v256_intrinsics_x86.h 18.9 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
/*
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 */

#ifndef _V256_INTRINSICS_H
#define _V256_INTRINSICS_H

#if !defined(__AVX2__)

#include "./v256_intrinsics_v128.h"

#else

// The _m256i type seems to cause problems for g++'s mangling prior to
// version 5, but adding -fabi-version=0 fixes this.
23
24
#if !defined(__clang__) && defined(__GNUC__) && __GNUC__ < 5 && \
    defined(__AVX2__) && defined(__cplusplus)
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
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
#pragma GCC optimize "-fabi-version=0"
#endif

#include <immintrin.h>
#include "./v128_intrinsics_x86.h"

typedef __m256i v256;

SIMD_INLINE uint32_t v256_low_u32(v256 a) {
  return (uint32_t)_mm_cvtsi128_si32(_mm256_extracti128_si256(a, 0));
}

SIMD_INLINE v64 v256_low_v64(v256 a) {
  return _mm_unpacklo_epi64(_mm256_extracti128_si256(a, 0), v64_zero());
}

SIMD_INLINE v128 v256_low_v128(v256 a) {
  return _mm256_extracti128_si256(a, 0);
}

SIMD_INLINE v128 v256_high_v128(v256 a) {
  return _mm256_extracti128_si256(a, 1);
}

SIMD_INLINE v256 v256_from_v128(v128 a, v128 b) {
  // gcc seems to be missing _mm256_set_m128i()
  return _mm256_insertf128_si256(
      _mm256_insertf128_si256(_mm256_setzero_si256(), b, 0), a, 1);
}

SIMD_INLINE v256 v256_from_v64(v64 a, v64 b, v64 c, v64 d) {
  return v256_from_v128(v128_from_v64(a, b), v128_from_v64(c, d));
}

SIMD_INLINE v256 v256_from_64(uint64_t a, uint64_t b, uint64_t c, uint64_t d) {
  return v256_from_v128(v128_from_64(a, b), v128_from_64(c, d));
}

SIMD_INLINE v256 v256_load_aligned(const void *p) {
  return _mm256_load_si256((const __m256i *)p);
}

SIMD_INLINE v256 v256_load_unaligned(const void *p) {
  return _mm256_loadu_si256((const __m256i *)p);
}

SIMD_INLINE void v256_store_aligned(void *p, v256 a) {
  _mm256_store_si256((__m256i *)p, a);
}

SIMD_INLINE void v256_store_unaligned(void *p, v256 a) {
  _mm256_storeu_si256((__m256i *)p, a);
}

SIMD_INLINE v256 v256_zero() { return _mm256_setzero_si256(); }

SIMD_INLINE v256 v256_dup_8(uint8_t x) { return _mm256_set1_epi8(x); }

SIMD_INLINE v256 v256_dup_16(uint16_t x) { return _mm256_set1_epi16(x); }

SIMD_INLINE v256 v256_dup_32(uint32_t x) { return _mm256_set1_epi32(x); }

SIMD_INLINE v256 v256_add_8(v256 a, v256 b) { return _mm256_add_epi8(a, b); }

SIMD_INLINE v256 v256_add_16(v256 a, v256 b) { return _mm256_add_epi16(a, b); }

SIMD_INLINE v256 v256_sadd_s16(v256 a, v256 b) {
  return _mm256_adds_epi16(a, b);
}

SIMD_INLINE v256 v256_add_32(v256 a, v256 b) { return _mm256_add_epi32(a, b); }

SIMD_INLINE v256 v256_padd_s16(v256 a) {
  return _mm256_madd_epi16(a, _mm256_set1_epi16(1));
}

SIMD_INLINE v256 v256_sub_8(v256 a, v256 b) { return _mm256_sub_epi8(a, b); }

SIMD_INLINE v256 v256_ssub_u8(v256 a, v256 b) { return _mm256_subs_epu8(a, b); }

SIMD_INLINE v256 v256_ssub_s8(v256 a, v256 b) { return _mm256_subs_epi8(a, b); }

SIMD_INLINE v256 v256_sub_16(v256 a, v256 b) { return _mm256_sub_epi16(a, b); }

SIMD_INLINE v256 v256_ssub_s16(v256 a, v256 b) {
  return _mm256_subs_epi16(a, b);
}

