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.
19904 Commits
32 Branches
824MB
Tree: ecc1330a6d
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'ecc1330a6d'
${ noResults }
FFmpeg/libavcodec/rv40.c

683 lines
26KB
Raw Blame History 

  1. /*

  2.  * RV40 decoder

  3.  * Copyright (c) 2007 Konstantin Shishkov

  4.  *

  5.  * This file is part of FFmpeg.

  6.  *

  7.  * FFmpeg 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.  * FFmpeg 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 FFmpeg; 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 libavcodec/rv40.c

  24.  * RV40 decoder

  25.  */

  26. 

  27. #include "avcodec.h"

  28. #include "dsputil.h"

  29. #include "mpegvideo.h"

  30. #include "golomb.h"

  31. 

  32. #include "rv34.h"

  33. #include "rv40vlc2.h"

  34. #include "rv40data.h"

  35. 

  36. static VLC aic_top_vlc;

  37. static VLC aic_mode1_vlc[AIC_MODE1_NUM], aic_mode2_vlc[AIC_MODE2_NUM];

  38. static VLC ptype_vlc[NUM_PTYPE_VLCS], btype_vlc[NUM_BTYPE_VLCS];

  39. 

  40. static const int16_t mode2_offs[] = {

  41.        0,  614, 1222, 1794, 2410,  3014,  3586,  4202,  4792, 5382, 5966, 6542,

  42.     7138, 7716, 8292, 8864, 9444, 10030, 10642, 11212, 11814

  43. };

  44. 

  45. /**

  46.  * Initialize all tables.

  47.  */

  48. static av_cold void rv40_init_tables(void)

  49. {

  50.     int i;

  51.     static VLC_TYPE aic_table[1 << AIC_TOP_BITS][2];

  52.     static VLC_TYPE aic_mode1_table[AIC_MODE1_NUM << AIC_MODE1_BITS][2];

  53.     static VLC_TYPE aic_mode2_table[11814][2];

  54.     static VLC_TYPE ptype_table[NUM_PTYPE_VLCS << PTYPE_VLC_BITS][2];

  55.     static VLC_TYPE btype_table[NUM_BTYPE_VLCS << BTYPE_VLC_BITS][2];

  56. 

  57.     aic_top_vlc.table = aic_table;

  58.     aic_top_vlc.table_allocated = 1 << AIC_TOP_BITS;

  59.     init_vlc(&aic_top_vlc, AIC_TOP_BITS, AIC_TOP_SIZE,

  60.              rv40_aic_top_vlc_bits,  1, 1,

  61.              rv40_aic_top_vlc_codes, 1, 1, INIT_VLC_USE_NEW_STATIC);

  62.     for(i = 0; i < AIC_MODE1_NUM; i++){

  63.         // Every tenth VLC table is empty

  64.         if((i % 10) == 9) continue;

  65.         aic_mode1_vlc[i].table = &aic_mode1_table[i << AIC_MODE1_BITS];

  66.         aic_mode1_vlc[i].table_allocated = 1 << AIC_MODE1_BITS;

  67.         init_vlc(&aic_mode1_vlc[i], AIC_MODE1_BITS, AIC_MODE1_SIZE,

  68.                  aic_mode1_vlc_bits[i],  1, 1,

  69.                  aic_mode1_vlc_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);

  70.     }

  71.     for(i = 0; i < AIC_MODE2_NUM; i++){

  72.         aic_mode2_vlc[i].table = &aic_mode2_table[mode2_offs[i]];

  73.         aic_mode2_vlc[i].table_allocated = mode2_offs[i + 1] - mode2_offs[i];

  74.         init_vlc(&aic_mode2_vlc[i], AIC_MODE2_BITS, AIC_MODE2_SIZE,

  75.                  aic_mode2_vlc_bits[i],  1, 1,

  76.                  aic_mode2_vlc_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);

