aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/h264_loopfilter.c
blob: 45e1cdf88db79654b22ff26b00ee7b8c125a3652 (plain) (blame)
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
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
/*
 * H.26L/H.264/AVC/JVT/14496-10/... loop filter
 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
 *
 * This file is part of Libav.
 *
 * Libav is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * Libav 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
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with Libav; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file
 * H.264 / AVC / MPEG4 part10 loop filter.
 * @author Michael Niedermayer <michaelni@gmx.at>
 */

#include "libavutil/internal.h"
#include "libavutil/intreadwrite.h"
#include "internal.h"
#include "avcodec.h"
#include "mpegvideo.h"
#include "h264.h"
#include "mathops.h"
#include "rectangle.h"

//#undef NDEBUG
#include <assert.h>

/* Deblocking filter (p153) */
static const uint8_t alpha_table[52*3] = {
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  4,  4,  5,  6,
     7,  8,  9, 10, 12, 13, 15, 17, 20, 22,
    25, 28, 32, 36, 40, 45, 50, 56, 63, 71,
    80, 90,101,113,127,144,162,182,203,226,
   255,255,
   255,255,255,255,255,255,255,255,255,255,255,255,255,
   255,255,255,255,255,255,255,255,255,255,255,255,255,
   255,255,255,255,255,255,255,255,255,255,255,255,255,
   255,255,255,255,255,255,255,255,255,255,255,255,255,
};
static const uint8_t beta_table[52*3] = {
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  2,  2,  2,  3,
     3,  3,  3,  4,  4,  4,  6,  6,  7,  7,
     8,  8,  9,  9, 10, 10, 11, 11, 12, 12,
    13, 13, 14, 14, 15, 15, 16, 16, 17, 17,
    18, 18,
    18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,
    18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,
    18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,
    18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,
};
static const uint8_t tc0_table[52*3][4] = {
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 1 },
    {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 1, 1 }, {-1, 0, 1, 1 }, {-1, 1, 1, 1 },
    {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 },
    {-1, 1, 1, 2 }, {-1, 1, 2, 3 }, {-1, 1, 2, 3 }, {-1, 2, 2, 3 }, {-1, 2, 2, 4 }, {-1, 2, 3, 4 },
    {-1, 2, 3, 4 }, {-1, 3, 3, 5 }, {-1, 3, 4, 6 }, {-1, 3, 4, 6 }, {-1, 4, 5, 7 }, {-1, 4, 5, 8 },
    {-1, 4, 6, 9 }, {-1, 5, 7,10 }, {-1, 6, 8,11 }, {-1, 6, 8,13 }, {-1, 7,10,14 }, {-1, 8,11,16 },
    {-1, 9,12,18 }, {-1,10,13,20 }, {-1,11,15,23 }, {-1,13,17,25 },
    {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
    {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
    {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
    {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
    {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
    {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
    {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
    {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
    {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
};

/* intra: 0 if this loopfilter call is guaranteed to be inter (bS < 4), 1 if it might be intra (bS == 4) */
static av_always_inline void filter_mb_edgev(uint8_t *pix, int stride,
                                             const int16_t bS[4],
                                             unsigned int qp, int a, int b,
                                             H264Context *h, int intra)
{
    const unsigned int index_a = qp + a;
    const int alpha = alpha_table[index_a];
    const int beta  = beta_table[qp + b];
    if (alpha ==0 || beta == 0) return;

    if( bS[0] < 4 || !intra ) {
        int8_t tc[4];
        tc[0] = tc0_table[index_a][bS[0]];
        tc[1] = tc0_table[index_a][bS[1]];
        tc[2] = tc0_table[index_a][bS[2]];
        tc[3] = tc0_table[index_a][bS[3]];
        h->h264dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc);
    } else {
        h->h264dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta);
    }
}

static av_always_inline void filter_mb_edgecv(uint8_t *pix, int stride,
                                              const int16_t bS[4],
                                              unsigned int qp, int a, int b,
                                              H264Context *h, int intra)
{
    const unsigned int index_a = qp + a;
    const int alpha = alpha_table[index_a];
    const int beta  = beta_table[qp + b];
    if (alpha ==0 || beta == 0) return;

