summaryrefslogtreecommitdiffstats
path: root/src/vppinfra/vector_sse42.h
blob: 50aa662a925aa1f445b6be624ace6e462204b117 (plain)
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
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
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
/*
 * Copyright (c) 2015 Cisco and/or its affiliates.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at:
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
/*
  Copyright (c) 2005 Eliot Dresselhaus

  Permission is hereby granted, free of charge, to any person obtaining
  a copy of this software and associated documentation files (the
  "Software"), to deal in the Software without restriction, including
  without limitation the rights to use, copy, modify, merge, publish,
  distribute, sublicense, and/or sell copies of the Software, and to
  permit persons to whom the Software is furnished to do so, subject to
  the following conditions:

  The above copyright notice and this permission notice shall be
  included in all copies or substantial portions of the Software.

  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/

#ifndef included_vector_sse2_h
#define included_vector_sse2_h

#include <vppinfra/error_bootstrap.h>	/* for ASSERT */
#include <x86intrin.h>

/* 128 bit interleaves. */
always_inline u8x16
u8x16_interleave_hi (u8x16 a, u8x16 b)
{
  return (u8x16) _mm_unpackhi_epi8 ((__m128i) a, (__m128i) b);
}

always_inline u8x16
u8x16_interleave_lo (u8x16 a, u8x16 b)
{
  return (u8x16) _mm_unpacklo_epi8 ((__m128i) a, (__m128i) b);
}

always_inline u16x8
u16x8_interleave_hi (u16x8 a, u16x8 b)
{
  return (u16x8) _mm_unpackhi_epi16 ((__m128i) a, (__m128i) b);
}

always_inline u16x8
u16x8_interleave_lo (u16x8 a, u16x8 b)
{
  return (u16x8) _mm_unpacklo_epi16 ((__m128i) a, (__m128i) b);
}

always_inline u32x4
u32x4_interleave_hi (u32x4 a, u32x4 b)
{
  return (u32x4) _mm_unpackhi_epi32 ((__m128i) a, (__m128i) b);
}

always_inline u32x4
u32x4_interleave_lo (u32x4 a, u32x4 b)
{
  return (u32x4) _mm_unpacklo_epi32 ((__m128i) a, (__m128i) b);
}

always_inline u64x2
u64x2_interleave_hi (u64x2 a, u64x2 b)
{
  return (u64x2) _mm_unpackhi_epi64 ((__m128i) a, (__m128i) b);
}

always_inline u64x2
u64x2_interleave_lo (u64x2 a, u64x2 b)
{
  return (u64x2) _mm_unpacklo_epi64 ((__m128i) a, (__m128i) b);
}

/* 64 bit interleaves. */
always_inline u8x8
u8x8_interleave_hi (u8x8 a, u8x8 b)
{
  return (u8x8) _m_punpckhbw ((__m64) a, (__m64) b);
}

always_inline u8x8
u8x8_interleave_lo (u8x8 a, u8x8 b)
{
  return (u8x8) _m_punpcklbw ((__m64) a, (__m64) b);
}

always_inline u16x4
u16x4_interleave_hi (u16x4 a, u16x4 b)
{
  return (u16x4) _m_punpckhwd ((__m64) a, (__m64) b);
}

always_inline u16x4
u16x4_interleave_lo (u16x4 a, u16x4 b)
{
  return (u16x4) _m_punpcklwd ((__m64) a, (__m64) b);
}

always_inline u32x2
u32x2_interleave_hi (u32x2 a, u32x2 b)
{
  return (u32x2) _m_punpckhdq ((__m64) a, (__m64) b);
}

always_inline u32x2
u32x2_interleave_lo (u32x2 a, u32x2 b)
{
  return (u32x2) _m_punpckldq ((__m64) a, (__m64) b);
}

/* 128 bit packs. */
always_inline u8x16
u16x8_pack (u16x8 lo, u16x8 hi)
{
  return (u8x16) _mm_packus_epi16 ((__m128i) lo, (__m128i) hi);
}

always_inline i8x16
i16x8_pack (i16x8 lo, i16x8 hi)
{
  return (i8x16) _mm_packs_epi16 ((__m128i) lo, (__m128i) hi);
}

always_inline u16x8
u32x4_pack (u32x4 lo, u32x4 hi)
{
  return (u16x8) _mm_packs_epi32 ((__m128i) lo, (__m128i) hi);
}