SIMD_INLINE v256 v256_sub_32(v256 a, v256 b) { return _mm256_sub_epi32(a, b); }

SIMD_INLINE v256 v256_abs_s16(v256 a) { return _mm256_abs_epi16(a); }

// AVX doesn't have the direct intrinsics to zip/unzip 8, 16, 32 bit
// lanes of lower or upper halves of a 256bit vector because the
// unpack/pack intrinsics operate on the 256 bit input vector as 2
// independent 128 bit vectors.
SIMD_INLINE v256 v256_ziplo_8(v256 a, v256 b) {
  return v256_from_v128(v128_ziphi_8(v256_low_v128(a), v256_low_v128(b)),
                        v128_ziplo_8(v256_low_v128(a), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_ziphi_8(v256 a, v256 b) {
  return v256_from_v128(v128_ziphi_8(v256_high_v128(a), v256_high_v128(b)),
                        v128_ziplo_8(v256_high_v128(a), v256_high_v128(b)));
}

SIMD_INLINE v256 v256_ziplo_16(v256 a, v256 b) {
  return v256_from_v128(v128_ziphi_16(v256_low_v128(a), v256_low_v128(b)),
                        v128_ziplo_16(v256_low_v128(a), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_ziphi_16(v256 a, v256 b) {
  return v256_from_v128(v128_ziphi_16(v256_high_v128(a), v256_high_v128(b)),
                        v128_ziplo_16(v256_high_v128(a), v256_high_v128(b)));
}

SIMD_INLINE v256 v256_ziplo_32(v256 a, v256 b) {
  return v256_from_v128(v128_ziphi_32(v256_low_v128(a), v256_low_v128(b)),
                        v128_ziplo_32(v256_low_v128(a), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_ziphi_32(v256 a, v256 b) {
  return v256_from_v128(v128_ziphi_32(v256_high_v128(a), v256_high_v128(b)),
                        v128_ziplo_32(v256_high_v128(a), v256_high_v128(b)));
}

SIMD_INLINE v256 v256_ziplo_64(v256 a, v256 b) {
  return v256_from_v128(v128_ziphi_64(v256_low_v128(a), v256_low_v128(b)),
                        v128_ziplo_64(v256_low_v128(a), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_ziphi_64(v256 a, v256 b) {
  return v256_from_v128(v128_ziphi_64(v256_high_v128(a), v256_high_v128(b)),
                        v128_ziplo_64(v256_high_v128(a), v256_high_v128(b)));
}

SIMD_INLINE v256 v256_ziplo_128(v256 a, v256 b) {
  return v256_from_v128(v256_low_v128(a), v256_low_v128(b));
}

SIMD_INLINE v256 v256_ziphi_128(v256 a, v256 b) {
  return v256_from_v128(v256_high_v128(a), v256_high_v128(b));
}

SIMD_INLINE v256 v256_zip_8(v128 a, v128 b) {
  return v256_from_v128(v128_ziphi_8(a, b), v128_ziplo_8(a, b));
}

SIMD_INLINE v256 v256_zip_16(v128 a, v128 b) {
  return v256_from_v128(v128_ziphi_16(a, b), v128_ziplo_16(a, b));
}

SIMD_INLINE v256 v256_zip_32(v128 a, v128 b) {
  return v256_from_v128(v128_ziphi_32(a, b), v128_ziplo_32(a, b));
}

SIMD_INLINE v256 v256_unziplo_8(v256 a, v256 b) {
  return v256_from_v128(v128_unziplo_8(v256_high_v128(a), v256_low_v128(a)),
                        v128_unziplo_8(v256_high_v128(b), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_unziphi_8(v256 a, v256 b) {
  return v256_from_v128(v128_unziphi_8(v256_high_v128(a), v256_low_v128(a)),
                        v128_unziphi_8(v256_high_v128(b), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_unziplo_16(v256 a, v256 b) {
  return v256_from_v128(v128_unziplo_16(v256_high_v128(a), v256_low_v128(a)),
                        v128_unziplo_16(v256_high_v128(b), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_unziphi_16(v256 a, v256 b) {
  return v256_from_v128(v128_unziphi_16(v256_high_v128(a), v256_low_v128(a)),
                        v128_unziphi_16(v256_high_v128(b), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_unziplo_32(v256 a, v256 b) {
  return v256_from_v128(v128_unziplo_32(v256_high_v128(a), v256_low_v128(a)),
                        v128_unziplo_32(v256_high_v128(b), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_unziphi_32(v256 a, v256 b) {
  return v256_from_v128(v128_unziphi_32(v256_high_v128(a), v256_low_v128(a)),
                        v128_unziphi_32(v256_high_v128(b), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_unpack_u8_s16(v128 a) {
  return v256_from_v128(v128_unpackhi_u8_s16(a), v128_unpacklo_u8_s16(a));
}