    if( bS[0] < 4 || !intra ) {
        int8_t tc[4];
        tc[0] = tc0_table[index_a][bS[0]]+1;
        tc[1] = tc0_table[index_a][bS[1]]+1;
        tc[2] = tc0_table[index_a][bS[2]]+1;
        tc[3] = tc0_table[index_a][bS[3]]+1;
        h->h264dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc);
    } else {
        h->h264dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta);
    }
}

static av_always_inline void filter_mb_mbaff_edgev(H264Context *h, uint8_t *pix,
                                                   int stride,
                                                   const int16_t bS[7], int bsi,
                                                   int qp, int a, int b,
                                                   int intra)
{
    const unsigned int index_a = qp + a;
    const int alpha = alpha_table[index_a];
    const int beta  = beta_table[qp + b];
    if (alpha ==0 || beta == 0) return;

    if( bS[0] < 4 || !intra ) {
        int8_t tc[4];
        tc[0] = tc0_table[index_a][bS[0*bsi]];
        tc[1] = tc0_table[index_a][bS[1*bsi]];
        tc[2] = tc0_table[index_a][bS[2*bsi]];
        tc[3] = tc0_table[index_a][bS[3*bsi]];
        h->h264dsp.h264_h_loop_filter_luma_mbaff(pix, stride, alpha, beta, tc);
    } else {
        h->h264dsp.h264_h_loop_filter_luma_mbaff_intra(pix, stride, alpha, beta);
    }
}

static av_always_inline void filter_mb_mbaff_edgecv(H264Context *h,
                                                    uint8_t *pix, int stride,
                                                    const int16_t bS[7],
                                                    int bsi, int qp, int a,
                                                    int b, int intra)
{
    const unsigned int index_a = qp + a;
    const int alpha = alpha_table[index_a];
    const int beta  = beta_table[qp + b];
    if (alpha ==0 || beta == 0) return;

    if( bS[0] < 4 || !intra ) {
        int8_t tc[4];
        tc[0] = tc0_table[index_a][bS[0*bsi]] + 1;
        tc[1] = tc0_table[index_a][bS[1*bsi]] + 1;
        tc[2] = tc0_table[index_a][bS[2*bsi]] + 1;
        tc[3] = tc0_table[index_a][bS[3*bsi]] + 1;
        h->h264dsp.h264_h_loop_filter_chroma_mbaff(pix, stride, alpha, beta, tc);
    } else {
        h->h264dsp.h264_h_loop_filter_chroma_mbaff_intra(pix, stride, alpha, beta);
    }
}

static av_always_inline void filter_mb_edgeh(uint8_t *pix, int stride,
                                             const int16_t bS[4],
                                             unsigned int qp, int a, int b,
                                             H264Context *h, int intra)
{
    const unsigned int index_a = qp + a;
    const int alpha = alpha_table[index_a];
    const int beta  = beta_table[qp + b];
    if (alpha ==0 || beta == 0) return;

    if( bS[0] < 4 || !intra ) {
        int8_t tc[4];
        tc[0] = tc0_table[index_a][bS[0]];
        tc[1] = tc0_table[index_a][bS[1]];
        tc[2] = tc0_table[index_a][bS[2]];
        tc[3] = tc0_table[index_a][bS[3]];
        h->h264dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc);
    } else {
        h->h264dsp.h264_v_loop_filter_luma_intra(pix, stride, alpha, beta);
    }
}

static av_always_inline void filter_mb_edgech(uint8_t *pix, int stride,
                                              const int16_t bS[4],
                                              unsigned int qp, int a, int b,
                                              H264Context *h, int intra)
{
    const unsigned int index_a = qp + a;
    const int alpha = alpha_table[index_a];
    const int beta  = beta_table[qp + b];
    if (alpha ==0 || beta == 0) return;

    if( bS[0] < 4 || !intra ) {
        int8_t tc[4];
        tc[0] = tc0_table[index_a][bS[0]]+1;
        tc[1] = tc0_table[index_a][bS[1]]+1;
        tc[2] = tc0_table[index_a][bS[2]]+1;
        tc[3] = tc0_table[index_a][bS[3]]+1;
        h->h264dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc);
    } else {
        h->h264dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta);
    }
}

static av_always_inline void h264_filter_mb_fast_internal(H264Context *h,
                                                          int mb_x, int mb_y,
                                                          uint8_t *img_y,
                                                          uint8_t *img_cb,
                                                          uint8_t *img_cr,
                                                          unsigned int linesize,
                                                          unsigned int uvlinesize,
                                                          int pixel_shift)
{
    MpegEncContext * const s = &h->s;
    int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));
    int chroma444 = CHROMA444;
    int chroma422 = CHROMA422;

