falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
29422 Commits
32 Branches
291MB
Tree: 7ce914fb5a
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '7ce914fb5a'
${ noResults }
FFmpeg/libavcodec/h264_loopfilter.c

753 lines
37KB
Raw Blame History 

  1. /*

  2.  * H.26L/H.264/AVC/JVT/14496-10/... loop filter

  3.  * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>

  4.  *

  5.  * This file is part of Libav.

  6.  *

  7.  * Libav is free software; you can redistribute it and/or

  8.  * modify it under the terms of the GNU Lesser General Public

  9.  * License as published by the Free Software Foundation; either

  10.  * version 2.1 of the License, or (at your option) any later version.

  11.  *

  12.  * Libav is distributed in the hope that it will be useful,

  13.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  14.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  15.  * Lesser General Public License for more details.

  16.  *

  17.  * You should have received a copy of the GNU Lesser General Public

  18.  * License along with Libav; if not, write to the Free Software

  19.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  20.  */

  21. 

  22. /**

  23.  * @file

  24.  * H.264 / AVC / MPEG4 part10 loop filter.

  25.  * @author Michael Niedermayer <michaelni@gmx.at>

  26.  */

  27. 

  28. #include "libavutil/intreadwrite.h"

  29. #include "internal.h"

  30. #include "dsputil.h"

  31. #include "avcodec.h"

  32. #include "mpegvideo.h"

  33. #include "h264.h"

  34. #include "mathops.h"

  35. #include "rectangle.h"

  36. 

  37. //#undef NDEBUG

  38. #include <assert.h>

  39. 

  40. /* Deblocking filter (p153) */

  41. static const uint8_t alpha_table[52*3] = {

  42.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  43.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  44.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  45.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  46.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  47.      0,  0,  0,  0,  0,  0,  4,  4,  5,  6,

  48.      7,  8,  9, 10, 12, 13, 15, 17, 20, 22,

  49.     25, 28, 32, 36, 40, 45, 50, 56, 63, 71,

  50.     80, 90,101,113,127,144,162,182,203,226,

  51.    255,255,

  52.    255,255,255,255,255,255,255,255,255,255,255,255,255,

  53.    255,255,255,255,255,255,255,255,255,255,255,255,255,

  54.    255,255,255,255,255,255,255,255,255,255,255,255,255,

  55.    255,255,255,255,255,255,255,255,255,255,255,255,255,

  56. };

  57. static const uint8_t beta_table[52*3] = {

  58.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  59.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  60.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  61.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  62.      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,

  63.      0,  0,  0,  0,  0,  0,  2,  2,  2,  3,

  64.      3,  3,  3,  4,  4,  4,  6,  6,  7,  7,

  65.      8,  8,  9,  9, 10, 10, 11, 11, 12, 12,

  66.     13, 13, 14, 14, 15, 15, 16, 16, 17, 17,

  67.     18, 18,

  68.     18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,

  69.     18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,

  70.     18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,

  71.     18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,

  72. };

  73. static const uint8_t tc0_table[52*3][4] = {

  74.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  75.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  76.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  77.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  78.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  79.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  80.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  81.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  82.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  83.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  84.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },

  85.     {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 1 },

  86.     {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 1, 1 }, {-1, 0, 1, 1 }, {-1, 1, 1, 1 },

  87.     {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 },

  88.     {-1, 1, 1, 2 }, {-1, 1, 2, 3 }, {-1, 1, 2, 3 }, {-1, 2, 2, 3 }, {-1, 2, 2, 4 }, {-1, 2, 3, 4 },

  89.     {-1, 2, 3, 4 }, {-1, 3, 3, 5 }, {-1, 3, 4, 6 }, {-1, 3, 4, 6 }, {-1, 4, 5, 7 }, {-1, 4, 5, 8 },

  90.     {-1, 4, 6, 9 }, {-1, 5, 7,10 }, {-1, 6, 8,11 }, {-1, 6, 8,13 }, {-1, 7,10,14 }, {-1, 8,11,16 },

  91.     {-1, 9,12,18 }, {-1,10,13,20 }, {-1,11,15,23 }, {-1,13,17,25 },

  92.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  93.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  94.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  95.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  96.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  97.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  98.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  99.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  100.     {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },

  101. };

  102. 

  103. /* intra: 0 if this loopfilter call is guaranteed to be inter (bS < 4), 1 if it might be intra (bS == 4) */

  104. static void av_always_inline filter_mb_edgev( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra ) {

  105.     const unsigned int index_a = qp + a;

  106.     const int alpha = alpha_table[index_a];

  107.     const int beta  = beta_table[qp + b];

  108.     if (alpha ==0 || beta == 0) return;

  109. 

