[95d55d]: src / video_dec / libmpeg2 / idct_mmx.c Maximize Restore History

Download this file

idct_mmx.c    742 lines (567 with data), 22.2 kB

  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
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
/*
* idct_mmx.c
* Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
* Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
*
* This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
* See http://libmpeg2.sourceforge.net/ for updates.
*
* mpeg2dec is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* mpeg2dec is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA.
*/
#include "config.h"
#if defined(ARCH_X86) || defined(ARCH_X86_64)
#include <inttypes.h>
#include "mpeg2_internal.h"
#include <xine/xineutils.h>
#include "xine_mmx.h"
#define ROW_SHIFT 11
#define COL_SHIFT 6
#define round(bias) ((int)(((bias)+0.5) * (1<<ROW_SHIFT)))
#define rounder(bias) {round (bias), round (bias)}
#if 0
/* C row IDCT - its just here to document the MMXEXT and MMX versions */
static inline void idct_row (int16_t * row, int offset,
int16_t * table, int32_t * rounder)
{
int C1, C2, C3, C4, C5, C6, C7;
int a0, a1, a2, a3, b0, b1, b2, b3;
row += offset;
C1 = table[1];
C2 = table[2];
C3 = table[3];
C4 = table[4];
C5 = table[5];
C6 = table[6];
C7 = table[7];
a0 = C4*row[0] + C2*row[2] + C4*row[4] + C6*row[6] + *rounder;
a1 = C4*row[0] + C6*row[2] - C4*row[4] - C2*row[6] + *rounder;
a2 = C4*row[0] - C6*row[2] - C4*row[4] + C2*row[6] + *rounder;
a3 = C4*row[0] - C2*row[2] + C4*row[4] - C6*row[6] + *rounder;
b0 = C1*row[1] + C3*row[3] + C5*row[5] + C7*row[7];
b1 = C3*row[1] - C7*row[3] - C1*row[5] - C5*row[7];
b2 = C5*row[1] - C1*row[3] + C7*row[5] + C3*row[7];
b3 = C7*row[1] - C5*row[3] + C3*row[5] - C1*row[7];
row[0] = (a0 + b0) >> ROW_SHIFT;
row[1] = (a1 + b1) >> ROW_SHIFT;
row[2] = (a2 + b2) >> ROW_SHIFT;
row[3] = (a3 + b3) >> ROW_SHIFT;
row[4] = (a3 - b3) >> ROW_SHIFT;
row[5] = (a2 - b2) >> ROW_SHIFT;
row[6] = (a1 - b1) >> ROW_SHIFT;
row[7] = (a0 - b0) >> ROW_SHIFT;
}
#endif
/* MMXEXT row IDCT */
#define mmxext_table(c1,c2,c3,c4,c5,c6,c7) { c4, c2, -c4, -c2, \
c4, c6, c4, c6, \
c1, c3, -c1, -c5, \
c5, c7, c3, -c7, \
c4, -c6, c4, -c6, \
-c4, c2, c4, -c2, \
c5, -c1, c3, -c1, \
c7, c3, c7, -c5 }
static inline void mmxext_row_head (int16_t * row, int offset, int16_t * table)
{
movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0
movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1
movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0
movq_m2r (*table, mm3); // mm3 = -C2 -C4 C2 C4
movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1
movq_m2r (*(table+4), mm4); // mm4 = C6 C4 C6 C4
pmaddwd_r2r (mm0, mm3); // mm3 = -C4*x4-C2*x6 C4*x0+C2*x2
pshufw_r2r (mm2, mm2, 0x4e); // mm2 = x2 x0 x6 x4
}
static inline void mmxext_row (int16_t * table, int32_t * rounder)