    int mb_xy = h->mb_xy;
    int left_type= h->left_type[LTOP];
    int top_type= h->top_type;

    int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
    int a = h->slice_alpha_c0_offset - qp_bd_offset;
    int b = h->slice_beta_offset - qp_bd_offset;

    int mb_type = s->current_picture.f.mb_type[mb_xy];
    int qp      = s->current_picture.f.qscale_table[mb_xy];
    int qp0     = s->current_picture.f.qscale_table[mb_xy - 1];
    int qp1     = s->current_picture.f.qscale_table[h->top_mb_xy];
    int qpc = get_chroma_qp( h, 0, qp );
    int qpc0 = get_chroma_qp( h, 0, qp0 );
    int qpc1 = get_chroma_qp( h, 0, qp1 );
    qp0 = (qp + qp0 + 1) >> 1;
    qp1 = (qp + qp1 + 1) >> 1;
    qpc0 = (qpc + qpc0 + 1) >> 1;
    qpc1 = (qpc + qpc1 + 1) >> 1;

    if( IS_INTRA(mb_type) ) {
        static const int16_t bS4[4] = {4,4,4,4};
        static const int16_t bS3[4] = {3,3,3,3};
        const int16_t *bSH = FIELD_PICTURE ? bS3 : bS4;
        if(left_type)
            filter_mb_edgev( &img_y[4*0<<pixel_shift], linesize, bS4, qp0, a, b, h, 1);
        if( IS_8x8DCT(mb_type) ) {
            filter_mb_edgev( &img_y[4*2<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
            if(top_type){
                filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1);
            }
            filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0);
        } else {
            filter_mb_edgev( &img_y[4*1<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
            filter_mb_edgev( &img_y[4*2<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
            filter_mb_edgev( &img_y[4*3<<pixel_shift], linesize, bS3, qp, a, b, h, 0);
            if(top_type){
                filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1);
            }
            filter_mb_edgeh( &img_y[4*1*linesize], linesize, bS3, qp, a, b, h, 0);
            filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0);
            filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, a, b, h, 0);
        }
        if(chroma){
            if(chroma444){
                if(left_type){
                    filter_mb_edgev( &img_cb[4*0<<pixel_shift], linesize, bS4, qpc0, a, b, h, 1);
                    filter_mb_edgev( &img_cr[4*0<<pixel_shift], linesize, bS4, qpc0, a, b, h, 1);
                }
                if( IS_8x8DCT(mb_type) ) {
                    filter_mb_edgev( &img_cb[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgev( &img_cr[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
                    if(top_type){
                        filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 );
                        filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 );
                    }
                    filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
                } else {
                    filter_mb_edgev( &img_cb[4*1<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgev( &img_cr[4*1<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgev( &img_cb[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgev( &img_cr[4*2<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgev( &img_cb[4*3<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgev( &img_cr[4*3<<pixel_shift], linesize, bS3, qpc, a, b, h, 0);
                    if(top_type){
                        filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1);
                        filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1);
                    }
                    filter_mb_edgeh( &img_cb[4*1*linesize], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgeh( &img_cr[4*1*linesize], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgeh( &img_cb[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);
                    filter_mb_edgeh( &img_cr[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);
                }
            }else if(chroma422){
                if(left_type){
                    filter_mb_edgecv(&img_cb[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
                    filter_mb_edgecv(&img_cr[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
                }
                filter_mb_edgecv(&img_cb[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
                filter_mb_edgecv(&img_cr[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
                if(top_type){
                    filter_mb_edgech(&img_cb[4*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
                    filter_mb_edgech(&img_cr[4*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
                }
                filter_mb_edgech(&img_cb[4*1*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
                filter_mb_edgech(&img_cr[4*1*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
                filter_mb_edgech(&img_cb[4*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
                filter_mb_edgech(&img_cr[4*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
                filter_mb_edgech(&img_cb[4*3*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
                filter_mb_edgech(&img_cr[4*3*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
            }else{
                if(left_type){
                    filter_mb_edgecv( &img_cb[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
                    filter_mb_edgecv( &img_cr[2*0<<pixel_shift], uvlinesize, bS4, qpc0, a, b, h, 1);
                }
                filter_mb_edgecv( &img_cb[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
                filter_mb_edgecv( &img_cr[2*2<<pixel_shift], uvlinesize, bS3, qpc, a, b, h, 0);
                if(top_type){
                    filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
                    filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);
                }
                filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
                filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);
            }
        }
        return;
    } else {
        LOCAL_ALIGNED_8(int16_t, bS, [2], [4][4]);
        int edges;
        if( IS_8x8DCT(mb_type) && (h->cbp&7) == 7 && !chroma444 ) {
            edges = 4;
            AV_WN64A(bS[0][0], 0x0002000200020002ULL);
            AV_WN64A(bS[0][2], 0x0002000200020002ULL);
            AV_WN64A(bS[1][0], 0x0002000200020002ULL);
            AV_WN64A(bS[1][2], 0x0002000200020002ULL);
        } else {
            int mask_edge1 = (3*(((5*mb_type)>>5)&1)) | (mb_type>>4); //(mb_type & (MB_TYPE_16x16 | MB_TYPE_8x16)) ? 3 : (mb_type & MB_TYPE_16x8) ? 1 : 0;
            int mask_edge0 = 3*((mask_edge1>>1) & ((5*left_type)>>5)&1); // (mb_type & (MB_TYPE_16x16 | MB_TYPE_8x16)) && (h->left_type[LTOP] & (MB_TYPE_16x16 | MB_TYPE_8x16)) ? 3 : 0;
            int step =  1+(mb_type>>24); //IS_8x8DCT(mb_type) ? 2 : 1;
            edges = 4 - 3*((mb_type>>3) & !(h->cbp & 15)); //(mb_type & MB_TYPE_16x16) && !(h->cbp & 15) ? 1 : 4;
            h->h264dsp.h264_loop_filter_strength( bS, h->non_zero_count_cache, h->ref_cache, h->mv_cache,
                                              h->list_count==2, edges, step, mask_edge0, mask_edge1, FIELD_PICTURE);
        }
        if( IS_INTRA(left_type) )
            AV_WN64A(bS[0][0], 0x0004000400040004ULL);
        if( IS_INTRA(top_type) )
            AV_WN64A(bS[1][0], FIELD_PICTURE ? 0x0003000300030003ULL : 0x0004000400040004ULL);

