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.
29107 Commits
32 Branches
449MB
Tree: 7c9079ab4c
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '7c9079ab4c'
${ noResults }
FFmpeg/libavcodec/h264_loopfilter.c

705 lines
34KB
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. static void av_always_inline filter_mb_edgev( uint8_t *pix, int stride, int16_t bS[4], unsigned int qp, H264Context *h) {

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

  105.     const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  106.     const int alpha = alpha_table[index_a];

  107.     const int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

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

  109. 

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

  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, int16_t bS[4], unsigned int qp, H264Context *h ) {

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

  123.     const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  124.     const int alpha = alpha_table[index_a];

  125.     const int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

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

  127. 

  128.     if( bS[0] < 4 ) {

  129.         int8_t tc[4];

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

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

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

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

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

  135.     } else {

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

  137.     }

  138. }

  139. 

  140. static void filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix, int stride, int16_t bS[7], int bsi, int qp ) {

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

  142.     int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  143.     int alpha = alpha_table[index_a];

  144.     int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

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

  146. 

  147.     if( bS[0] < 4 ) {

  148.         int8_t tc[4];

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

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

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

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

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

  154.     } else {

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

  156.     }

  157. }

  158. static void filter_mb_mbaff_edgecv( H264Context *h, uint8_t *pix, int stride, int16_t bS[7], int bsi, int qp ) {

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

  160.     int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  161.     int alpha = alpha_table[index_a];

  162.     int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

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

  164. 

  165.     if( bS[0] < 4 ) {

  166.         int8_t tc[4];

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

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

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

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

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

  172.     } else {

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

  174.     }

  175. }

  176. 

  177. static void av_always_inline filter_mb_edgeh( uint8_t *pix, int stride, int16_t bS[4], unsigned int qp, H264Context *h ) {

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

  179.     const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  180.     const int alpha = alpha_table[index_a];

  181.     const int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

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

  183. 

  184.     if( bS[0] < 4 ) {

  185.         int8_t tc[4];

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

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

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

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

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

  191.     } else {

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

  193.     }

  194. }

  195. 

  196. static void av_always_inline filter_mb_edgech( uint8_t *pix, int stride, int16_t bS[4], unsigned int qp, H264Context *h ) {

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

  198.     const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;

  199.     const int alpha = alpha_table[index_a];

  200.     const int beta  = beta_table[qp - qp_bd_offset + h->slice_beta_offset];

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

  202. 

  203.     if( bS[0] < 4 ) {

  204.         int8_t tc[4];

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

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

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

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

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

  210.     } else {

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

  212.     }

  213. }

  214. 

  215. 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) {

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

  217.     int mb_xy;

  218.     int mb_type, left_type;

  219.     int qp, qp0, qp1, qpc, qpc0, qpc1, qp_thresh;

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

  221. 

  222.     mb_xy = h->mb_xy;

  223. 

  224.     if(!h->top_type || !h->h264dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff || CHROMA444) {

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

  226.         return;

  227.     }

  228.     assert(!FRAME_MBAFF);

  229.     left_type= h->left_type[0];

  230. 

  231.     mb_type = s->current_picture.mb_type[mb_xy];

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

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

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

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

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

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

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

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

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

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

  242.     qp_thresh = 15+52 - h->slice_alpha_c0_offset;

  243.     if(qp <= qp_thresh && qp0 <= qp_thresh && qp1 <= qp_thresh &&

  244.        qpc <= qp_thresh && qpc0 <= qp_thresh && qpc1 <= qp_thresh)

  245.         return;

  246. 

  247.     if( IS_INTRA(mb_type) ) {

  248.         int16_t bS4[4] = {4,4,4,4};

  249.         int16_t bS3[4] = {3,3,3,3};

  250.         int16_t *bSH = FIELD_PICTURE ? bS3 : bS4;

  251.         if(left_type)

  252.             filter_mb_edgev( &img_y[4*0], linesize, bS4, qp0, h);

  253.         if( IS_8x8DCT(mb_type) ) {

  254.             filter_mb_edgev( &img_y[4*2], linesize, bS3, qp, h);

  255.             filter_mb_edgeh( &img_y[4*0*linesize], linesize, bSH, qp1, h);

  256.             filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, h);