{
movq_m2r (*(table+8), mm1); // mm1 = -C5 -C1 C3 C1
pmaddwd_r2r (mm2, mm4); // mm4 = C4*x0+C6*x2 C4*x4+C6*x6
pmaddwd_m2r (*(table+16), mm0); // mm0 = C4*x4-C6*x6 C4*x0-C6*x2
pshufw_r2r (mm6, mm6, 0x4e); // mm6 = x3 x1 x7 x5
movq_m2r (*(table+12), mm7); // mm7 = -C7 C3 C7 C5
pmaddwd_r2r (mm5, mm1); // mm1 = -C1*x5-C5*x7 C1*x1+C3*x3
paddd_m2r (*rounder, mm3); // mm3 += rounder
pmaddwd_r2r (mm6, mm7); // mm7 = C3*x1-C7*x3 C5*x5+C7*x7
pmaddwd_m2r (*(table+20), mm2); // mm2 = C4*x0-C2*x2 -C4*x4+C2*x6
paddd_r2r (mm4, mm3); // mm3 = a1 a0 + rounder
pmaddwd_m2r (*(table+24), mm5); // mm5 = C3*x5-C1*x7 C5*x1-C1*x3
movq_r2r (mm3, mm4); // mm4 = a1 a0 + rounder
pmaddwd_m2r (*(table+28), mm6); // mm6 = C7*x1-C5*x3 C7*x5+C3*x7
paddd_r2r (mm7, mm1); // mm1 = b1 b0
paddd_m2r (*rounder, mm0); // mm0 += rounder
psubd_r2r (mm1, mm3); // mm3 = a1-b1 a0-b0 + rounder
psrad_i2r (ROW_SHIFT, mm3); // mm3 = y6 y7
paddd_r2r (mm4, mm1); // mm1 = a1+b1 a0+b0 + rounder
paddd_r2r (mm2, mm0); // mm0 = a3 a2 + rounder
psrad_i2r (ROW_SHIFT, mm1); // mm1 = y1 y0
paddd_r2r (mm6, mm5); // mm5 = b3 b2
movq_r2r (mm0, mm4); // mm4 = a3 a2 + rounder
paddd_r2r (mm5, mm0); // mm0 = a3+b3 a2+b2 + rounder
psubd_r2r (mm5, mm4); // mm4 = a3-b3 a2-b2 + rounder
}
static inline void mmxext_row_tail (int16_t * row, int store)
{
psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2
psrad_i2r (ROW_SHIFT, mm4); // mm4 = y4 y5
packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0
packssdw_r2r (mm3, mm4); // mm4 = y6 y7 y4 y5
movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0
pshufw_r2r (mm4, mm4, 0xb1); // mm4 = y7 y6 y5 y4
/* slot */
movq_r2m (mm4, *(row+store+4)); // save y7 y6 y5 y4
}
static inline void mmxext_row_mid (int16_t * row, int store,
int offset, int16_t * table)
{
movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0
psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2
movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1
psrad_i2r (ROW_SHIFT, mm4); // mm4 = y4 y5
packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0
movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1
packssdw_r2r (mm3, mm4); // mm4 = y6 y7 y4 y5
movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0
movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0
pshufw_r2r (mm4, mm4, 0xb1); // mm4 = y7 y6 y5 y4
movq_m2r (*table, mm3); // mm3 = -C2 -C4 C2 C4
movq_r2m (mm4, *(row+store+4)); // save y7 y6 y5 y4
pmaddwd_r2r (mm0, mm3); // mm3 = -C4*x4-C2*x6 C4*x0+C2*x2
movq_m2r (*(table+4), mm4); // mm4 = C6 C4 C6 C4
pshufw_r2r (mm2, mm2, 0x4e); // mm2 = x2 x0 x6 x4
}
/* MMX row IDCT */
#define mmx_table(c1,c2,c3,c4,c5,c6,c7) { c4, c2, c4, c6, \
c4, c6, -c4, -c2, \
c1, c3, c3, -c7, \
c5, c7, -c1, -c5, \
c4, -c6, c4, -c2, \
-c4, c2, c4, -c6, \
c5, -c1, c7, -c5, \
c7, c3, c3, -c1 }