#define FILTER(hv,dir,edge,intra)\
        if(AV_RN64A(bS[dir][edge])) {                                   \
            filter_mb_edge##hv( &img_y[4*edge*(dir?linesize:1<<pixel_shift)], linesize, bS[dir][edge], edge ? qp : qp##dir, a, b, h, intra );\
            if(chroma){\
                if(chroma444){\
                    filter_mb_edge##hv( &img_cb[4*edge*(dir?linesize:1<<pixel_shift)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
                    filter_mb_edge##hv( &img_cr[4*edge*(dir?linesize:1<<pixel_shift)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
                } else if(!(edge&1)) {\
                    filter_mb_edgec##hv( &img_cb[2*edge*(dir?uvlinesize:1<<pixel_shift)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
                    filter_mb_edgec##hv( &img_cr[2*edge*(dir?uvlinesize:1<<pixel_shift)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\
                }\
            }\
        }
        if(left_type)
            FILTER(v,0,0,1);
        if( edges == 1 ) {
            if(top_type)
                FILTER(h,1,0,1);
        } else if( IS_8x8DCT(mb_type) ) {
            FILTER(v,0,2,0);
            if(top_type)
                FILTER(h,1,0,1);
            FILTER(h,1,2,0);
        } else {
            FILTER(v,0,1,0);
            FILTER(v,0,2,0);
            FILTER(v,0,3,0);
            if(top_type)
                FILTER(h,1,0,1);
            FILTER(h,1,1,0);
            FILTER(h,1,2,0);
            FILTER(h,1,3,0);
        }
#undef FILTER
    }
}

void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize) {
    assert(!FRAME_MBAFF);
    if(!h->h264dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff) {
        ff_h264_filter_mb(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize);
        return;
    }