  257.         } else {

  258.             filter_mb_edgev( &img_y[4*1], linesize, bS3, qp, h);

  259.             filter_mb_edgev( &img_y[4*2], linesize, bS3, qp, h);

  260.             filter_mb_edgev( &img_y[4*3], linesize, bS3, qp, h);

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

  262.             filter_mb_edgeh( &img_y[4*1*linesize], linesize, bS3, qp, h);

  263.             filter_mb_edgeh( &img_y[4*2*linesize], linesize, bS3, qp, h);

  264.             filter_mb_edgeh( &img_y[4*3*linesize], linesize, bS3, qp, h);

  265.         }

  266.         if(chroma){

  267.             if(left_type){

  268.                 filter_mb_edgecv( &img_cb[2*0], uvlinesize, bS4, qpc0, h);

  269.                 filter_mb_edgecv( &img_cr[2*0], uvlinesize, bS4, qpc0, h);

  270.             }

  271.             filter_mb_edgecv( &img_cb[2*2], uvlinesize, bS3, qpc, h);

  272.             filter_mb_edgecv( &img_cr[2*2], uvlinesize, bS3, qpc, h);

  273.             filter_mb_edgech( &img_cb[2*0*uvlinesize], uvlinesize, bSH, qpc1, h);

  274.             filter_mb_edgech( &img_cb[2*2*uvlinesize], uvlinesize, bS3, qpc, h);

  275.             filter_mb_edgech( &img_cr[2*0*uvlinesize], uvlinesize, bSH, qpc1, h);

  276.             filter_mb_edgech( &img_cr[2*2*uvlinesize], uvlinesize, bS3, qpc, h);

  277.         }

  278.         return;

  279.     } else {

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

  281.         int edges;

  282.         if( IS_8x8DCT(mb_type) && (h->cbp&7) == 7 ) {

  283.             edges = 4;

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

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

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

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

  288.         } else {

  289.             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;

  290.             int mask_edge0 = 3*((mask_edge1>>1) & ((5*left_type)>>5)&1); // (mb_type & (MB_TYPE_16x16 | MB_TYPE_8x16)) && (h->left_type[0] & (MB_TYPE_16x16 | MB_TYPE_8x16)) ? 3 : 0;

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

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

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

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

  295.         }

  296.         if( IS_INTRA(left_type) )

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

  298.         if( IS_INTRA(h->top_type) )

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

  300. 

  301. #define FILTER(hv,dir,edge)\

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

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

  304.             if(chroma && !(edge&1)) {\

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

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

  307.             }\

  308.         }

  309.         if(left_type)

  310.             FILTER(v,0,0);

  311.         if( edges == 1 ) {

  312.             FILTER(h,1,0);

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

  314.             FILTER(v,0,2);

  315.             FILTER(h,1,0);

  316.             FILTER(h,1,2);

  317.         } else {

  318.             FILTER(v,0,1);

  319.             FILTER(v,0,2);

  320.             FILTER(v,0,3);

  321.             FILTER(h,1,0);

  322.             FILTER(h,1,1);

  323.             FILTER(h,1,2);

  324.             FILTER(h,1,3);

  325.         }

  326. #undef FILTER

  327.     }

  328. }

  329. 

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

  331.     int v;

  332. 

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

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

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

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

  337. 

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

  339.         if(!v)

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

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

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

  343. 

  344.         if(v){

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

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

  347.                 return 1;

  348.             return

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

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

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

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

  353.         }

  354.     }

  355. 

  356.     return v;

  357. }

  358. 

  359. 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 chroma, int chroma444, int dir) {

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

  361.     int edge;

  362.     int chroma_qp_avg[2];

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

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

  365. 

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

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

  368.                                               {0,3,1,1,3,3,3,3}};

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

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

  371. 

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

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

  374. 

  375.     if(mbm_type && !first_vertical_edge_done){

  376. 

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

  378.             && IS_INTERLACED(mbm_type&~mb_type)

  379.             ) {

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

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

  382.             // frame macroblock.

  383.             //

  384.             unsigned int tmp_linesize   = 2 *   linesize;

  385.             unsigned int tmp_uvlinesize = 2 * uvlinesize;

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

  387.             int j;

  388. 