static inline void mmx_row_head (int16_t * row, int offset, int16_t * table)
{
movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0
movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1
movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0
movq_m2r (*table, mm3); // mm3 = C6 C4 C2 C4
movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1
punpckldq_r2r (mm0, mm0); // mm0 = x2 x0 x2 x0
movq_m2r (*(table+4), mm4); // mm4 = -C2 -C4 C6 C4
pmaddwd_r2r (mm0, mm3); // mm3 = C4*x0+C6*x2 C4*x0+C2*x2
movq_m2r (*(table+8), mm1); // mm1 = -C7 C3 C3 C1
punpckhdq_r2r (mm2, mm2); // mm2 = x6 x4 x6 x4
}
static inline void mmx_row (int16_t * table, int32_t * rounder)
{
pmaddwd_r2r (mm2, mm4); // mm4 = -C4*x4-C2*x6 C4*x4+C6*x6
punpckldq_r2r (mm5, mm5); // mm5 = x3 x1 x3 x1
pmaddwd_m2r (*(table+16), mm0); // mm0 = C4*x0-C2*x2 C4*x0-C6*x2
punpckhdq_r2r (mm6, mm6); // mm6 = x7 x5 x7 x5
movq_m2r (*(table+12), mm7); // mm7 = -C5 -C1 C7 C5
pmaddwd_r2r (mm5, mm1); // mm1 = C3*x1-C7*x3 C1*x1+C3*x3
paddd_m2r (*rounder, mm3); // mm3 += rounder
pmaddwd_r2r (mm6, mm7); // mm7 = -C1*x5-C5*x7 C5*x5+C7*x7
pmaddwd_m2r (*(table+20), mm2); // mm2 = C4*x4-C6*x6 -C4*x4+C2*x6
paddd_r2r (mm4, mm3); // mm3 = a1 a0 + rounder
pmaddwd_m2r (*(table+24), mm5); // mm5 = C7*x1-C5*x3 C5*x1-C1*x3
movq_r2r (mm3, mm4); // mm4 = a1 a0 + rounder
pmaddwd_m2r (*(table+28), mm6); // mm6 = C3*x5-C1*x7 C7*x5+C3*x7
paddd_r2r (mm7, mm1); // mm1 = b1 b0
paddd_m2r (*rounder, mm0); // mm0 += rounder
psubd_r2r (mm1, mm3); // mm3 = a1-b1 a0-b0 + rounder
psrad_i2r (ROW_SHIFT, mm3); // mm3 = y6 y7
paddd_r2r (mm4, mm1); // mm1 = a1+b1 a0+b0 + rounder
paddd_r2r (mm2, mm0); // mm0 = a3 a2 + rounder
psrad_i2r (ROW_SHIFT, mm1); // mm1 = y1 y0
paddd_r2r (mm6, mm5); // mm5 = b3 b2
movq_r2r (mm0, mm7); // mm7 = a3 a2 + rounder
paddd_r2r (mm5, mm0); // mm0 = a3+b3 a2+b2 + rounder
psubd_r2r (mm5, mm7); // mm7 = a3-b3 a2-b2 + rounder
}
static inline void mmx_row_tail (int16_t * row, int store)
{
psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2
psrad_i2r (ROW_SHIFT, mm7); // mm7 = y4 y5
packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0
packssdw_r2r (mm3, mm7); // mm7 = y6 y7 y4 y5
movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0
movq_r2r (mm7, mm4); // mm4 = y6 y7 y4 y5
pslld_i2r (16, mm7); // mm7 = y7 0 y5 0
psrld_i2r (16, mm4); // mm4 = 0 y6 0 y4
por_r2r (mm4, mm7); // mm7 = y7 y6 y5 y4
/* slot */
movq_r2m (mm7, *(row+store+4)); // save y7 y6 y5 y4
}
static inline void mmx_row_mid (int16_t * row, int store,
int offset, int16_t * table)
{
movq_m2r (*(row+offset), mm2); // mm2 = x6 x4 x2 x0
psrad_i2r (ROW_SHIFT, mm0); // mm0 = y3 y2
movq_m2r (*(row+offset+4), mm5); // mm5 = x7 x5 x3 x1
psrad_i2r (ROW_SHIFT, mm7); // mm7 = y4 y5
packssdw_r2r (mm0, mm1); // mm1 = y3 y2 y1 y0
movq_r2r (mm5, mm6); // mm6 = x7 x5 x3 x1
packssdw_r2r (mm3, mm7); // mm7 = y6 y7 y4 y5
movq_r2r (mm2, mm0); // mm0 = x6 x4 x2 x0
movq_r2m (mm1, *(row+store)); // save y3 y2 y1 y0
movq_r2r (mm7, mm1); // mm1 = y6 y7 y4 y5
punpckldq_r2r (mm0, mm0); // mm0 = x2 x0 x2 x0
psrld_i2r (16, mm7); // mm7 = 0 y6 0 y4
movq_m2r (*table, mm3); // mm3 = C6 C4 C2 C4
pslld_i2r (16, mm1); // mm1 = y7 0 y5 0
movq_m2r (*(table+4), mm4); // mm4 = -C2 -C4 C6 C4
por_r2r (mm1, mm7); // mm7 = y7 y6 y5 y4
movq_m2r (*(table+8), mm1); // mm1 = -C7 C3 C3 C1
punpckhdq_r2r (mm2, mm2); // mm2 = x6 x4 x6 x4
movq_r2m (mm7, *(row+store+4)); // save y7 y6 y5 y4
pmaddwd_r2r (mm0, mm3); // mm3 = C4*x0+C6*x2 C4*x0+C2*x2
}
#if 0
// C column IDCT - its just here to document the MMXEXT and MMX versions
static inline void idct_col (int16_t * col, int offset)
{
/* multiplication - as implemented on mmx */
#define F(c,x) (((c) * (x)) >> 16)
/* saturation - it helps us handle torture test cases */
#define S(x) (((x)>32767) ? 32767 : ((x)<-32768) ? -32768 : (x))
int16_t x0, x1, x2, x3, x4, x5, x6, x7;
int16_t y0, y1, y2, y3, y4, y5, y6, y7;
int16_t a0, a1, a2, a3, b0, b1, b2, b3;
int16_t u04, v04, u26, v26, u17, v17, u35, v35, u12, v12;
col += offset;
x0 = col[0*8];
x1 = col[1*8];
x2 = col[2*8];
x3 = col[3*8];
x4 = col[4*8];
x5 = col[5*8];
x6 = col[6*8];
x7 = col[7*8];
u04 = S (x0 + x4);
v04 = S (x0 - x4);
u26 = S (F (T2, x6) + x2);
v26 = S (F (T2, x2) - x6);
a0 = S (u04 + u26);
a1 = S (v04 + v26);
a2 = S (v04 - v26);
a3 = S (u04 - u26);
u17 = S (F (T1, x7) + x1);
v17 = S (F (T1, x1) - x7);
u35 = S (F (T3, x5) + x3);
v35 = S (F (T3, x3) - x5);
b0 = S (u17 + u35);
b3 = S (v17 - v35);
u12 = S (u17 - u35);
v12 = S (v17 + v35);
u12 = S (2 * F (C4, u12));
v12 = S (2 * F (C4, v12));
b1 = S (u12 + v12);
b2 = S (u12 - v12);
y0 = S (a0 + b0) >> COL_SHIFT;
y1 = S (a1 + b1) >> COL_SHIFT;
y2 = S (a2 + b2) >> COL_SHIFT;
y3 = S (a3 + b3) >> COL_SHIFT;
y4 = S (a3 - b3) >> COL_SHIFT;
y5 = S (a2 - b2) >> COL_SHIFT;
y6 = S (a1 - b1) >> COL_SHIFT;
y7 = S (a0 - b0) >> COL_SHIFT;
col[0*8] = y0;
col[1*8] = y1;
col[2*8] = y2;
col[3*8] = y3;
col[4*8] = y4;
col[5*8] = y5;
col[6*8] = y6;
col[7*8] = y7;
}
#endif
// MMX column IDCT
static inline void idct_col (int16_t * col, int offset)
{
#define T1 13036
#define T2 27146
#define T3 43790
#define C4 23170
static short _T1[] ATTR_ALIGN(8) = {T1,T1,T1,T1};
static short _T2[] ATTR_ALIGN(8) = {T2,T2,T2,T2};
static short _T3[] ATTR_ALIGN(8) = {T3,T3,T3,T3};
static short _C4[] ATTR_ALIGN(8) = {C4,C4,C4,C4};
/* column code adapted from peter gubanov */
/* http://www.elecard.com/peter/idct.shtml */
movq_m2r (*_T1, mm0); // mm0 = T1
movq_m2r (*(col+offset+1*8), mm1); // mm1 = x1
movq_r2r (mm0, mm2); // mm2 = T1
movq_m2r (*(col+offset+7*8), mm4); // mm4 = x7
pmulhw_r2r (mm1, mm0); // mm0 = T1*x1
movq_m2r (*_T3, mm5); // mm5 = T3
pmulhw_r2r (mm4, mm2); // mm2 = T1*x7