#if CONFIG_SMALL
    h264_filter_mb_fast_internal(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, h->pixel_shift);
#else
    if(h->pixel_shift){
        h264_filter_mb_fast_internal(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, 1);
    }else{
        h264_filter_mb_fast_internal(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, 0);
    }
#endif
}

static int check_mv(H264Context *h, long b_idx, long bn_idx, int mvy_limit){
    int v;

    v= h->ref_cache[0][b_idx] != h->ref_cache[0][bn_idx];
    if(!v && h->ref_cache[0][b_idx]!=-1)
        v= h->mv_cache[0][b_idx][0] - h->mv_cache[0][bn_idx][0] + 3 >= 7U |
           FFABS( h->mv_cache[0][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= mvy_limit;

    if(h->list_count==2){
        if(!v)
            v = h->ref_cache[1][b_idx] != h->ref_cache[1][bn_idx] |
                h->mv_cache[1][b_idx][0] - h->mv_cache[1][bn_idx][0] + 3 >= 7U |
                FFABS( h->mv_cache[1][b_idx][1] - h->mv_cache[1][bn_idx][1] ) >= mvy_limit;

        if(v){
            if(h->ref_cache[0][b_idx] != h->ref_cache[1][bn_idx] |
               h->ref_cache[1][b_idx] != h->ref_cache[0][bn_idx])
                return 1;
            return
                h->mv_cache[0][b_idx][0] - h->mv_cache[1][bn_idx][0] + 3 >= 7U |
                FFABS( h->mv_cache[0][b_idx][1] - h->mv_cache[1][bn_idx][1] ) >= mvy_limit |
                h->mv_cache[1][b_idx][0] - h->mv_cache[0][bn_idx][0] + 3 >= 7U |
                FFABS( h->mv_cache[1][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= mvy_limit;
        }
    }

    return v;
}

static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize, int mb_xy, int mb_type, int mvy_limit, int first_vertical_edge_done, int a, int b, int chroma, int dir) {
    MpegEncContext * const s = &h->s;
    int edge;
    int chroma_qp_avg[2];
    int chroma444 = CHROMA444;
    int chroma422 = CHROMA422;
    const int mbm_xy = dir == 0 ? mb_xy -1 : h->top_mb_xy;
    const int mbm_type = dir == 0 ? h->left_type[LTOP] : h->top_type;

    // how often to recheck mv-based bS when iterating between edges
    static const uint8_t mask_edge_tab[2][8]={{0,3,3,3,1,1,1,1},
                                              {0,3,1,1,3,3,3,3}};
    const int mask_edge = mask_edge_tab[dir][(mb_type>>3)&7];
    const int edges = mask_edge== 3 && !(h->cbp&15) ? 1 : 4;

    // how often to recheck mv-based bS when iterating along each edge
    const int mask_par0 = mb_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir));

    if(mbm_type && !first_vertical_edge_done){

        if (FRAME_MBAFF && (dir == 1) && ((mb_y&1) == 0)
            && IS_INTERLACED(mbm_type&~mb_type)
            ) {
            // This is a special case in the norm where the filtering must
            // be done twice (one each of the field) even if we are in a
            // frame macroblock.
            //
            unsigned int tmp_linesize   = 2 *   linesize;
            unsigned int tmp_uvlinesize = 2 * uvlinesize;
            int mbn_xy = mb_xy - 2 * s->mb_stride;
            int j;

            for(j=0; j<2; j++, mbn_xy += s->mb_stride){
                DECLARE_ALIGNED(8, int16_t, bS)[4];
                int qp;
                if (IS_INTRA(mb_type | s->current_picture.f.mb_type[mbn_xy])) {
                    AV_WN64A(bS, 0x0003000300030003ULL);
                } else {
                    if (!CABAC && IS_8x8DCT(s->current_picture.f.mb_type[mbn_xy])) {
                        bS[0]= 1+((h->cbp_table[mbn_xy] & 0x4000)||h->non_zero_count_cache[scan8[0]+0]);
                        bS[1]= 1+((h->cbp_table[mbn_xy] & 0x4000)||h->non_zero_count_cache[scan8[0]+1]);
                        bS[2]= 1+((h->cbp_table[mbn_xy] & 0x8000)||h->non_zero_count_cache[scan8[0]+2]);
                        bS[3]= 1+((h->cbp_table[mbn_xy] & 0x8000)||h->non_zero_count_cache[scan8[0]+3]);
                    }else{
                    const uint8_t *mbn_nnz = h->non_zero_count[mbn_xy] + 3*4;
                    int i;
                    for( i = 0; i < 4; i++ ) {
                        bS[i] = 1 + !!(h->non_zero_count_cache[scan8[0]+i] | mbn_nnz[i]);
                    }
                    }
                }
                // Do not use s->qscale as luma quantizer because it has not the same
                // value in IPCM macroblocks.
                qp = (s->current_picture.f.qscale_table[mb_xy] + s->current_picture.f.qscale_table[mbn_xy] + 1) >> 1;
                tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize);
                { int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
                filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, a, b, h, 0 );
                chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1;
                chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1;
                if (chroma) {
                    if (chroma444) {
                        filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);
                        filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0);
                    } else {
                        filter_mb_edgech(&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);
                        filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0);
                    }
                }
            }
        }else{
            DECLARE_ALIGNED(8, int16_t, bS)[4];
            int qp;