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

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

  391.                 int qp;

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

  393.                     AV_WN64A(bS, 0x0003000300030003ULL);

  394.                 } else {

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

  396.                         bS[0]= 1+((h->cbp_table[mbn_xy] & 4)||h->non_zero_count_cache[scan8[0]+0]);

  397.                         bS[1]= 1+((h->cbp_table[mbn_xy] & 4)||h->non_zero_count_cache[scan8[0]+1]);

  398.                         bS[2]= 1+((h->cbp_table[mbn_xy] & 8)||h->non_zero_count_cache[scan8[0]+2]);

  399.                         bS[3]= 1+((h->cbp_table[mbn_xy] & 8)||h->non_zero_count_cache[scan8[0]+3]);

  400.                     }else{

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

  402.                     int i;

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

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

  405.                     }

  406.                     }

  407.                 }

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

  409.                 // value in IPCM macroblocks.

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

  411.                 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);

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

  413.                 filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, h );

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

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

  416.                 if (chroma) {

  417.                     if (chroma444) {

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

  419.                         filter_mb_edgeh (&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], h);

  420.                     } else {

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

  422.                         filter_mb_edgech(&img_cr[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[1], h);

  423.                     }

  424.                 }

  425.             }

  426.         }else{

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

  428.             int qp;

  429. 

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

  431.                 AV_WN64A(bS, 0x0003000300030003ULL);

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

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

  434.                 )

  435.                     AV_WN64A(bS, 0x0004000400040004ULL);

  436.             } else {

  437.                 int i;

  438.                 int mv_done;

  439. 

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

  441.                     AV_WN64A(bS, 0x0001000100010001ULL);

  442.                     mv_done = 1;

  443.                 }

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

  445.                     int b_idx= 8 + 4;

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

  447. 

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

  449.                     mv_done = 1;

  450.                 }

  451.                 else

  452.                     mv_done = 0;

  453. 

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

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

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

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

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

  459. 

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

  461.                         h->non_zero_count_cache[bn_idx] ) {

  462.                         bS[i] = 2;

  463.                     }

  464.                     else if(!mv_done)

  465.                     {

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

  467.                     }

  468.                 }

  469.             }

  470. 

  471.             /* Filter edge */

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

  473.             // value in IPCM macroblocks.

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

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

  476.                 //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]);

  477.                 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);

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

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

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

  481.                 if( dir == 0 ) {

  482.                     filter_mb_edgev( &img_y[0], linesize, bS, qp, h );

  483.                     if (chroma) {

  484.                         if (chroma444) {

  485.                             filter_mb_edgev ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h);

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

  487.                         } else {

  488.                             filter_mb_edgecv( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h);

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

  490.                         }

  491.                     }

  492.                 } else {

  493.                     filter_mb_edgeh( &img_y[0], linesize, bS, qp, h );

  494.                     if (chroma) {

  495.                         if (chroma444) {

  496.                             filter_mb_edgeh ( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h);

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

  498.                         } else {

  499.                             filter_mb_edgech( &img_cb[0], uvlinesize, bS, chroma_qp_avg[0], h);

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

  501.                         }

  502.                     }

  503.                 }

  504.             }

  505.         }

  506.     }

  507. 

  508.     /* Calculate bS */

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

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

  511.         int qp;

  512. 

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

  514.             continue;

  515. 

  516.         if( IS_INTRA(mb_type)) {

  517.             AV_WN64A(bS, 0x0003000300030003ULL);

  518.         } else {

  519.             int i;

  520.             int mv_done;

  521. 

  522.             if( edge & mask_edge ) {

  523.                 AV_ZERO64(bS);

  524.                 mv_done = 1;

  525.             }

  526.             else if( mask_par0 ) {

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

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

  529. 

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

  531.                 mv_done = 1;

  532.             }

  533.             else

  534.                 mv_done = 0;

  535. 

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

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

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

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

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

  541. 

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

  543.                     h->non_zero_count_cache[bn_idx] ) {

  544.                     bS[i] = 2;

  545.                 }

  546.                 else if(!mv_done)

  547.                 {

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

  549.                 }

  550.             }

  551. 

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

  553.                 continue;

  554.         }

  555. 

  556.         /* Filter edge */

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

  558.         // value in IPCM macroblocks.