  77.     }

  78.     for(i = 0; i < NUM_PTYPE_VLCS; i++){

  79.         ptype_vlc[i].table = &ptype_table[i << PTYPE_VLC_BITS];

  80.         ptype_vlc[i].table_allocated = 1 << PTYPE_VLC_BITS;

  81.         init_vlc_sparse(&ptype_vlc[i], PTYPE_VLC_BITS, PTYPE_VLC_SIZE,

  82.                          ptype_vlc_bits[i],  1, 1,

  83.                          ptype_vlc_codes[i], 1, 1,

  84.                          ptype_vlc_syms,     1, 1, INIT_VLC_USE_NEW_STATIC);

  85.     }

  86.     for(i = 0; i < NUM_BTYPE_VLCS; i++){

  87.         btype_vlc[i].table = &btype_table[i << BTYPE_VLC_BITS];

  88.         btype_vlc[i].table_allocated = 1 << BTYPE_VLC_BITS;

  89.         init_vlc_sparse(&btype_vlc[i], BTYPE_VLC_BITS, BTYPE_VLC_SIZE,

  90.                          btype_vlc_bits[i],  1, 1,

  91.                          btype_vlc_codes[i], 1, 1,

  92.                          btype_vlc_syms,     1, 1, INIT_VLC_USE_NEW_STATIC);

  93.     }

  94. }

  95. 

  96. /**

  97.  * Get stored dimension from bitstream.

  98.  *

  99.  * If the width/height is the standard one then it's coded as a 3-bit index.

  100.  * Otherwise it is coded as escaped 8-bit portions.

  101.  */

  102. static int get_dimension(GetBitContext *gb, const int *dim)

  103. {

  104.     int t   = get_bits(gb, 3);

  105.     int val = dim[t];

  106.     if(val < 0)

  107.         val = dim[get_bits1(gb) - val];

  108.     if(!val){

  109.         do{

  110.             t = get_bits(gb, 8);

  111.             val += t << 2;

  112.         }while(t == 0xFF);

  113.     }

  114.     return val;

  115. }

  116. 

  117. /**

  118.  * Get encoded picture size - usually this is called from rv40_parse_slice_header.

  119.  */

  120. static void rv40_parse_picture_size(GetBitContext *gb, int *w, int *h)

  121. {

  122.     *w = get_dimension(gb, rv40_standard_widths);

  123.     *h = get_dimension(gb, rv40_standard_heights);

  124. }

  125. 

  126. static int rv40_parse_slice_header(RV34DecContext *r, GetBitContext *gb, SliceInfo *si)

  127. {

  128.     int mb_bits;

  129.     int w = r->s.width, h = r->s.height;

  130.     int mb_size;

  131. 

  132.     memset(si, 0, sizeof(SliceInfo));

  133.     if(get_bits1(gb))

  134.         return -1;

  135.     si->type = get_bits(gb, 2);

  136.     if(si->type == 1) si->type = 0;

  137.     si->quant = get_bits(gb, 5);

  138.     if(get_bits(gb, 2))

  139.         return -1;

  140.     si->vlc_set = get_bits(gb, 2);

  141.     skip_bits1(gb);

  142.     si->pts = get_bits(gb, 13);

  143.     if(!si->type || !get_bits1(gb))

  144.         rv40_parse_picture_size(gb, &w, &h);

  145.     if(avcodec_check_dimensions(r->s.avctx, w, h) < 0)

  146.         return -1;

  147.     si->width  = w;

  148.     si->height = h;

  149.     mb_size = ((w + 15) >> 4) * ((h + 15) >> 4);

  150.     mb_bits = ff_rv34_get_start_offset(gb, mb_size);

  151.     si->start = get_bits(gb, mb_bits);

  152. 

  153.     return 0;

  154. }

  155. 

  156. /**

  157.  * Decode 4x4 intra types array.