/* 64 bit packs. */
always_inline u8x8
u16x4_pack (u16x4 lo, u16x4 hi)
{
  return (u8x8) _m_packuswb ((__m64) lo, (__m64) hi);
}

always_inline i8x8
i16x4_pack (i16x4 lo, i16x4 hi)
{
  return (i8x8) _m_packsswb ((__m64) lo, (__m64) hi);
}

always_inline u16x4
u32x2_pack (u32x2 lo, u32x2 hi)
{
  return (u16x4) _m_packssdw ((__m64) lo, (__m64) hi);
}

always_inline i16x4
i32x2_pack (i32x2 lo, i32x2 hi)
{
  return (i16x4) _m_packssdw ((__m64) lo, (__m64) hi);
}

#ifndef __ICC
always_inline u64x2
u64x2_read_lo (u64x2 x, u64 * a)
{
  return (u64x2) _mm_loadl_pi ((__m128) x, (__m64 *) a);
}

always_inline u64x2
u64x2_read_hi (u64x2 x, u64 * a)
{
  return (u64x2) _mm_loadh_pi ((__m128) x, (__m64 *) a);
}

always_inline void
u64x2_write_lo (u64x2 x, u64 * a)
{
  _mm_storel_pi ((__m64 *) a, (__m128) x);
}

always_inline void
u64x2_write_hi (u64x2 x, u64 * a)
{
  _mm_storeh_pi ((__m64 *) a, (__m128) x);
}
#endif

/* Unaligned loads/stores. */

#define _(t)						\
  always_inline void t##_store_unaligned (t x, void * a)	\
  { _mm_storeu_si128 ((__m128i *) a, (__m128i) x); }	\
  always_inline t t##_load_unaligned (void * a)		\
  { return (t) _mm_loadu_si128 ((__m128i *) a); }

_(u8x16) _(u16x8) _(u32x4) _(u64x2) _(i8x16) _(i16x8) _(i32x4) _(i64x2)
#undef _
#define _signed_binop(n,m,f,g)                                         \
  /* Unsigned */                                                       \
  always_inline u##n##x##m                                             \
  u##n##x##m##_##f (u##n##x##m x, u##n##x##m y)                        \
  { return (u##n##x##m) _mm_##g##n ((__m128i) x, (__m128i) y); } \
                                                                       \
  /* Signed */                                                         \
  always_inline i##n##x##m                                             \
  i##n##x##m##_##f (i##n##x##m x, i##n##x##m y)                        \
  { return (i##n##x##m) _mm_##g##n ((__m128i) x, (__m128i) y); }
/* Addition/subtraction with saturation. */
  _signed_binop (8, 16, add_saturate, adds_epu)
_signed_binop (16, 8, add_saturate, adds_epu)
_signed_binop (8, 16, sub_saturate, subs_epu)
_signed_binop (16, 8, sub_saturate, subs_epu)
/* Multiplication. */
     always_inline i16x8 i16x8_mul_lo (i16x8 x, i16x8 y)
{
  return (i16x8) _mm_mullo_epi16 ((__m128i) x, (__m128i) y);
}

always_inline u16x8
u16x8_mul_lo (u16x8 x, u16x8 y)
{
  return (u16x8) _mm_mullo_epi16 ((__m128i) x, (__m128i) y);
}

always_inline i16x8
i16x8_mul_hi (i16x8 x, i16x8 y)
{
  return (i16x8) _mm_mulhi_epu16 ((__m128i) x, (__m128i) y);
}

always_inline u16x8
u16x8_mul_hi (u16x8 x, u16x8 y)
{
  return (u16x8) _mm_mulhi_epu16 ((__m128i) x, (__m128i) y);
}