movq_m2r (*(col+offset+5*8), mm6); // mm6 = x5
movq_r2r (mm5, mm7); // mm7 = T3-1
movq_m2r (*(col+offset+3*8), mm3); // mm3 = x3
psubsw_r2r (mm4, mm0); // mm0 = v17
movq_m2r (*_T2, mm4); // mm4 = T2
pmulhw_r2r (mm3, mm5); // mm5 = (T3-1)*x3
paddsw_r2r (mm2, mm1); // mm1 = u17
pmulhw_r2r (mm6, mm7); // mm7 = (T3-1)*x5
/* slot */
movq_r2r (mm4, mm2); // mm2 = T2
paddsw_r2r (mm3, mm5); // mm5 = T3*x3
pmulhw_m2r (*(col+offset+2*8), mm4);// mm4 = T2*x2
paddsw_r2r (mm6, mm7); // mm7 = T3*x5
psubsw_r2r (mm6, mm5); // mm5 = v35
paddsw_r2r (mm3, mm7); // mm7 = u35
movq_m2r (*(col+offset+6*8), mm3); // mm3 = x6
movq_r2r (mm0, mm6); // mm6 = v17
pmulhw_r2r (mm3, mm2); // mm2 = T2*x6
psubsw_r2r (mm5, mm0); // mm0 = b3
psubsw_r2r (mm3, mm4); // mm4 = v26
paddsw_r2r (mm6, mm5); // mm5 = v12
movq_r2m (mm0, *(col+offset+3*8)); // save b3 in scratch0
movq_r2r (mm1, mm6); // mm6 = u17
paddsw_m2r (*(col+offset+2*8), mm2);// mm2 = u26
paddsw_r2r (mm7, mm6); // mm6 = b0
psubsw_r2r (mm7, mm1); // mm1 = u12
movq_r2r (mm1, mm7); // mm7 = u12
movq_m2r (*(col+offset+0*8), mm3); // mm3 = x0
paddsw_r2r (mm5, mm1); // mm1 = u12+v12
movq_m2r (*_C4, mm0); // mm0 = C4/2
psubsw_r2r (mm5, mm7); // mm7 = u12-v12
movq_r2m (mm6, *(col+offset+5*8)); // save b0 in scratch1
pmulhw_r2r (mm0, mm1); // mm1 = b1/2
movq_r2r (mm4, mm6); // mm6 = v26
pmulhw_r2r (mm0, mm7); // mm7 = b2/2
movq_m2r (*(col+offset+4*8), mm5); // mm5 = x4
movq_r2r (mm3, mm0); // mm0 = x0
psubsw_r2r (mm5, mm3); // mm3 = v04
paddsw_r2r (mm5, mm0); // mm0 = u04
paddsw_r2r (mm3, mm4); // mm4 = a1
movq_r2r (mm0, mm5); // mm5 = u04
psubsw_r2r (mm6, mm3); // mm3 = a2
paddsw_r2r (mm2, mm5); // mm5 = a0
paddsw_r2r (mm1, mm1); // mm1 = b1
psubsw_r2r (mm2, mm0); // mm0 = a3
paddsw_r2r (mm7, mm7); // mm7 = b2
movq_r2r (mm3, mm2); // mm2 = a2
movq_r2r (mm4, mm6); // mm6 = a1
paddsw_r2r (mm7, mm3); // mm3 = a2+b2
psraw_i2r (COL_SHIFT, mm3); // mm3 = y2
paddsw_r2r (mm1, mm4); // mm4 = a1+b1
psraw_i2r (COL_SHIFT, mm4); // mm4 = y1
psubsw_r2r (mm1, mm6); // mm6 = a1-b1
movq_m2r (*(col+offset+5*8), mm1); // mm1 = b0
psubsw_r2r (mm7, mm2); // mm2 = a2-b2
psraw_i2r (COL_SHIFT, mm6); // mm6 = y6
movq_r2r (mm5, mm7); // mm7 = a0
movq_r2m (mm4, *(col+offset+1*8)); // save y1
psraw_i2r (COL_SHIFT, mm2); // mm2 = y5
movq_r2m (mm3, *(col+offset+2*8)); // save y2
paddsw_r2r (mm1, mm5); // mm5 = a0+b0
movq_m2r (*(col+offset+3*8), mm4); // mm4 = b3
psubsw_r2r (mm1, mm7); // mm7 = a0-b0
psraw_i2r (COL_SHIFT, mm5); // mm5 = y0
movq_r2r (mm0, mm3); // mm3 = a3
movq_r2m (mm2, *(col+offset+5*8)); // save y5
psubsw_r2r (mm4, mm3); // mm3 = a3-b3
psraw_i2r (COL_SHIFT, mm7); // mm7 = y7
paddsw_r2r (mm0, mm4); // mm4 = a3+b3
movq_r2m (mm5, *(col+offset+0*8)); // save y0
psraw_i2r (COL_SHIFT, mm3); // mm3 = y4
movq_r2m (mm6, *(col+offset+6*8)); // save y6