  110.     if( bS[0] < 4 || !intra ) {

  111.         int8_t tc[4];

  112.         tc[0] = tc0_table[index_a][bS[0]];

  113.         tc[1] = tc0_table[index_a][bS[1]];

  114.         tc[2] = tc0_table[index_a][bS[2]];

  115.         tc[3] = tc0_table[index_a][bS[3]];

  116.         h->h264dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc);

  117.     } else {

  118.         h->h264dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta);

  119.     }

  120. }

  121. static void av_always_inline filter_mb_edgecv( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra ) {

  122.     const unsigned int index_a = qp + a;

  123.     const int alpha = alpha_table[index_a];

  124.     const int beta  = beta_table[qp + b];

  125.     if (alpha ==0 || beta == 0) return;

  126. 

  127.     if( bS[0] < 4 || !intra ) {

  128.         int8_t tc[4];

  129.         tc[0] = tc0_table[index_a][bS[0]]+1;

  130.         tc[1] = tc0_table[index_a][bS[1]]+1;

  131.         tc[2] = tc0_table[index_a][bS[2]]+1;

  132.         tc[3] = tc0_table[index_a][bS[3]]+1;

  133.         h->h264dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc);

  134.     } else {

  135.         h->h264dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta);

  136.     }

  137. }

  138. 

  139. static void av_always_inline 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 ) {

  140.     const unsigned int index_a = qp + a;

  141.     const int alpha = alpha_table[index_a];

  142.     const int beta  = beta_table[qp + b];

  143.     if (alpha ==0 || beta == 0) return;

  144. 

  145.     if( bS[0] < 4 || !intra ) {

  146.         int8_t tc[4];

  147.         tc[0] = tc0_table[index_a][bS[0*bsi]];

  148.         tc[1] = tc0_table[index_a][bS[1*bsi]];

  149.         tc[2] = tc0_table[index_a][bS[2*bsi]];

  150.         tc[3] = tc0_table[index_a][bS[3*bsi]];

  151.         h->h264dsp.h264_h_loop_filter_luma_mbaff(pix, stride, alpha, beta, tc);

  152.     } else {

  153.         h->h264dsp.h264_h_loop_filter_luma_mbaff_intra(pix, stride, alpha, beta);

  154.     }

  155. }

  156. static void av_always_inline 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 ) {

  157.     const unsigned int index_a = qp + a;

  158.     const int alpha = alpha_table[index_a];

  159.     const int beta  = beta_table[qp + b];

  160.     if (alpha ==0 || beta == 0) return;

  161. 

  162.     if( bS[0] < 4 || !intra ) {

  163.         int8_t tc[4];

  164.         tc[0] = tc0_table[index_a][bS[0*bsi]] + 1;

  165.         tc[1] = tc0_table[index_a][bS[1*bsi]] + 1;

  166.         tc[2] = tc0_table[index_a][bS[2*bsi]] + 1;

  167.         tc[3] = tc0_table[index_a][bS[3*bsi]] + 1;

  168.         h->h264dsp.h264_h_loop_filter_chroma_mbaff(pix, stride, alpha, beta, tc);

  169.     } else {

  170.         h->h264dsp.h264_h_loop_filter_chroma_mbaff_intra(pix, stride, alpha, beta);

  171.     }

  172. }

  173. 

  174. static void av_always_inline filter_mb_edgeh( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra ) {

  175.     const unsigned int index_a = qp + a;

  176.     const int alpha = alpha_table[index_a];

  177.     const int beta  = beta_table[qp + b];

  178.     if (alpha ==0 || beta == 0) return;

  179. 

  180.     if( bS[0] < 4 || !intra ) {

  181.         int8_t tc[4];

  182.         tc[0] = tc0_table[index_a][bS[0]];

  183.         tc[1] = tc0_table[index_a][bS[1]];

  184.         tc[2] = tc0_table[index_a][bS[2]];

  185.         tc[3] = tc0_table[index_a][bS[3]];

  186.         h->h264dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc);

  187.     } else {

  188.         h->h264dsp.h264_v_loop_filter_luma_intra(pix, stride, alpha, beta);

  189.     }

  190. }

  191. 

  192. static void av_always_inline filter_mb_edgech( uint8_t *pix, int stride, const int16_t bS[4], unsigned int qp, int a, int b, H264Context *h, int intra ) {

  193.     const unsigned int index_a = qp + a;

  194.     const int alpha = alpha_table[index_a];

  195.     const int beta  = beta_table[qp + b];

  196.     if (alpha ==0 || beta == 0) return;

  197. 

  198.     if( bS[0] < 4 || !intra ) {

  199.         int8_t tc[4];

  200.         tc[0] = tc0_table[index_a][bS[0]]+1;

  201.         tc[1] = tc0_table[index_a][bS[1]]+1;

  202.         tc[2] = tc0_table[index_a][bS[2]]+1;

  203.         tc[3] = tc0_table[index_a][bS[3]]+1;

  204.         h->h264dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc);

  205.     } else {

  206.         h->h264dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta);

  207.     }

  208. }

  209. 

  210. 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) {

  211.     MpegEncContext * const s = &h->s;

  212.     int mb_xy;

  213.     int mb_type, left_type, top_type;

  214.     int qp, qp0, qp1, qpc, qpc0, qpc1;

  215.     int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));

  216.     int chroma444 = CHROMA444;

  217.     int qp_bd_offset, a, b;

  218. 