            if( IS_INTRA(mb_type|mbm_type)) {
                AV_WN64A(bS, 0x0003000300030003ULL);
                if (   (!IS_INTERLACED(mb_type|mbm_type))
                    || ((FRAME_MBAFF || (s->picture_structure != PICT_FRAME)) && (dir == 0))
                )
                    AV_WN64A(bS, 0x0004000400040004ULL);
            } else {
                int i;
                int mv_done;

                if( dir && FRAME_MBAFF && IS_INTERLACED(mb_type ^ mbm_type)) {
                    AV_WN64A(bS, 0x0001000100010001ULL);
                    mv_done = 1;
                }
                else if( mask_par0 && ((mbm_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir)))) ) {
                    int b_idx= 8 + 4;
                    int bn_idx= b_idx - (dir ? 8:1);

                    bS[0] = bS[1] = bS[2] = bS[3] = check_mv(h, 8 + 4, bn_idx, mvy_limit);
                    mv_done = 1;
                }
                else
                    mv_done = 0;

                for( i = 0; i < 4; i++ ) {
                    int x = dir == 0 ? 0 : i;
                    int y = dir == 0 ? i    : 0;
                    int b_idx= 8 + 4 + x + 8*y;
                    int bn_idx= b_idx - (dir ? 8:1);

                    if( h->non_zero_count_cache[b_idx] |
                        h->non_zero_count_cache[bn_idx] ) {
                        bS[i] = 2;
                    }
                    else if(!mv_done)
                    {
                        bS[i] = check_mv(h, b_idx, bn_idx, mvy_limit);
                    }
                }
            }

            /* Filter edge */
            // Do not use s->qscale as luma quantizer because it has not the same
            // value in IPCM macroblocks.
            if(bS[0]+bS[1]+bS[2]+bS[3]){
                qp = (s->current_picture.f.qscale_table[mb_xy] + s->current_picture.f.qscale_table[mbm_xy] + 1) >> 1;
                tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
                chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1;
                chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1;
                if( dir == 0 ) {
                    filter_mb_edgev( &img_y[0], linesize, bS, qp, a, b, h, 1 );
                    if (chroma) {
                        if (chroma444) {
                            filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
                            filter_mb_edgev ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
                        } else {
                            filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
                            filter_mb_edgecv( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
                        }
                    }
                } else {
                    filter_mb_edgeh( &img_y[0], linesize, bS, qp, a, b, h, 1 );
                    if (chroma) {
                        if (chroma444) {
                            filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
                            filter_mb_edgeh ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
                        } else {
                            filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);
                            filter_mb_edgech( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);
                        }
                    }
                }
            }
        }
    }

    /* Calculate bS */
    for( edge = 1; edge < edges; edge++ ) {
        DECLARE_ALIGNED(8, int16_t, bS)[4];
        int qp;
        const int deblock_edge = !IS_8x8DCT(mb_type & (edge<<24)); // (edge&1) && IS_8x8DCT(mb_type)

        if (!deblock_edge && (!chroma422 || dir == 0))
            continue;

        if( IS_INTRA(mb_type)) {
            AV_WN64A(bS, 0x0003000300030003ULL);
        } else {
            int i;
            int mv_done;

            if( edge & mask_edge ) {
                AV_ZERO64(bS);
                mv_done = 1;
            }
            else if( mask_par0 ) {
                int b_idx= 8 + 4 + edge * (dir ? 8:1);
                int bn_idx= b_idx - (dir ? 8:1);

                bS[0] = bS[1] = bS[2] = bS[3] = check_mv(h, b_idx, bn_idx, mvy_limit);
                mv_done = 1;
            }
            else
                mv_done = 0;

            for( i = 0; i < 4; i++ ) {
                int x = dir == 0 ? edge : i;
                int y = dir == 0 ? i    : edge;
                int b_idx= 8 + 4 + x + 8*y;
                int bn_idx= b_idx - (dir ? 8:1);

                if( h->non_zero_count_cache[b_idx] |
                    h->non_zero_count_cache[bn_idx] ) {
                    bS[i] = 2;
                }
                else if(!mv_done)
                {
                    bS[i] = check_mv(h, b_idx, bn_idx, mvy_limit);
                }
            }

            if(bS[0]+bS[1]+bS[2]+bS[3] == 0)
                continue;
        }

        /* Filter edge */
        // Do not use s->qscale as luma quantizer because it has not the same
        // value in IPCM macroblocks.
        qp = s->current_picture.f.qscale_table[mb_xy];
        tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
        if( dir == 0 ) {
            filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, a, b, h, 0 );
            if (chroma) {
                if (chroma444) {
                    filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
                    filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
                } else if( (edge&1) == 0 ) {
                    filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
                    filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
                }
            }
        } else {
            if (chroma422) {
                if (deblock_edge)
                    filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0);
                if (chroma) {
                    filter_mb_edgech(&img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
                    filter_mb_edgech(&img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
                }
            } else {
                filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0);
                if (chroma) {
                    if (chroma444) {
                        filter_mb_edgeh (&img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
                        filter_mb_edgeh (&img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
                    } else if ((edge&1) == 0) {
                        filter_mb_edgech(&img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);
                        filter_mb_edgech(&img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);
                    }
                }
            }
        }
    }
}

void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize) {
    MpegEncContext * const s = &h->s;
    const int mb_xy= mb_x + mb_y*s->mb_stride;
    const int mb_type = s->current_picture.f.mb_type[mb_xy];
    const int mvy_limit = IS_INTERLACED(mb_type) ? 2 : 4;
    int first_vertical_edge_done = 0;
    av_unused int dir;
    int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));
    int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
    int a = h->slice_alpha_c0_offset - qp_bd_offset;
    int b = h->slice_beta_offset - qp_bd_offset;

    if (FRAME_MBAFF
            // and current and left pair do not have the same interlaced type
            && IS_INTERLACED(mb_type^h->left_type[LTOP])
            // and left mb is in available to us
            && h->left_type[LTOP]) {
        /* First vertical edge is different in MBAFF frames
         * There are 8 different bS to compute and 2 different Qp
         */
        DECLARE_ALIGNED(8, int16_t, bS)[8];
        int qp[2];
        int bqp[2];
        int rqp[2];
        int mb_qp, mbn0_qp, mbn1_qp;
        int i;
        first_vertical_edge_done = 1;

        if( IS_INTRA(mb_type) ) {
            AV_WN64A(&bS[0], 0x0004000400040004ULL);
            AV_WN64A(&bS[4], 0x0004000400040004ULL);
        } else {
            static const uint8_t offset[2][2][8]={
                {
                    {3+4*0, 3+4*0, 3+4*0, 3+4*0, 3+4*1, 3+4*1, 3+4*1, 3+4*1},
                    {3+4*2, 3+4*2, 3+4*2, 3+4*2, 3+4*3, 3+4*3, 3+4*3, 3+4*3},
                },{
                    {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},
                    {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},
                }
            };
            const uint8_t *off= offset[MB_FIELD][mb_y&1];
            for( i = 0; i < 8; i++ ) {
                int j= MB_FIELD ? i>>2 : i&1;
                int mbn_xy = h->left_mb_xy[LEFT(j)];
                int mbn_type= h->left_type[LEFT(j)];

                if( IS_INTRA( mbn_type ) )
                    bS[i] = 4;
                else{
                    bS[i] = 1 + !!(h->non_zero_count_cache[12+8*(i>>1)] |
                         ((!h->pps.cabac && IS_8x8DCT(mbn_type)) ?
                            (h->cbp_table[mbn_xy] & (((MB_FIELD ? (i&2) : (mb_y&1)) ? 8 : 2) << 12))
                                                                       :
                            h->non_zero_count[mbn_xy][ off[i] ]));
                }
            }
        }