  158.  */

  159. static int rv40_decode_intra_types(RV34DecContext *r, GetBitContext *gb, int8_t *dst)

  160. {

  161.     MpegEncContext *s = &r->s;

  162.     int i, j, k, v;

  163.     int A, B, C;

  164.     int pattern;

  165.     int8_t *ptr;

  166. 

  167.     for(i = 0; i < 4; i++, dst += s->b4_stride){

  168.         if(!i && s->first_slice_line){

  169.             pattern = get_vlc2(gb, aic_top_vlc.table, AIC_TOP_BITS, 1);

  170.             dst[0] = (pattern >> 2) & 2;

  171.             dst[1] = (pattern >> 1) & 2;

  172.             dst[2] =  pattern       & 2;

  173.             dst[3] = (pattern << 1) & 2;

  174.             continue;

  175.         }

  176.         ptr = dst;

  177.         for(j = 0; j < 4; j++){

  178.             /* Coefficients are read using VLC chosen by the prediction pattern

  179.              * The first one (used for retrieving a pair of coefficients) is

  180.              * constructed from the top, top right and left coefficients

  181.              * The second one (used for retrieving only one coefficient) is

  182.              * top + 10 * left.

  183.              */

  184.             A = ptr[-s->b4_stride + 1]; // it won't be used for the last coefficient in a row

  185.             B = ptr[-s->b4_stride];

  186.             C = ptr[-1];

  187.             pattern = A + (B << 4) + (C << 8);

  188.             for(k = 0; k < MODE2_PATTERNS_NUM; k++)

  189.                 if(pattern == rv40_aic_table_index[k])

  190.                     break;

  191.             if(j < 3 && k < MODE2_PATTERNS_NUM){ //pattern is found, decoding 2 coefficients

  192.                 v = get_vlc2(gb, aic_mode2_vlc[k].table, AIC_MODE2_BITS, 2);

  193.                 *ptr++ = v/9;

  194.                 *ptr++ = v%9;

  195.                 j++;

  196.             }else{

  197.                 if(B != -1 && C != -1)

  198.                     v = get_vlc2(gb, aic_mode1_vlc[B + C*10].table, AIC_MODE1_BITS, 1);

  199.                 else{ // tricky decoding

  200.                     v = 0;

  201.                     switch(C){

  202.                     case -1: // code 0 -> 1, 1 -> 0

  203.                         if(B < 2)

  204.                             v = get_bits1(gb) ^ 1;

  205.                         break;

  206.                     case  0:

  207.                     case  2: // code 0 -> 2, 1 -> 0

  208.                         v = (get_bits1(gb) ^ 1) << 1;

  209.                         break;

  210.                     }

  211.                 }

  212.                 *ptr++ = v;

  213.             }

  214.         }

  215.     }

  216.     return 0;

  217. }

  218. 

  219. /**

  220.  * Decode macroblock information.

  221.  */

  222. static int rv40_decode_mb_info(RV34DecContext *r)

  223. {

  224.     MpegEncContext *s = &r->s;

  225.     GetBitContext *gb = &s->gb;

  226.     int q, i;

  227.     int prev_type = 0;

  228.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  229.     int blocks[RV34_MB_TYPES] = {0};

  230.     int count = 0;

  231. 

  232.     if(!r->s.mb_skip_run)

  233.         r->s.mb_skip_run = svq3_get_ue_golomb(gb) + 1;

  234. 

  235.     if(--r->s.mb_skip_run)

  236.          return RV34_MB_SKIP;

  237. 

  238.     if(r->avail_cache[5-1])

  239.         blocks[r->mb_type[mb_pos - 1]]++;

  240.     if(r->avail_cache[5-4]){

  241.         blocks[r->mb_type[mb_pos - s->mb_stride]]++;

  242.         if(r->avail_cache[5-2])

  243.             blocks[r->mb_type[mb_pos - s->mb_stride + 1]]++;

  244.         if(r->avail_cache[5-5])

  245.             blocks[r->mb_type[mb_pos - s->mb_stride - 1]]++;

  246.     }

  247. 