  219.     mb_xy = h->mb_xy;

  220. 

  221.     if(!h->h264dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff) {

  222.         ff_h264_filter_mb(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize);

  223.         return;

  224.     }

  225.     assert(!FRAME_MBAFF);

  226.     left_type= h->left_type[LTOP];

  227.     top_type= h->top_type;

  228. 

  229.     mb_type = s->current_picture.f.mb_type[mb_xy];

  230.     qp      = s->current_picture.f.qscale_table[mb_xy];

  231.     qp0     = s->current_picture.f.qscale_table[mb_xy - 1];

  232.     qp1     = s->current_picture.f.qscale_table[h->top_mb_xy];

  233.     qpc = get_chroma_qp( h, 0, qp );

  234.     qpc0 = get_chroma_qp( h, 0, qp0 );

  235.     qpc1 = get_chroma_qp( h, 0, qp1 );

  236.     qp0 = (qp + qp0 + 1) >> 1;

  237.     qp1 = (qp + qp1 + 1) >> 1;

  238.     qpc0 = (qpc + qpc0 + 1) >> 1;

  239.     qpc1 = (qpc + qpc1 + 1) >> 1;

  240.     qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);

  241.     a = h->slice_alpha_c0_offset - qp_bd_offset;

  242.     b = h->slice_beta_offset - qp_bd_offset;

  243. 

  244.     if( IS_INTRA(mb_type) ) {

  245.         static const int16_t bS4[4] = {4,4,4,4};

  246.         static const int16_t bS3[4] = {3,3,3,3};

  247.         const int16_t *bSH = FIELD_PICTURE ? bS3 : bS4;

  248.         if(left_type)

  249.             filter_mb_edgev( &img_y[4*0], linesize, bS4, qp0, a, b, h, 1);

  250.         if( IS_8x8DCT(mb_type) ) {

  251.             filter_mb_edgev( &img_y[4*2], linesize, bS3, qp, a, b, h, 0);

  252.             if(top_type){

  253.                 filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1);

  254.             }

  255.             filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0);

  256.         } else {

  257.             filter_mb_edgev( &img_y[4*1], linesize, bS3, qp, a, b, h, 0);

  258.             filter_mb_edgev( &img_y[4*2], linesize, bS3, qp, a, b, h, 0);

  259.             filter_mb_edgev( &img_y[4*3], linesize, bS3, qp, a, b, h, 0);

  260.             if(top_type){

  261.                 filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, a, b, h, 1);

  262.             }

  263.             filter_mb_edgeh( &img_y[4*1*linesize], linesize, bS3, qp, a, b, h, 0);

  264.             filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, a, b, h, 0);

  265.             filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, a, b, h, 0);

  266.         }

  267.         if(chroma){

  268.             if(chroma444){

  269.                 if(left_type){

  270.                     filter_mb_edgev( &img_cb[4*0], linesize, bS4, qpc0, a, b, h, 1);

  271.                     filter_mb_edgev( &img_cr[4*0], linesize, bS4, qpc0, a, b, h, 1);

  272.                 }

  273.                 if( IS_8x8DCT(mb_type) ) {

  274.                     filter_mb_edgev( &img_cb[4*2], linesize, bS3, qpc, a, b, h, 0);

  275.                     filter_mb_edgev( &img_cr[4*2], linesize, bS3, qpc, a, b, h, 0);

  276.                     if(top_type){

  277.                         filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 );

  278.                         filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1 );

  279.                     }

  280.                     filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);

  281.                     filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);

  282.                 } else {

  283.                     filter_mb_edgev( &img_cb[4*1], linesize, bS3, qpc, a, b, h, 0);

  284.                     filter_mb_edgev( &img_cr[4*1], linesize, bS3, qpc, a, b, h, 0);

  285.                     filter_mb_edgev( &img_cb[4*2], linesize, bS3, qpc, a, b, h, 0);

  286.                     filter_mb_edgev( &img_cr[4*2], linesize, bS3, qpc, a, b, h, 0);

  287.                     filter_mb_edgev( &img_cb[4*3], linesize, bS3, qpc, a, b, h, 0);

  288.                     filter_mb_edgev( &img_cr[4*3], linesize, bS3, qpc, a, b, h, 0);

  289.                     if(top_type){

  290.                         filter_mb_edgeh( &img_cb[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1);

  291.                         filter_mb_edgeh( &img_cr[4*0*linesize], linesize, bSH, qpc1, a, b, h, 1);

  292.                     }

  293.                     filter_mb_edgeh( &img_cb[4*1*linesize], linesize, bS3, qpc, a, b, h, 0);

  294.                     filter_mb_edgeh( &img_cr[4*1*linesize], linesize, bS3, qpc, a, b, h, 0);

  295.                     filter_mb_edgeh( &img_cb[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);