/* 128 bit shifts. */

#define _(p,a,b,c,f)           \
  always_inline p##a##x##b p##a##x##b##_ishift_##c (p##a##x##b x, int i)       \
  { return (p##a##x##b) _mm_##f##i_epi##a ((__m128i) x, i); }                  \
                                                                               \
  always_inline p##a##x##b p##a##x##b##_shift_##c (p##a##x##b x, p##a##x##b y) \
  { return (p##a##x##b) _mm_##f##_epi##a ((__m128i) x, (__m128i) y); }

_(u, 16, 8, left, sll)
_(u, 32, 4, left, sll)
_(u, 64, 2, left, sll)
_(u, 16, 8, right, srl)
_(u, 32, 4, right, srl)
_(u, 64, 2, right, srl)
_(i, 16, 8, left, sll)
_(i, 32, 4, left, sll)
_(i, 64, 2, left, sll) _(i, 16, 8, right, sra) _(i, 32, 4, right, sra)
#undef _
/* 64 bit shifts. */
  always_inline u16x4
u16x4_shift_left (u16x4 x, u16x4 i)
{
  return (u16x4) _m_psllw ((__m64) x, (__m64) i);
};

always_inline u32x2
u32x2_shift_left (u32x2 x, u32x2 i)
{
  return (u32x2) _m_pslld ((__m64) x, (__m64) i);
};

always_inline u16x4
u16x4_shift_right (u16x4 x, u16x4 i)
{
  return (u16x4) _m_psrlw ((__m64) x, (__m64) i);
};

always_inline u32x2
u32x2_shift_right (u32x2 x, u32x2 i)
{
  return (u32x2) _m_psrld ((__m64) x, (__m64) i);
};

always_inline i16x4
i16x4_shift_left (i16x4 x, i16x4 i)
{
  return (i16x4) _m_psllw ((__m64) x, (__m64) i);
};

always_inline i32x2
i32x2_shift_left (i32x2 x, i32x2 i)
{
  return (i32x2) _m_pslld ((__m64) x, (__m64) i);
};

always_inline i16x4
i16x4_shift_right (i16x4 x, i16x4 i)
{
  return (i16x4) _m_psraw ((__m64) x, (__m64) i);
};

always_inline i32x2
i32x2_shift_right (i32x2 x, i32x2 i)
{
  return (i32x2) _m_psrad ((__m64) x, (__m64) i);
};

#define u8x16_word_shift_left(a,n)  (u8x16) _mm_slli_si128((__m128i) a, n)
#define u8x16_word_shift_right(a,n) (u8x16) _mm_srli_si128((__m128i) a, n)

#define i8x16_word_shift_left(a,n) \
  ((i8x16) u8x16_word_shift_left((u8x16) (a), (n)))
#define i8x16_word_shift_right(a,n) \
  ((i8x16) u8x16_word_shift_right((u8x16) (a), (n)))

#define u16x8_word_shift_left(a,n) \
  ((u16x8) u8x16_word_shift_left((u8x16) (a), (n) * sizeof (u16)))
#define i16x8_word_shift_left(a,n) \
  ((u16x8) u8x16_word_shift_left((u8x16) (a), (n) * sizeof (u16)))
#define u16x8_word_shift_right(a,n) \
  ((u16x8) u8x16_word_shift_right((u8x16) (a), (n) * sizeof (u16)))
#define i16x8_word_shift_right(a,n) \
  ((i16x8) u8x16_word_shift_right((u8x16) (a), (n) * sizeof (u16)))

#define u32x4_word_shift_left(a,n) \
  ((u32x4) u8x16_word_shift_left((u8x16) (a), (n) * sizeof (u32)))
#define i32x4_word_shift_left(a,n) \
  ((u32x4) u8x16_word_shift_left((u8x16) (a), (n) * sizeof (u32)))
#define u32x4_word_shift_right(a,n) \
  ((u32x4) u8x16_word_shift_right((u8x16) (a), (n) * sizeof (u32)))
#define i32x4_word_shift_right(a,n) \
  ((i32x4) u8x16_word_shift_right((u8x16) (a), (n) * sizeof (u32)))