SIMD_INLINE v256 v256_unpacklo_u8_s16(v256 a) {
  return v256_from_v128(v128_unpackhi_u8_s16(v256_low_v128(a)),
                        v128_unpacklo_u8_s16(v256_low_v128(a)));
}

SIMD_INLINE v256 v256_unpackhi_u8_s16(v256 a) {
  return v256_from_v128(v128_unpackhi_u8_s16(v256_high_v128(a)),
                        v128_unpacklo_u8_s16(v256_high_v128(a)));
}

SIMD_INLINE v256 v256_pack_s32_s16(v256 a, v256 b) {
  return v256_from_v128(v128_pack_s32_s16(v256_high_v128(a), v256_low_v128(a)),
                        v128_pack_s32_s16(v256_high_v128(b), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_pack_s16_u8(v256 a, v256 b) {
  return v256_from_v128(v128_pack_s16_u8(v256_high_v128(a), v256_low_v128(a)),
                        v128_pack_s16_u8(v256_high_v128(b), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_pack_s16_s8(v256 a, v256 b) {
  return v256_from_v128(v128_pack_s16_s8(v256_high_v128(a), v256_low_v128(a)),
                        v128_pack_s16_s8(v256_high_v128(b), v256_low_v128(b)));
}

SIMD_INLINE v256 v256_unpack_u16_s32(v128 a) {
  return v256_from_v128(v128_unpackhi_u16_s32(a), v128_unpacklo_u16_s32(a));
}

SIMD_INLINE v256 v256_unpack_s16_s32(v128 a) {
  return v256_from_v128(v128_unpackhi_s16_s32(a), v128_unpacklo_s16_s32(a));
}

SIMD_INLINE v256 v256_unpacklo_u16_s32(v256 a) {
  return v256_from_v128(v128_unpackhi_u16_s32(v256_low_v128(a)),
                        v128_unpacklo_u16_s32(v256_low_v128(a)));
}

SIMD_INLINE v256 v256_unpacklo_s16_s32(v256 a) {
  return v256_from_v128(v128_unpackhi_s16_s32(v256_low_v128(a)),
                        v128_unpacklo_s16_s32(v256_low_v128(a)));
}

SIMD_INLINE v256 v256_unpackhi_u16_s32(v256 a) {
  return v256_from_v128(v128_unpackhi_u16_s32(v256_high_v128(a)),
                        v128_unpacklo_u16_s32(v256_high_v128(a)));
}

SIMD_INLINE v256 v256_unpackhi_s16_s32(v256 a) {
  return v256_from_v128(v128_unpackhi_s16_s32(v256_high_v128(a)),
                        v128_unpacklo_s16_s32(v256_high_v128(a)));
}
SIMD_INLINE v256 v256_shuffle_8(v256 a, v256 pattern) {
  v128 c16 = v128_dup_8(16);
  v128 hi = v256_high_v128(pattern);
  v128 lo = v256_low_v128(pattern);
  v128 maskhi = v128_cmplt_s8(hi, c16);
  v128 masklo = v128_cmplt_s8(lo, c16);
  return v256_from_v128(
      v128_or(v128_and(v128_shuffle_8(v256_low_v128(a), hi), maskhi),
              v128_andn(v128_shuffle_8(v256_high_v128(a), v128_sub_8(hi, c16)),
                        maskhi)),
      v128_or(v128_and(v128_shuffle_8(v256_low_v128(a), lo), masklo),
              v128_andn(v128_shuffle_8(v256_high_v128(a), v128_sub_8(lo, c16)),
                        masklo)));
}

SIMD_INLINE v256 v256_pshuffle_8(v256 a, v256 pattern) {
  return _mm256_shuffle_epi8(a, pattern);
}

SIMD_INLINE int64_t v256_dotp_s16(v256 a, v256 b) {
  v256 r = _mm256_madd_epi16(a, b);
#if defined(__x86_64__)
  v128 t;
  r = _mm256_add_epi64(_mm256_cvtepi32_epi64(v256_high_v128(r)),
                       _mm256_cvtepi32_epi64(v256_low_v128(r)));
  t = v256_low_v128(_mm256_add_epi64(
      r, _mm256_permute2x128_si256(r, r, _MM_SHUFFLE(2, 0, 0, 1))));
  return _mm_cvtsi128_si64(_mm_add_epi64(t, _mm_srli_si128(t, 8)));
#else
  v128 l = v256_low_v128(r);
  v128 h = v256_high_v128(r);
  return (int64_t)_mm_cvtsi128_si32(l) +
         (int64_t)_mm_cvtsi128_si32(_mm_srli_si128(l, 4)) +
         (int64_t)_mm_cvtsi128_si32(_mm_srli_si128(l, 8)) +
         (int64_t)_mm_cvtsi128_si32(_mm_srli_si128(l, 12)) +
         (int64_t)_mm_cvtsi128_si32(h) +
         (int64_t)_mm_cvtsi128_si32(_mm_srli_si128(h, 4)) +
         (int64_t)_mm_cvtsi128_si32(_mm_srli_si128(h, 8)) +
         (int64_t)_mm_cvtsi128_si32(_mm_srli_si128(h, 12));
#endif
}

SIMD_INLINE uint64_t v256_hadd_u8(v256 a) {
  v256 t = _mm256_sad_epu8(a, _mm256_setzero_si256());
  v128 lo = v256_low_v128(t);
  v128 hi = v256_high_v128(t);
  lo = v128_add_32(lo, hi);
  return v64_low_u32(v128_low_v64(lo)) + v128_low_u32(v128_high_v64(lo));
}

typedef v256 sad256_internal;

SIMD_INLINE sad256_internal v256_sad_u8_init() {
  return _mm256_setzero_si256();
}

/* Implementation dependent return value.  Result must be finalised with
   v256_sad_sum().
   The result for more than 32 v256_sad_u8() calls is undefined. */
SIMD_INLINE sad256_internal v256_sad_u8(sad256_internal s, v256 a, v256 b) {
  return _mm256_add_epi64(s, _mm256_sad_epu8(a, b));
}

SIMD_INLINE uint32_t v256_sad_u8_sum(sad256_internal s) {
  v256 t = _mm256_add_epi32(s, _mm256_unpackhi_epi64(s, s));
  return v128_low_u32(_mm_add_epi32(v256_high_v128(t), v256_low_v128(t)));
}

typedef v256 ssd256_internal;

SIMD_INLINE ssd256_internal v256_ssd_u8_init() {
  return _mm256_setzero_si256();
}

/* Implementation dependent return value.  Result must be finalised with
 * v256_ssd_sum(). */
SIMD_INLINE ssd256_internal v256_ssd_u8(ssd256_internal s, v256 a, v256 b) {
  v256 l = _mm256_sub_epi16(_mm256_unpacklo_epi8(a, _mm256_setzero_si256()),
                            _mm256_unpacklo_epi8(b, _mm256_setzero_si256()));
  v256 h = _mm256_sub_epi16(_mm256_unpackhi_epi8(a, _mm256_setzero_si256()),
                            _mm256_unpackhi_epi8(b, _mm256_setzero_si256()));
  v256 rl = _mm256_madd_epi16(l, l);
  v256 rh = _mm256_madd_epi16(h, h);
  v128 c = _mm_cvtsi32_si128(32);
  rl = _mm256_add_epi32(rl, _mm256_srli_si256(rl, 8));
  rl = _mm256_add_epi32(rl, _mm256_srli_si256(rl, 4));
  rh = _mm256_add_epi32(rh, _mm256_srli_si256(rh, 8));
  rh = _mm256_add_epi32(rh, _mm256_srli_si256(rh, 4));
  return _mm256_add_epi64(
      s,
      _mm256_srl_epi64(_mm256_sll_epi64(_mm256_unpacklo_epi64(rl, rh), c), c));
}

SIMD_INLINE uint32_t v256_ssd_u8_sum(ssd256_internal s) {
  v256 t = _mm256_add_epi32(s, _mm256_unpackhi_epi64(s, s));
  return v128_low_u32(_mm_add_epi32(v256_high_v128(t), v256_low_v128(t)));
}