  296.                     filter_mb_edgeh( &img_cr[4*2*linesize], linesize, bS3, qpc, a, b, h, 0);

  297.                     filter_mb_edgeh( &img_cb[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);

  298.                     filter_mb_edgeh( &img_cr[4*3*linesize], linesize, bS3, qpc, a, b, h, 0);

  299.                 }

  300.             }else{

  301.                 if(left_type){

  302.                     filter_mb_edgecv( &img_cb[2*0], uvlinesize, bS4, qpc0, a, b, h, 1);

  303.                     filter_mb_edgecv( &img_cr[2*0], uvlinesize, bS4, qpc0, a, b, h, 1);

  304.                 }

  305.                 filter_mb_edgecv( &img_cb[2*2], uvlinesize, bS3, qpc, a, b, h, 0);

  306.                 filter_mb_edgecv( &img_cr[2*2], uvlinesize, bS3, qpc, a, b, h, 0);

  307.                 if(top_type){

  308.                     filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);

  309.                     filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, a, b, h, 1);

  310.                 }

  311.                 filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);

  312.                 filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, a, b, h, 0);

  313.             }

  314.         }

  315.         return;

  316.     } else {

  317.         LOCAL_ALIGNED_8(int16_t, bS, [2], [4][4]);

  318.         int edges;

  319.         if( IS_8x8DCT(mb_type) && (h->cbp&7) == 7 && !chroma444 ) {

  320.             edges = 4;

  321.             AV_WN64A(bS[0][0], 0x0002000200020002ULL);

  322.             AV_WN64A(bS[0][2], 0x0002000200020002ULL);

  323.             AV_WN64A(bS[1][0], 0x0002000200020002ULL);

  324.             AV_WN64A(bS[1][2], 0x0002000200020002ULL);

  325.         } else {

  326.             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;

  327.             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;

  328.             int step =  1+(mb_type>>24); //IS_8x8DCT(mb_type) ? 2 : 1;

  329.             edges = 4 - 3*((mb_type>>3) & !(h->cbp & 15)); //(mb_type & MB_TYPE_16x16) && !(h->cbp & 15) ? 1 : 4;

  330.             h->h264dsp.h264_loop_filter_strength( bS, h->non_zero_count_cache, h->ref_cache, h->mv_cache,

  331.                                               h->list_count==2, edges, step, mask_edge0, mask_edge1, FIELD_PICTURE);

  332.         }

  333.         if( IS_INTRA(left_type) )

  334.             AV_WN64A(bS[0][0], 0x0004000400040004ULL);

  335.         if( IS_INTRA(top_type) )

  336.             AV_WN64A(bS[1][0], FIELD_PICTURE ? 0x0003000300030003ULL : 0x0004000400040004ULL);

  337. 

  338. #define FILTER(hv,dir,edge,intra)\

  339.         if(AV_RN64A(bS[dir][edge])) {                                   \

  340.             filter_mb_edge##hv( &img_y[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qp : qp##dir, a, b, h, intra );\

  341.             if(chroma){\

  342.                 if(chroma444){\

  343.                     filter_mb_edge##hv( &img_cb[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\

  344.                     filter_mb_edge##hv( &img_cr[4*edge*(dir?linesize:1)], linesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\

  345.                 } else if(!(edge&1)) {\

  346.                     filter_mb_edgec##hv( &img_cb[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\

  347.                     filter_mb_edgec##hv( &img_cr[2*edge*(dir?uvlinesize:1)], uvlinesize, bS[dir][edge], edge ? qpc : qpc##dir, a, b, h, intra );\

  348.                 }\

  349.             }\

  350.         }

  351.         if(left_type)

  352.             FILTER(v,0,0,1);

  353.         if( edges == 1 ) {

  354.             if(top_type)

  355.                 FILTER(h,1,0,1);

  356.         } else if( IS_8x8DCT(mb_type) ) {

  357.             FILTER(v,0,2,0);

  358.             if(top_type)

  359.                 FILTER(h,1,0,1);

  360.             FILTER(h,1,2,0);

  361.         } else {

  362.             FILTER(v,0,1,0);

  363.             FILTER(v,0,2,0);

  364.             FILTER(v,0,3,0);

  365.             if(top_type)

  366.                 FILTER(h,1,0,1);

  367.             FILTER(h,1,1,0);

  368.             FILTER(h,1,2,0);

  369.             FILTER(h,1,3,0);

  370.         }

  371. #undef FILTER

  372.     }

  373. }

  374. 

  375. static int check_mv(H264Context *h, long b_idx, long bn_idx, int mvy_limit){

  376.     int v;

  377. 

  378.     v= h->ref_cache[0][b_idx] != h->ref_cache[0][bn_idx];

  379.     if(!v && h->ref_cache[0][b_idx]!=-1)

  380.         v= h->mv_cache[0][b_idx][0] - h->mv_cache[0][bn_idx][0] + 3 >= 7U |

  381.            FFABS( h->mv_cache[0][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= mvy_limit;

  382. 