#define u64x2_word_shift_left(a,n) \
  ((u64x2) u8x16_word_shift_left((u8x16) (a), (n) * sizeof (u64)))
#define i64x2_word_shift_left(a,n) \
  ((u64x2) u8x16_word_shift_left((u8x16) (a), (n) * sizeof (u64)))
#define u64x2_word_shift_right(a,n) \
  ((u64x2) u8x16_word_shift_right((u8x16) (a), (n) * sizeof (u64)))
#define i64x2_word_shift_right(a,n) \
  ((i64x2) u8x16_word_shift_right((u8x16) (a), (n) * sizeof (u64)))

/* SSE2 has no rotate instructions: use shifts to simulate them. */
#define _(t,n,lr1,lr2)					\
  always_inline t##x##n					\
  t##x##n##_irotate_##lr1 (t##x##n w, int i)		\
  {							\
    ASSERT (i >= 0 && i <= BITS (t));			\
    return (t##x##n##_ishift_##lr1 (w, i)		\
	    | t##x##n##_ishift_##lr2 (w, BITS (t) - i)); \
  }							\
							\
  always_inline t##x##n					\
  t##x##n##_rotate_##lr1 (t##x##n w, t##x##n i)		\
  {							\
    t##x##n j = t##x##n##_splat (BITS (t));		\
    return (t##x##n##_shift_##lr1 (w, i)		\
	    | t##x##n##_shift_##lr2 (w, j - i));	\
  }

_(u16, 8, left, right);
_(u16, 8, right, left);
_(u32, 4, left, right);
_(u32, 4, right, left);
_(u64, 2, left, right);
_(u64, 2, right, left);

#undef _

#ifndef __clang__
#define _(t,n,lr1,lr2)						\
  always_inline t##x##n						\
  t##x##n##_word_rotate2_##lr1 (t##x##n w0, t##x##n w1, int i)	\
  {								\
    int m = sizeof (t##x##n) / sizeof (t);			\
    ASSERT (i >= 0 && i < m);					\
    return (t##x##n##_word_shift_##lr1 (w0, i)			\
	    | t##x##n##_word_shift_##lr2 (w1, m - i));		\
  }								\
								\
  always_inline t##x##n						\
  t##x##n##_word_rotate_##lr1 (t##x##n w0, int i)		\
  { return t##x##n##_word_rotate2_##lr1 (w0, w0, i); }

_(u8, 16, left, right);
_(u8, 16, right, left);
_(u16, 8, left, right);
_(u16, 8, right, left);
_(u32, 4, left, right);
_(u32, 4, right, left);
_(u64, 2, left, right);
_(u64, 2, right, left);

#undef _
#endif

always_inline int
u8x16_is_all_zero (u8x16 x)
{
  return _mm_testz_si128 ((__m128i) x, (__m128i) x);
}

always_inline int
u16x8_is_all_zero (u16x8 x)
{
  return _mm_testz_si128 ((__m128i) x, (__m128i) x);
}

always_inline int
u32x4_is_all_zero (u32x4 x)
{
  return _mm_testz_si128 ((__m128i) x, (__m128i) x);
}

always_inline int
u64x2_is_all_zero (u64x2 x)
{
  return _mm_testz_si128 ((__m128i) x, (__m128i) x);
}

#define u32x4_select(A,MASK)						\
({									\
  u32x4 _x, _y;								\
  _x = (A);								\
  asm volatile ("pshufd %[mask], %[x], %[y]"				\
		: /* outputs */ [y] "=x" (_y)				\
		: /* inputs */  [x] "x" (_x), [mask] "i" (MASK));	\
  _y;									\
})

#define u32x4_splat_word(x,i)			\
  u32x4_select ((x), (((i) << (2*0))		\
		      | ((i) << (2*1))		\
		      | ((i) << (2*2))		\
		      | ((i) << (2*3))))

/* Extract low order 32 bit word. */
always_inline u32
u32x4_get0 (u32x4 x)
{
  u32 result;
  asm volatile ("movd %[x], %[result]": /* outputs */ [result] "=r" (result)
		: /* inputs */ [x] "x" (x));
  return result;
}

always_inline u32x4
u32x4_set0 (u32 x)
{
  u32x4 result;
  asm volatile ("movd %[x], %[result]": /* outputs */ [result] "=x" (result)
		: /* inputs */ [x] "r" (x));
  return result;
}

always_inline i32x4
i32x4_set0 (i32 x)
{
  return (i32x4) u32x4_set0 ((u32) x);
}

always_inline i32
i32x4_get0 (i32x4 x)
{
  return (i32) u32x4_get0 ((u32x4) x);
}