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

  560.         //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]);

  561.         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);

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

  563.         if( dir == 0 ) {

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

  565.             if (chroma) {

  566.                 if (chroma444) {

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

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

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

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

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

  572.                 }

  573.             }

  574.         } else {

  575.             filter_mb_edgeh( &img_y[4*edge*linesize], linesize, bS, qp, h );

  576.             if (chroma) {

  577.                 if (chroma444) {

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

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

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

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

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

  583.                 }

  584.             }

  585.         }

  586.     }

  587. }

  588. 

  589. 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) {

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

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

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

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

  594.     int first_vertical_edge_done = 0;

  595.     av_unused int dir;

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

  597. 

  598.     if (FRAME_MBAFF

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

  600.             && IS_INTERLACED(mb_type^h->left_type[0])

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

  602.             && h->left_type[0]) {

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

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

  605.          */

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

  607.         int qp[2];

  608.         int bqp[2];

  609.         int rqp[2];

  610.         int mb_qp, mbn0_qp, mbn1_qp;

  611.         int i;

  612.         first_vertical_edge_done = 1;

  613. 

  614.         if( IS_INTRA(mb_type) ) {

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

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

  617.         } else {

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

  619.                 {

  620.                     {3+4*0, 3+4*0, 3+4*0, 3+4*0, 3+4*1, 3+4*1, 3+4*1, 3+4*1},

  621.                     {3+4*2, 3+4*2, 3+4*2, 3+4*2, 3+4*3, 3+4*3, 3+4*3, 3+4*3},

  622.                 },{

  623.                     {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},

  624.                     {3+4*0, 3+4*1, 3+4*2, 3+4*3, 3+4*0, 3+4*1, 3+4*2, 3+4*3},

  625.                 }

  626.             };

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

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

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

  630.                 int mbn_xy = h->left_mb_xy[j];

  631.                 int mbn_type= h->left_type[j];

  632. 

  633.                 if( IS_INTRA( mbn_type ) )

  634.                     bS[i] = 4;

  635.                 else{

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

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

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

  639.                                                                        :

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

  641.                 }

  642.             }

  643.         }

  644. 

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

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

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

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

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

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

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

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

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

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

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

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

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

  658. 

  659.         /* Filter edge */

  660.         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);

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

  662.         if(MB_FIELD){

  663.             filter_mb_mbaff_edgev ( h, img_y                ,   linesize, bS  , 1, qp [0] );

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

  665.             if (chroma){

  666.                 if (CHROMA444) {

  667.                     filter_mb_mbaff_edgev ( h, img_cb,                uvlinesize, bS  , 1, bqp[0] );

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

  669.                     filter_mb_mbaff_edgev ( h, img_cr,                uvlinesize, bS  , 1, rqp[0] );

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

  671.                 }else{

  672.                     filter_mb_mbaff_edgecv( h, img_cb,                uvlinesize, bS  , 1, bqp[0] );

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

  674.                     filter_mb_mbaff_edgecv( h, img_cr,                uvlinesize, bS  , 1, rqp[0] );

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

  676.                 }

  677.             }

  678.         }else{

  679.             filter_mb_mbaff_edgev ( h, img_y              , 2*  linesize, bS  , 2, qp [0] );

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

  681.             if (chroma){

  682.                 if (CHROMA444) {

  683.                     filter_mb_mbaff_edgev ( h, img_cb,              2*uvlinesize, bS  , 2, bqp[0] );

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

  685.                     filter_mb_mbaff_edgev ( h, img_cr,              2*uvlinesize, bS  , 2, rqp[0] );

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

  687.                 }else{

  688.                     filter_mb_mbaff_edgecv( h, img_cb,              2*uvlinesize, bS  , 2, bqp[0] );

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

  690.                     filter_mb_mbaff_edgecv( h, img_cr,              2*uvlinesize, bS  , 2, rqp[0] );

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

  692.                 }

  693.             }

  694.         }

  695.     }

  696. 

  697. #if CONFIG_SMALL

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

  699.         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, chroma, CHROMA444, dir);

  700. #else

  701.     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, chroma, CHROMA444, 0);

  702.     filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, chroma, CHROMA444, 1);

  703. #endif

  704. }