  383.     if(h->list_count==2){

  384.         if(!v)

  385.             v = h->ref_cache[1][b_idx] != h->ref_cache[1][bn_idx] |

  386.                 h->mv_cache[1][b_idx][0] - h->mv_cache[1][bn_idx][0] + 3 >= 7U |

  387.                 FFABS( h->mv_cache[1][b_idx][1] - h->mv_cache[1][bn_idx][1] ) >= mvy_limit;

  388. 

  389.         if(v){

  390.             if(h->ref_cache[0][b_idx] != h->ref_cache[1][bn_idx] |

  391.                h->ref_cache[1][b_idx] != h->ref_cache[0][bn_idx])

  392.                 return 1;

  393.             return

  394.                 h->mv_cache[0][b_idx][0] - h->mv_cache[1][bn_idx][0] + 3 >= 7U |

  395.                 FFABS( h->mv_cache[0][b_idx][1] - h->mv_cache[1][bn_idx][1] ) >= mvy_limit |

  396.                 h->mv_cache[1][b_idx][0] - h->mv_cache[0][bn_idx][0] + 3 >= 7U |

  397.                 FFABS( h->mv_cache[1][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= mvy_limit;

  398.         }

  399.     }

  400. 

  401.     return v;

  402. }

  403. 

  404. 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 chroma444, int dir) {

  405.     MpegEncContext * const s = &h->s;

  406.     int edge;

  407.     int chroma_qp_avg[2];

  408.     const int mbm_xy = dir == 0 ? mb_xy -1 : h->top_mb_xy;

  409.     const int mbm_type = dir == 0 ? h->left_type[LTOP] : h->top_type;

  410. 

  411.     // how often to recheck mv-based bS when iterating between edges

  412.     static const uint8_t mask_edge_tab[2][8]={{0,3,3,3,1,1,1,1},

  413.                                               {0,3,1,1,3,3,3,3}};

  414.     const int mask_edge = mask_edge_tab[dir][(mb_type>>3)&7];

  415.     const int edges = mask_edge== 3 && !(h->cbp&15) ? 1 : 4;

  416. 

  417.     // how often to recheck mv-based bS when iterating along each edge

  418.     const int mask_par0 = mb_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir));

  419. 

  420.     if(mbm_type && !first_vertical_edge_done){

  421. 

  422.         if (FRAME_MBAFF && (dir == 1) && ((mb_y&1) == 0)

  423.             && IS_INTERLACED(mbm_type&~mb_type)

  424.             ) {

  425.             // This is a special case in the norm where the filtering must

  426.             // be done twice (one each of the field) even if we are in a

  427.             // frame macroblock.

  428.             //

  429.             unsigned int tmp_linesize   = 2 *   linesize;

  430.             unsigned int tmp_uvlinesize = 2 * uvlinesize;

  431.             int mbn_xy = mb_xy - 2 * s->mb_stride;

  432.             int j;

  433. 

  434.             for(j=0; j<2; j++, mbn_xy += s->mb_stride){

  435.                 DECLARE_ALIGNED(8, int16_t, bS)[4];

  436.                 int qp;

  437.                 if (IS_INTRA(mb_type | s->current_picture.f.mb_type[mbn_xy])) {

  438.                     AV_WN64A(bS, 0x0003000300030003ULL);

  439.                 } else {

  440.                     if (!CABAC && IS_8x8DCT(s->current_picture.f.mb_type[mbn_xy])) {

  441.                         bS[0]= 1+((h->cbp_table[mbn_xy] & 0x4000)||h->non_zero_count_cache[scan8[0]+0]);

  442.                         bS[1]= 1+((h->cbp_table[mbn_xy] & 0x4000)||h->non_zero_count_cache[scan8[0]+1]);

  443.                         bS[2]= 1+((h->cbp_table[mbn_xy] & 0x8000)||h->non_zero_count_cache[scan8[0]+2]);

  444.                         bS[3]= 1+((h->cbp_table[mbn_xy] & 0x8000)||h->non_zero_count_cache[scan8[0]+3]);

  445.                     }else{

  446.                     const uint8_t *mbn_nnz = h->non_zero_count[mbn_xy] + 3*4;

  447.                     int i;

  448.                     for( i = 0; i < 4; i++ ) {

  449.                         bS[i] = 1 + !!(h->non_zero_count_cache[scan8[0]+i] | mbn_nnz[i]);

  450.                     }

  451.                     }

  452.                 }

  453.                 // Do not use s->qscale as luma quantizer because it has not the same

  454.                 // value in IPCM macroblocks.

  455.                 qp = (s->current_picture.f.qscale_table[mb_xy] + s->current_picture.f.qscale_table[mbn_xy] + 1) >> 1;

  456.                 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);

  457.                 { int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }

  458.                 filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, a, b, h, 0 );

  459.                 chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1;

  460.                 chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbn_xy]) + 1) >> 1;

  461.                 if (chroma) {

  462.                     if (chroma444) {

  463.                         filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);

  464.                         filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0);

  465.                     } else {

  466.                         filter_mb_edgech(&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0);

  467.                         filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], a, b, h, 0);

  468.                     }

  469.                 }

  470.             }

  471.         }else{

  472.             DECLARE_ALIGNED(8, int16_t, bS)[4];

  473.             int qp;

  474. 

  475.             if( IS_INTRA(mb_type|mbm_type)) {

  476.                 AV_WN64A(bS, 0x0003000300030003ULL);

  477.                 if (   (!IS_INTERLACED(mb_type|mbm_type))

  478.                     || ((FRAME_MBAFF || (s->picture_structure != PICT_FRAME)) && (dir == 0))

  479.                 )

  480.                     AV_WN64A(bS, 0x0004000400040004ULL);

  481.             } else {

  482.                 int i;

  483.                 int mv_done;

  484. 

  485.                 if( dir && FRAME_MBAFF && IS_INTERLACED(mb_type ^ mbm_type)) {

  486.                     AV_WN64A(bS, 0x0001000100010001ULL);

  487.                     mv_done = 1;

  488.                 }

  489.                 else if( mask_par0 && ((mbm_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir)))) ) {

  490.                     int b_idx= 8 + 4;

  491.                     int bn_idx= b_idx - (dir ? 8:1);

  492. 

  493.                     bS[0] = bS[1] = bS[2] = bS[3] = check_mv(h, 8 + 4, bn_idx, mvy_limit);

  494.                     mv_done = 1;

  495.                 }

  496.                 else

  497.                     mv_done = 0;

  498. 

  499.                 for( i = 0; i < 4; i++ ) {

  500.                     int x = dir == 0 ? 0 : i;

  501.                     int y = dir == 0 ? i    : 0;

  502.                     int b_idx= 8 + 4 + x + 8*y;

  503.                     int bn_idx= b_idx - (dir ? 8:1);

  504. 

  505.                     if( h->non_zero_count_cache[b_idx] |

  506.                         h->non_zero_count_cache[bn_idx] ) {

  507.                         bS[i] = 2;

  508.                     }

  509.                     else if(!mv_done)

  510.                     {

  511.                         bS[i] = check_mv(h, b_idx, bn_idx, mvy_limit);

  512.                     }

  513.                 }

  514.             }

  515. 

  516.             /* Filter edge */

  517.             // Do not use s->qscale as luma quantizer because it has not the same

  518.             // value in IPCM macroblocks.

  519.             if(bS[0]+bS[1]+bS[2]+bS[3]){

  520.                 qp = (s->current_picture.f.qscale_table[mb_xy] + s->current_picture.f.qscale_table[mbm_xy] + 1) >> 1;

  521.                 //tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp[0], s->current_picture.qscale_table[mbn_xy]);

  522.                 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);

  523.                 //{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }

  524.                 chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1;

  525.                 chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mbm_xy]) + 1) >> 1;

  526.                 if( dir == 0 ) {

  527.                     filter_mb_edgev( &img_y[0], linesize, bS, qp, a, b, h, 1 );

  528.                     if (chroma) {

  529.                         if (chroma444) {

  530.                             filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);

  531.                             filter_mb_edgev ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);

  532.                         } else {

  533.                             filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);

  534.                             filter_mb_edgecv( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);

  535.                         }

  536.                     }

  537.                 } else {

  538.                     filter_mb_edgeh( &img_y[0], linesize, bS, qp, a, b, h, 1 );

  539.                     if (chroma) {

  540.                         if (chroma444) {

  541.                             filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);

  542.                             filter_mb_edgeh ( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);

  543.                         } else {

  544.                             filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], a, b, h, 1);

  545.                             filter_mb_edgech( &img_cr[0], uvlinesize, bS, chroma_qp_avg[1], a, b, h, 1);

  546.                         }

  547.                     }

  548.                 }

  549.             }

  550.         }

  551.     }

  552. 

  553.     /* Calculate bS */

  554.     for( edge = 1; edge < edges; edge++ ) {

  555.         DECLARE_ALIGNED(8, int16_t, bS)[4];

  556.         int qp;

  557. 

  558.         if( IS_8x8DCT(mb_type & (edge<<24)) ) // (edge&1) && IS_8x8DCT(mb_type)

  559.             continue;

  560. 

  561.         if( IS_INTRA(mb_type)) {

  562.             AV_WN64A(bS, 0x0003000300030003ULL);

  563.         } else {

  564.             int i;

  565.             int mv_done;

  566. 

  567.             if( edge & mask_edge ) {

  568.                 AV_ZERO64(bS);

  569.                 mv_done = 1;

  570.             }

  571.             else if( mask_par0 ) {

  572.                 int b_idx= 8 + 4 + edge * (dir ? 8:1);

  573.                 int bn_idx= b_idx - (dir ? 8:1);

  574. 

  575.                 bS[0] = bS[1] = bS[2] = bS[3] = check_mv(h, b_idx, bn_idx, mvy_limit);

  576.                 mv_done = 1;

  577.             }

  578.             else

  579.                 mv_done = 0;

  580. 

  581.             for( i = 0; i < 4; i++ ) {

  582.                 int x = dir == 0 ? edge : i;

  583.                 int y = dir == 0 ? i    : edge;

  584.                 int b_idx= 8 + 4 + x + 8*y;

  585.                 int bn_idx= b_idx - (dir ? 8:1);

  586. 

  587.                 if( h->non_zero_count_cache[b_idx] |

  588.                     h->non_zero_count_cache[bn_idx] ) {

  589.                     bS[i] = 2;

  590.                 }

  591.                 else if(!mv_done)

  592.                 {

  593.                     bS[i] = check_mv(h, b_idx, bn_idx, mvy_limit);

  594.                 }

  595.             }

  596. 

  597.             if(bS[0]+bS[1]+bS[2]+bS[3] == 0)

  598.                 continue;

  599.         }

  600. 

  601.         /* Filter edge */

  602.         // Do not use s->qscale as luma quantizer because it has not the same

  603.         // value in IPCM macroblocks.

  604.         qp = s->current_picture.f.qscale_table[mb_xy];

  605.         //tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp[0], s->current_picture.qscale_table[mbn_xy]);

  606.         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);

  607.         //{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }

  608.         if( dir == 0 ) {

  609.             filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, a, b, h, 0 );

  610.             if (chroma) {

  611.                 if (chroma444) {

  612.                     filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);

  613.                     filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);

  614.                 } else if( (edge&1) == 0 ) {

  615.                     filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);

  616.                     filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);

  617.                 }

  618.             }

  619.         } else {

  620.             filter_mb_edgeh( &img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0 );

  621.             if (chroma) {

  622.                 if (chroma444) {

  623.                     filter_mb_edgeh ( &img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);

  624.                     filter_mb_edgeh ( &img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);

  625.                 } else if( (edge&1) == 0 ) {

  626.                     filter_mb_edgech( &img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0);

  627.                     filter_mb_edgech( &img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0);

  628.                 }

  629.             }

  630.         }

  631.     }

  632. }

  633. 

  634. 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) {

  635.     MpegEncContext * const s = &h->s;

  636.     const int mb_xy= mb_x + mb_y*s->mb_stride;

  637.     const int mb_type = s->current_picture.f.mb_type[mb_xy];

  638.     const int mvy_limit = IS_INTERLACED(mb_type) ? 2 : 4;

  639.     int first_vertical_edge_done = 0;

  640.     av_unused int dir;

  641.     int chroma = !(CONFIG_GRAY && (s->flags&CODEC_FLAG_GRAY));

  642.     int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);

  643.     int a = h->slice_alpha_c0_offset - qp_bd_offset;

  644.     int b = h->slice_beta_offset - qp_bd_offset;

  645. 

  646.     if (FRAME_MBAFF

  647.             // and current and left pair do not have the same interlaced type

  648.             && IS_INTERLACED(mb_type^h->left_type[LTOP])

  649.             // and left mb is in available to us

  650.             && h->left_type[LTOP]) {

  651.         /* First vertical edge is different in MBAFF frames

  652.          * There are 8 different bS to compute and 2 different Qp

  653.          */

  654.         DECLARE_ALIGNED(8, int16_t, bS)[8];

  655.         int qp[2];

  656.         int bqp[2];

  657.         int rqp[2];

  658.         int mb_qp, mbn0_qp, mbn1_qp;

  659.         int i;

  660.         first_vertical_edge_done = 1;

  661. 

  662.         if( IS_INTRA(mb_type) ) {

  663.             AV_WN64A(&bS[0], 0x0004000400040004ULL);

  664.             AV_WN64A(&bS[4], 0x0004000400040004ULL);

  665.         } else {

  666.             static const uint8_t offset[2][2][8]={

  667.                 {

  668.                     {3+4*0, 3+4*0, 3+4*0, 3+4*0, 3+4*1, 3+4*1, 3+4*1, 3+4*1},

  669.                     {3+4*2, 3+4*2, 3+4*2, 3+4*2, 3+4*3, 3+4*3, 3+4*3, 3+4*3},

  670.                 },{

  671.                     {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},

  672.                     {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},

  673.                 }

  674.             };

  675.             const uint8_t *off= offset[MB_FIELD][mb_y&1];

  676.             for( i = 0; i < 8; i++ ) {

  677.                 int j= MB_FIELD ? i>>2 : i&1;

  678.                 int mbn_xy = h->left_mb_xy[LEFT(j)];

  679.                 int mbn_type= h->left_type[LEFT(j)];

  680. 

  681.                 if( IS_INTRA( mbn_type ) )

  682.                     bS[i] = 4;

  683.                 else{

  684.                     bS[i] = 1 + !!(h->non_zero_count_cache[12+8*(i>>1)] |

  685.                          ((!h->pps.cabac && IS_8x8DCT(mbn_type)) ?