/* Converts all ones/zeros compare mask to bitmap. */
always_inline u32
u8x16_compare_byte_mask (u8x16 x)
{
  return _mm_movemask_epi8 ((__m128i) x);
}

extern u8 u32x4_compare_word_mask_table[256];

always_inline u32
u32x4_compare_word_mask (u32x4 x)
{
  u32 m = u8x16_compare_byte_mask ((u8x16) x);
  return (u32x4_compare_word_mask_table[(m >> 0) & 0xff]
	  | (u32x4_compare_word_mask_table[(m >> 8) & 0xff] << 2));
}

always_inline u32
u8x16_zero_byte_mask (u8x16 x)
{
  u8x16 zero = { 0 };
  return u8x16_compare_byte_mask (u8x16_is_equal (x, zero));
}

always_inline u32
u16x8_zero_byte_mask (u16x8 x)
{
  u16x8 zero = { 0 };
  return u8x16_compare_byte_mask ((u8x16) u16x8_is_equal (x, zero));
}

always_inline u32
u32x4_zero_byte_mask (u32x4 x)
{
  u32x4 zero = { 0 };
  return u8x16_compare_byte_mask ((u8x16) u32x4_is_equal (x, zero));
}

always_inline u8x16
u8x16_max (u8x16 x, u8x16 y)
{
  return (u8x16) _mm_max_epu8 ((__m128i) x, (__m128i) y);
}

always_inline u32
u8x16_max_scalar (u8x16 x)
{
  x = u8x16_max (x, u8x16_word_shift_right (x, 8));
  x = u8x16_max (x, u8x16_word_shift_right (x, 4));
  x = u8x16_max (x, u8x16_word_shift_right (x, 2));
  x = u8x16_max (x, u8x16_word_shift_right (x, 1));
  return _mm_extract_epi16 ((__m128i) x, 0) & 0xff;
}

always_inline u8x16
u8x16_min (u8x16 x, u8x16 y)
{
  return (u8x16) _mm_min_epu8 ((__m128i) x, (__m128i) y);
}

always_inline u8
u8x16_min_scalar (u8x16 x)
{
  x = u8x16_min (x, u8x16_word_shift_right (x, 8));
  x = u8x16_min (x, u8x16_word_shift_right (x, 4));
  x = u8x16_min (x, u8x16_word_shift_right (x, 2));
  x = u8x16_min (x, u8x16_word_shift_right (x, 1));
  return _mm_extract_epi16 ((__m128i) x, 0) & 0xff;
}

always_inline i16x8
i16x8_max (i16x8 x, i16x8 y)
{
  return (i16x8) _mm_max_epi16 ((__m128i) x, (__m128i) y);
}

always_inline i16
i16x8_max_scalar (i16x8 x)
{
  x = i16x8_max (x, i16x8_word_shift_right (x, 4));
  x = i16x8_max (x, i16x8_word_shift_right (x, 2));
  x = i16x8_max (x, i16x8_word_shift_right (x, 1));
  return _mm_extract_epi16 ((__m128i) x, 0);
}

always_inline i16x8
i16x8_min (i16x8 x, i16x8 y)
{
  return (i16x8) _mm_min_epi16 ((__m128i) x, (__m128i) y);
}

always_inline i16
i16x8_min_scalar (i16x8 x)
{
  x = i16x8_min (x, i16x8_word_shift_right (x, 4));
  x = i16x8_min (x, i16x8_word_shift_right (x, 2));
  x = i16x8_min (x, i16x8_word_shift_right (x, 1));
  return _mm_extract_epi16 ((__m128i) x, 0);
}

static_always_inline u16
u8x16_msb_mask (u8x16 v)
{
  return _mm_movemask_epi8 ((__m128i) v);
}

#undef _signed_binop

#endif /* included_vector_sse2_h */

/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */