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64]
; GFNIAVX2-NEXT: vpxor %ymm3, %ymm0, %ymm0
; GFNIAVX2-NEXT: vpsubb %ymm3, %ymm0, %ymm0
; GFNIAVX2-NEXT: vpsrlw $1, %ymm1, %ymm1
; GFNIAVX2-NEXT: vpand %ymm2, %ymm1, %ymm1
; GFNIAVX2-NEXT: vpxor %ymm3, %ymm1, %ymm1
; GFNIAVX2-NEXT: vpsubb %ymm3, %ymm1, %ymm1
; GFNIAVX2-NEXT: retq
;
; GFNIAVX512-LABEL: splatconstant_ashr_v64i8:
; GFNIAVX512: # %bb.0:
; GFNIAVX512-NEXT: vpsrlw $1, %zmm0, %zmm0
; GFNIAVX512-NEXT: vmovdqa64 {{.*#+}} zmm1 = [64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64,64]
; GFNIAVX512-NEXT: vpternlogq $108, {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm1, %zmm0
; GFNIAVX512-NEXT: vpsubb %zmm1, %zmm0, %zmm0
; GFNIAVX512-NEXT: retq
%shift = ashr <64 x i8> %a, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
ret <64 x i8> %shift
}
|