SIMD_INLINE v256 v256_or(v256 a, v256 b) { return _mm256_or_si256(a, b); }

SIMD_INLINE v256 v256_xor(v256 a, v256 b) { return _mm256_xor_si256(a, b); }

SIMD_INLINE v256 v256_and(v256 a, v256 b) { return _mm256_and_si256(a, b); }

SIMD_INLINE v256 v256_andn(v256 a, v256 b) { return _mm256_andnot_si256(b, a); }

SIMD_INLINE v256 v256_mul_s16(v64 a, v64 b) {
  v128 lo_bits = v128_mullo_s16(a, b);
  v128 hi_bits = v128_mulhi_s16(a, b);
  return v256_from_v128(v128_ziphi_16(hi_bits, lo_bits),
                        v128_ziplo_16(hi_bits, lo_bits));
}

SIMD_INLINE v256 v256_mullo_s16(v256 a, v256 b) {
  return _mm256_mullo_epi16(a, b);
}

SIMD_INLINE v256 v256_mulhi_s16(v256 a, v256 b) {
  return _mm256_mulhi_epi16(a, b);
}

SIMD_INLINE v256 v256_mullo_s32(v256 a, v256 b) {
  return _mm256_mullo_epi32(a, b);
}

SIMD_INLINE v256 v256_madd_s16(v256 a, v256 b) {
  return _mm256_madd_epi16(a, b);
}

SIMD_INLINE v256 v256_madd_us8(v256 a, v256 b) {
  return _mm256_maddubs_epi16(a, b);
}

SIMD_INLINE v256 v256_avg_u8(v256 a, v256 b) { return _mm256_avg_epu8(a, b); }

SIMD_INLINE v256 v256_rdavg_u8(v256 a, v256 b) {
  return _mm256_sub_epi8(
      _mm256_avg_epu8(a, b),
      _mm256_and_si256(_mm256_xor_si256(a, b), v256_dup_8(1)));
}

SIMD_INLINE v256 v256_avg_u16(v256 a, v256 b) { return _mm256_avg_epu16(a, b); }

SIMD_INLINE v256 v256_min_u8(v256 a, v256 b) { return _mm256_min_epu8(a, b); }

SIMD_INLINE v256 v256_max_u8(v256 a, v256 b) { return _mm256_max_epu8(a, b); }

SIMD_INLINE v256 v256_min_s8(v256 a, v256 b) { return _mm256_min_epi8(a, b); }

SIMD_INLINE v256 v256_max_s8(v256 a, v256 b) { return _mm256_max_epi8(a, b); }

SIMD_INLINE v256 v256_min_s16(v256 a, v256 b) { return _mm256_min_epi16(a, b); }

SIMD_INLINE v256 v256_max_s16(v256 a, v256 b) { return _mm256_max_epi16(a, b); }

SIMD_INLINE v256 v256_cmpgt_s8(v256 a, v256 b) {
  return _mm256_cmpgt_epi8(a, b);
}

SIMD_INLINE v256 v256_cmplt_s8(v256 a, v256 b) {
  return v256_andn(_mm256_cmpgt_epi8(b, a), _mm256_cmpeq_epi8(b, a));
}

SIMD_INLINE v256 v256_cmpeq_8(v256 a, v256 b) {
  return _mm256_cmpeq_epi8(a, b);
}

SIMD_INLINE v256 v256_cmpgt_s16(v256 a, v256 b) {
  return _mm256_cmpgt_epi16(a, b);
}

SIMD_INLINE v256 v256_cmplt_s16(v256 a, v256 b) {
  return v256_andn(_mm256_cmpgt_epi16(b, a), _mm256_cmpeq_epi16(b, a));
}

SIMD_INLINE v256 v256_cmpeq_16(v256 a, v256 b) {
  return _mm256_cmpeq_epi16(a, b);
}

SIMD_INLINE v256 v256_shl_8(v256 a, unsigned int c) {
  return _mm256_and_si256(_mm256_set1_epi8((uint8_t)(0xff << c)),
                          _mm256_sll_epi16(a, _mm_cvtsi32_si128(c)));
}

SIMD_INLINE v256 v256_shr_u8(v256 a, unsigned int c) {
  return _mm256_and_si256(_mm256_set1_epi8(0xff >> c),
                          _mm256_srl_epi16(a, _mm_cvtsi32_si128(c)));
}