  686.                             (h->cbp_table[mbn_xy] & (((MB_FIELD ? (i&2) : (mb_y&1)) ? 8 : 2) << 12))

  687.                                                                        :

  688.                             h->non_zero_count[mbn_xy][ off[i] ]));

  689.                 }

  690.             }

  691.         }

  692. 

  693.         mb_qp   = s->current_picture.f.qscale_table[mb_xy];

  694.         mbn0_qp = s->current_picture.f.qscale_table[h->left_mb_xy[0]];

  695.         mbn1_qp = s->current_picture.f.qscale_table[h->left_mb_xy[1]];

  696.         qp[0] = ( mb_qp + mbn0_qp + 1 ) >> 1;

  697.         bqp[0] = ( get_chroma_qp( h, 0, mb_qp ) +

  698.                    get_chroma_qp( h, 0, mbn0_qp ) + 1 ) >> 1;

  699.         rqp[0] = ( get_chroma_qp( h, 1, mb_qp ) +

  700.                    get_chroma_qp( h, 1, mbn0_qp ) + 1 ) >> 1;

  701.         qp[1] = ( mb_qp + mbn1_qp + 1 ) >> 1;

  702.         bqp[1] = ( get_chroma_qp( h, 0, mb_qp ) +

  703.                    get_chroma_qp( h, 0, mbn1_qp ) + 1 ) >> 1;

  704.         rqp[1] = ( get_chroma_qp( h, 1, mb_qp ) +

  705.                    get_chroma_qp( h, 1, mbn1_qp ) + 1 ) >> 1;

  706. 

  707.         /* Filter edge */

  708.         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);

  709.         { int i; for (i = 0; i < 8; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }

  710.         if(MB_FIELD){

  711.             filter_mb_mbaff_edgev ( h, img_y                ,   linesize, bS  , 1, qp [0], a, b, 1 );

  712.             filter_mb_mbaff_edgev ( h, img_y  + 8*  linesize,   linesize, bS+4, 1, qp [1], a, b, 1 );

  713.             if (chroma){

  714.                 if (CHROMA444) {

  715.                     filter_mb_mbaff_edgev ( h, img_cb,                uvlinesize, bS  , 1, bqp[0], a, b, 1 );

  716.                     filter_mb_mbaff_edgev ( h, img_cb + 8*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );

  717.                     filter_mb_mbaff_edgev ( h, img_cr,                uvlinesize, bS  , 1, rqp[0], a, b, 1 );

  718.                     filter_mb_mbaff_edgev ( h, img_cr + 8*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );

  719.                 }else{

  720.                     filter_mb_mbaff_edgecv( h, img_cb,                uvlinesize, bS  , 1, bqp[0], a, b, 1 );

  721.                     filter_mb_mbaff_edgecv( h, img_cb + 4*uvlinesize, uvlinesize, bS+4, 1, bqp[1], a, b, 1 );

  722.                     filter_mb_mbaff_edgecv( h, img_cr,                uvlinesize, bS  , 1, rqp[0], a, b, 1 );

  723.                     filter_mb_mbaff_edgecv( h, img_cr + 4*uvlinesize, uvlinesize, bS+4, 1, rqp[1], a, b, 1 );

  724.                 }

  725.             }

  726.         }else{

  727.             filter_mb_mbaff_edgev ( h, img_y              , 2*  linesize, bS  , 2, qp [0], a, b, 1 );

  728.             filter_mb_mbaff_edgev ( h, img_y  +   linesize, 2*  linesize, bS+1, 2, qp [1], a, b, 1 );

  729.             if (chroma){

  730.                 if (CHROMA444) {

  731.                     filter_mb_mbaff_edgev ( h, img_cb,              2*uvlinesize, bS  , 2, bqp[0], a, b, 1 );

  732.                     filter_mb_mbaff_edgev ( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 );

  733.                     filter_mb_mbaff_edgev ( h, img_cr,              2*uvlinesize, bS  , 2, rqp[0], a, b, 1 );

  734.                     filter_mb_mbaff_edgev ( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 );

  735.                 }else{

  736.                     filter_mb_mbaff_edgecv( h, img_cb,              2*uvlinesize, bS  , 2, bqp[0], a, b, 1 );

  737.                     filter_mb_mbaff_edgecv( h, img_cb + uvlinesize, 2*uvlinesize, bS+1, 2, bqp[1], a, b, 1 );

  738.                     filter_mb_mbaff_edgecv( h, img_cr,              2*uvlinesize, bS  , 2, rqp[0], a, b, 1 );

  739.                     filter_mb_mbaff_edgecv( h, img_cr + uvlinesize, 2*uvlinesize, bS+1, 2, rqp[1], a, b, 1 );

  740.                 }

  741.             }

  742.         }

  743.     }

  744. 

  745. #if CONFIG_SMALL

  746.     for( dir = 0; dir < 2; dir++ )

  747.         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, CHROMA444, dir);

  748. #else

  749.     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, CHROMA444, 0);

  750.     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, CHROMA444, 1);

  751. #endif

  752. }