psraw_i2r (COL_SHIFT, mm4); // mm4 = y3
movq_r2m (mm7, *(col+offset+7*8)); // save y7
movq_r2m (mm3, *(col+offset+4*8)); // save y4
movq_r2m (mm4, *(col+offset+3*8)); // save y3
}
static int32_t rounder0[] ATTR_ALIGN(8) =
rounder ((1 << (COL_SHIFT - 1)) - 0.5);
static int32_t rounder4[] ATTR_ALIGN(8) = rounder (0);
static int32_t rounder1[] ATTR_ALIGN(8) =
rounder (1.25683487303); /* C1*(C1/C4+C1+C7)/2 */
static int32_t rounder7[] ATTR_ALIGN(8) =
rounder (-0.25); /* C1*(C7/C4+C7-C1)/2 */
static int32_t rounder2[] ATTR_ALIGN(8) =
rounder (0.60355339059); /* C2 * (C6+C2)/2 */
static int32_t rounder6[] ATTR_ALIGN(8) =
rounder (-0.25); /* C2 * (C6-C2)/2 */
static int32_t rounder3[] ATTR_ALIGN(8) =
rounder (0.087788325588); /* C3*(-C3/C4+C3+C5)/2 */
static int32_t rounder5[] ATTR_ALIGN(8) =
rounder (-0.441341716183); /* C3*(-C5/C4+C5-C3)/2 */
#define declare_idct(idct,table,idct_row_head,idct_row,idct_row_tail,idct_row_mid) \
static inline void idct (int16_t * block) \
{ \
static int16_t table04[] ATTR_ALIGN(16) = \
table (22725, 21407, 19266, 16384, 12873, 8867, 4520); \
static int16_t table17[] ATTR_ALIGN(16) = \
table (31521, 29692, 26722, 22725, 17855, 12299, 6270); \
static int16_t table26[] ATTR_ALIGN(16) = \
table (29692, 27969, 25172, 21407, 16819, 11585, 5906); \
static int16_t table35[] ATTR_ALIGN(16) = \
table (26722, 25172, 22654, 19266, 15137, 10426, 5315); \
\
idct_row_head (block, 0*8, table04); \
idct_row (table04, rounder0); \
idct_row_mid (block, 0*8, 4*8, table04); \
idct_row (table04, rounder4); \
idct_row_mid (block, 4*8, 1*8, table17); \
idct_row (table17, rounder1); \
idct_row_mid (block, 1*8, 7*8, table17); \
idct_row (table17, rounder7); \
idct_row_mid (block, 7*8, 2*8, table26); \
idct_row (table26, rounder2); \
idct_row_mid (block, 2*8, 6*8, table26); \
idct_row (table26, rounder6); \
idct_row_mid (block, 6*8, 3*8, table35); \
idct_row (table35, rounder3); \
idct_row_mid (block, 3*8, 5*8, table35); \
idct_row (table35, rounder5); \
idct_row_tail (block, 5*8); \
\
idct_col (block, 0); \
idct_col (block, 4); \
}
#define COPY_MMX(offset,r0,r1,r2) \
do { \
movq_m2r (*(block+offset), r0); \
dest += stride; \
movq_m2r (*(block+offset+4), r1); \
movq_r2m (r2, *dest); \
packuswb_r2r (r1, r0); \
} while (0)
static void block_copy (int16_t * block, uint8_t * dest, int stride)
{
movq_m2r (*(block+0*8), mm0);
movq_m2r (*(block+0*8+4), mm1);
movq_m2r (*(block+1*8), mm2);
packuswb_r2r (mm1, mm0);
movq_m2r (*(block+1*8+4), mm3);
movq_r2m (mm0, *dest);
packuswb_r2r (mm3, mm2);
COPY_MMX (2*8, mm0, mm1, mm2);
COPY_MMX (3*8, mm2, mm3, mm0);
COPY_MMX (4*8, mm0, mm1, mm2);
COPY_MMX (5*8, mm2, mm3, mm0);
COPY_MMX (6*8, mm0, mm1, mm2);
COPY_MMX (7*8, mm2, mm3, mm0);
movq_r2m (mm2, *(dest+stride));
}
#define ADD_MMX(offset,r1,r2,r3,r4) \
do { \
movq_m2r (*(dest+2*stride), r1); \
packuswb_r2r (r4, r3); \