SIMD_INLINE v256 v256_shr_s8(v256 a, unsigned int c) {
  __m128i x = _mm_cvtsi32_si128(c + 8);
  return _mm256_packs_epi16(_mm256_sra_epi16(_mm256_unpacklo_epi8(a, a), x),
                            _mm256_sra_epi16(_mm256_unpackhi_epi8(a, a), x));
}

SIMD_INLINE v256 v256_shl_16(v256 a, unsigned int c) {
  return _mm256_sll_epi16(a, _mm_cvtsi32_si128(c));
}

SIMD_INLINE v256 v256_shr_u16(v256 a, unsigned int c) {
  return _mm256_srl_epi16(a, _mm_cvtsi32_si128(c));
}

SIMD_INLINE v256 v256_shr_s16(v256 a, unsigned int c) {
  return _mm256_sra_epi16(a, _mm_cvtsi32_si128(c));
}

SIMD_INLINE v256 v256_shl_32(v256 a, unsigned int c) {
  return _mm256_sll_epi32(a, _mm_cvtsi32_si128(c));
}

SIMD_INLINE v256 v256_shr_u32(v256 a, unsigned int c) {
  return _mm256_srl_epi32(a, _mm_cvtsi32_si128(c));
}

SIMD_INLINE v256 v256_shr_s32(v256 a, unsigned int c) {
  return _mm256_sra_epi32(a, _mm_cvtsi32_si128(c));
}

/* These intrinsics require immediate values, so we must use #defines
   to enforce that. */
// _mm256_slli_si256 works on 128 bit lanes and can't be used
#define v256_shl_n_byte(a, n)                                                 \
  ((n) < 16                                                                   \
       ? v256_from_v128(v128_or(v128_shl_n_byte(v256_high_v128(a), n),        \
                                v128_shr_n_byte(v256_low_v128(a), 16 - (n))), \
                        v128_shl_n_byte(v256_low_v128(a), n))                 \
       : v256_from_v128(v128_shl_n_byte(v256_low_v128(a), (n)-16),            \
                        v128_zero()))

// _mm256_srli_si256 works on 128 bit lanes and can't be used
#define v256_shr_n_byte(a, n)                                                 \
  ((n) < 16                                                                   \
       ? _mm256_alignr_epi8(                                                  \
             _mm256_permute2x128_si256(a, a, _MM_SHUFFLE(2, 0, 0, 1)), a, n)  \
       : ((n) > 16                                                            \
              ? _mm256_srli_si256(                                            \
                    _mm256_permute2x128_si256(a, a, _MM_SHUFFLE(2, 0, 0, 1)), \
                    (n)-16)                                                   \
              : _mm256_permute2x128_si256(a, a, _MM_SHUFFLE(2, 0, 0, 1))))

// _mm256_alignr_epi8 works on two 128 bit lanes and can't be used
#define v256_align(a, b, c) \
  ((c) ? v256_or(v256_shr_n_byte(b, c), v256_shl_n_byte(a, 32 - c)) : b)

#define v256_shl_n_8(a, c)                                   \
  _mm256_and_si256(_mm256_set1_epi8((uint8_t)(0xff << (c))), \
                   _mm256_slli_epi16(a, c))
#define v256_shr_n_u8(a, c) \
  _mm256_and_si256(_mm256_set1_epi8(0xff >> (c)), _mm256_srli_epi16(a, c))
#define v256_shr_n_s8(a, c)                                                  \
  _mm256_packs_epi16(_mm256_srai_epi16(_mm256_unpacklo_epi8(a, a), (c) + 8), \
                     _mm256_srai_epi16(_mm256_unpackhi_epi8(a, a), (c) + 8))
#define v256_shl_n_16(a, c) _mm256_slli_epi16(a, c)
#define v256_shr_n_u16(a, c) _mm256_srli_epi16(a, c)
#define v256_shr_n_s16(a, c) _mm256_srai_epi16(a, c)
#define v256_shl_n_32(a, c) _mm256_slli_epi32(a, c)
#define v256_shr_n_u32(a, c) _mm256_srli_epi32(a, c)
#define v256_shr_n_s32(a, c) _mm256_srai_epi32(a, c)
#endif

#endif /* _V256_INTRINSICS_H */