        mb_qp   = s->current_picture.f.qscale_table[mb_xy];
        mbn0_qp = s->current_picture.f.qscale_table[h->left_mb_xy[0]];
        mbn1_qp = s->current_picture.f.qscale_table[h->left_mb_xy[1]];
        qp[0] = ( mb_qp + mbn0_qp + 1 ) >> 1;
        bqp[0] = ( get_chroma_qp( h, 0, mb_qp ) +
                   get_chroma_qp( h, 0, mbn0_qp ) + 1 ) >> 1;
        rqp[0] = ( get_chroma_qp( h, 1, mb_qp ) +
                   get_chroma_qp( h, 1, mbn0_qp ) + 1 ) >> 1;
        qp[1] = ( mb_qp + mbn1_qp + 1 ) >> 1;
        bqp[1] = ( get_chroma_qp( h, 0, mb_qp ) +
                   get_chroma_qp( h, 0, mbn1_qp ) + 1 ) >> 1;
        rqp[1] = ( get_chroma_qp( h, 1, mb_qp ) +
                   get_chroma_qp( h, 1, mbn1_qp ) + 1 ) >> 1;

        /* Filter edge */
        tprintf(s->avctx, "filter mb:%d/%d MBAFF, QPy:%d/%d, QPb:%d/%d QPr:%d/%d ls:%d uvls:%d", mb_x, mb_y, qp[0], qp[1], bqp[0], bqp[1], rqp[0], rqp[1], linesize, uvlinesize);
        { int i; for (i = 0; i < 8; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
        if(MB_FIELD){
            filter_mb_mbaff_edgev ( h, img_y                ,   linesize, bS  , 1, qp [0], a, b, 1 );
            filter_mb_mbaff_edgev ( h, img_y  + 8*  linesize,   linesize, bS+4, 1, qp [1], a, b, 1 );
            if (chroma){
                if (CHROMA444) {
                    filter_mb_mbaff_edgev ( h, img_cb,                uvlinesize, bS  , 1, bqp[0], a, b, 1 );
                    filter_mb_mbaff_edgev ( h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );
                    filter_mb_mbaff_edgev ( h, img_cr,                uvlinesize, bS  , 1, rqp[0], a, b, 1 );
                    filter_mb_mbaff_edgev ( h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );
                } else if (CHROMA422) {
                    filter_mb_mbaff_edgecv(h, img_cb,                uvlinesize, bS  , 1, bqp[0], a, b, 1);
                    filter_mb_mbaff_edgecv(h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1);
                    filter_mb_mbaff_edgecv(h, img_cr,                uvlinesize, bS  , 1, rqp[0], a, b, 1);
                    filter_mb_mbaff_edgecv(h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1);
                }else{
                    filter_mb_mbaff_edgecv( h, img_cb,                uvlinesize, bS  , 1, bqp[0], a, b, 1 );
                    filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );
                    filter_mb_mbaff_edgecv( h, img_cr,                uvlinesize, bS  , 1, rqp[0], a, b, 1 );
                    filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );
                }
            }
        }else{
            filter_mb_mbaff_edgev ( h, img_y              , 2*  linesize, bS  , 2, qp [0], a, b, 1 );
            filter_mb_mbaff_edgev ( h, img_y  +   linesize, 2*  linesize, bS+1, 2, qp [1], a, b, 1 );
            if (chroma){
                if (CHROMA444) {
                    filter_mb_mbaff_edgev ( h, img_cb,              2*uvlinesize, bS  , 2, bqp[0], a, b, 1 );
                    filter_mb_mbaff_edgev ( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 );
                    filter_mb_mbaff_edgev ( h, img_cr,              2*uvlinesize, bS  , 2, rqp[0], a, b, 1 );
                    filter_mb_mbaff_edgev ( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 );
                }else{
                    filter_mb_mbaff_edgecv( h, img_cb,              2*uvlinesize, bS  , 2, bqp[0], a, b, 1 );
                    filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 );
                    filter_mb_mbaff_edgecv( h, img_cr,              2*uvlinesize, bS  , 2, rqp[0], a, b, 1 );
                    filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 );
                }
            }
        }
    }

#if CONFIG_SMALL
    for( dir = 0; dir < 2; dir++ )
        filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, dir ? 0 : first_vertical_edge_done, a, b, chroma, dir);
#else
    filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, first_vertical_edge_done, a, b, chroma, 0);
    filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0,                        a, b, chroma, 1);
#endif
}