movq_r2r (r1, r2); \
dest += stride; \
movq_r2m (r3, *dest); \
punpcklbw_r2r (mm0, r1); \
paddsw_m2r (*(block+offset), r1); \
punpckhbw_r2r (mm0, r2); \
paddsw_m2r (*(block+offset+4), r2); \
} while (0)
static void block_add (int16_t * block, uint8_t * dest, int stride)
{
movq_m2r (*dest, mm1);
pxor_r2r (mm0, mm0);
movq_m2r (*(dest+stride), mm3);
movq_r2r (mm1, mm2);
punpcklbw_r2r (mm0, mm1);
movq_r2r (mm3, mm4);
paddsw_m2r (*(block+0*8), mm1);
punpckhbw_r2r (mm0, mm2);
paddsw_m2r (*(block+0*8+4), mm2);
punpcklbw_r2r (mm0, mm3);
paddsw_m2r (*(block+1*8), mm3);
packuswb_r2r (mm2, mm1);
punpckhbw_r2r (mm0, mm4);
movq_r2m (mm1, *dest);
paddsw_m2r (*(block+1*8+4), mm4);
ADD_MMX (2*8, mm1, mm2, mm3, mm4);
ADD_MMX (3*8, mm3, mm4, mm1, mm2);
ADD_MMX (4*8, mm1, mm2, mm3, mm4);
ADD_MMX (5*8, mm3, mm4, mm1, mm2);
ADD_MMX (6*8, mm1, mm2, mm3, mm4);
ADD_MMX (7*8, mm3, mm4, mm1, mm2);
packuswb_r2r (mm4, mm3);
movq_r2m (mm3, *(dest+stride));
}
static inline void block_zero (int16_t * block) {
pxor_r2r (mm0, mm0);
movq_r2m (mm0, *(block+0*4));
movq_r2m (mm0, *(block+1*4));
movq_r2m (mm0, *(block+2*4));
movq_r2m (mm0, *(block+3*4));
movq_r2m (mm0, *(block+4*4));
movq_r2m (mm0, *(block+5*4));
movq_r2m (mm0, *(block+6*4));
movq_r2m (mm0, *(block+7*4));
movq_r2m (mm0, *(block+8*4));
movq_r2m (mm0, *(block+9*4));
movq_r2m (mm0, *(block+10*4));
movq_r2m (mm0, *(block+11*4));
movq_r2m (mm0, *(block+12*4));
movq_r2m (mm0, *(block+13*4));
movq_r2m (mm0, *(block+14*4));
movq_r2m (mm0, *(block+15*4));
}
declare_idct (mmxext_idct, mmxext_table,
mmxext_row_head, mmxext_row, mmxext_row_tail, mmxext_row_mid)
void mpeg2_idct_copy_mmxext (int16_t * block, uint8_t * dest, int stride)
{
mmxext_idct (block);
block_copy (block, dest, stride);
block_zero (block);
}
void mpeg2_idct_add_mmxext (int16_t * block, uint8_t * dest, int stride)
{
mmxext_idct (block);
block_add (block, dest, stride);
block_zero (block);
}
void mpeg2_idct_mmxext (int16_t * block)
{
mmxext_idct (block);
}
declare_idct (mmx_idct, mmx_table,
mmx_row_head, mmx_row, mmx_row_tail, mmx_row_mid)
void mpeg2_idct_copy_mmx (int16_t * block, uint8_t * dest, int stride)
{
mmx_idct (block);
block_copy (block, dest, stride);
block_zero (block);
}
void mpeg2_idct_add_mmx (int16_t * block, uint8_t * dest, int stride)
{
mmx_idct (block);
block_add (block, dest, stride);
block_zero (block);
}
void mpeg2_idct_mmx (int16_t * block)
{
mmx_idct (block);
}
void mpeg2_zero_block_mmx (int16_t * block)
{
block_zero (block);
}
void mpeg2_idct_mmx_init (void)
{
int i, j;
/* the mmx/mmxext idct uses a reordered input, so we patch scan tables */
for (i = 0; i < 64; i++) {
j = mpeg2_scan_norm[i];
mpeg2_scan_norm[i] = (j & 0x38) | ((j & 6) >> 1) | ((j & 1) << 2);
j = mpeg2_scan_alt[i];
mpeg2_scan_alt[i] = (j & 0x38) | ((j & 6) >> 1) | ((j & 1) << 2);
}
}
#endif