  248.     for(i = 0; i < RV34_MB_TYPES; i++){

  249.         if(blocks[i] > count){

  250.             count = blocks[i];

  251.             prev_type = i;

  252.         }

  253.     }

  254.     if(s->pict_type == FF_P_TYPE){

  255.         prev_type = block_num_to_ptype_vlc_num[prev_type];

  256.         q = get_vlc2(gb, ptype_vlc[prev_type].table, PTYPE_VLC_BITS, 1);

  257.         if(q < PBTYPE_ESCAPE)

  258.             return q;

  259.         q = get_vlc2(gb, ptype_vlc[prev_type].table, PTYPE_VLC_BITS, 1);

  260.         av_log(s->avctx, AV_LOG_ERROR, "Dquant for P-frame\n");

  261.     }else{

  262.         prev_type = block_num_to_btype_vlc_num[prev_type];

  263.         q = get_vlc2(gb, btype_vlc[prev_type].table, BTYPE_VLC_BITS, 1);

  264.         if(q < PBTYPE_ESCAPE)

  265.             return q;

  266.         q = get_vlc2(gb, btype_vlc[prev_type].table, BTYPE_VLC_BITS, 1);

  267.         av_log(s->avctx, AV_LOG_ERROR, "Dquant for B-frame\n");

  268.     }

  269.     return 0;

  270. }

  271. 

  272. #define CLIP_SYMM(a, b) av_clip(a, -(b), b)

  273. /**

  274.  * weaker deblocking very similar to the one described in 4.4.2 of JVT-A003r1

  275.  */

  276. static inline void rv40_weak_loop_filter(uint8_t *src, const int step,

  277.                                          const int filter_p1, const int filter_q1,

  278.                                          const int alpha, const int beta,

  279.                                          const int lim_p0q0,

  280.                                          const int lim_q1, const int lim_p1,

  281.                                          const int diff_p1p0, const int diff_q1q0,

  282.                                          const int diff_p1p2, const int diff_q1q2)

  283. {

  284.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  285.     int t, u, diff;

  286. 

  287.     t = src[0*step] - src[-1*step];

  288.     if(!t)

  289.         return;

  290.     u = (alpha * FFABS(t)) >> 7;

  291.     if(u > 3 - (filter_p1 && filter_q1))

  292.         return;

  293. 

  294.     t <<= 2;

  295.     if(filter_p1 && filter_q1)

  296.         t += src[-2*step] - src[1*step];

  297.     diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0);

  298.     src[-1*step] = cm[src[-1*step] + diff];

  299.     src[ 0*step] = cm[src[ 0*step] - diff];

  300.     if(FFABS(diff_p1p2) <= beta && filter_p1){

  301.         t = (diff_p1p0 + diff_p1p2 - diff) >> 1;

  302.         src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)];

  303.     }

  304.     if(FFABS(diff_q1q2) <= beta && filter_q1){

  305.         t = (diff_q1q0 + diff_q1q2 + diff) >> 1;

  306.         src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)];

  307.     }

  308. }

  309. 

  310. static inline void rv40_adaptive_loop_filter(uint8_t *src, const int step,

  311.                                              const int stride, const int dmode,

  312.                                              const int lim_q1, const int lim_p1,

  313.                                              const int alpha,

  314.                                              const int beta, const int beta2,

  315.                                              const int chroma, const int edge)

  316. {

  317.     int diff_p1p0[4], diff_q1q0[4], diff_p1p2[4], diff_q1q2[4];

  318.     int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0;

  319.     uint8_t *ptr;

  320.     int flag_strong0 = 1, flag_strong1 = 1;

  321.     int filter_p1, filter_q1;

  322.     int i;

  323.     int lims;

  324. 

  325.     for(i = 0, ptr = src; i < 4; i++, ptr += stride){

  326.         diff_p1p0[i] = ptr[-2*step] - ptr[-1*step];

  327.         diff_q1q0[i] = ptr[ 1*step] - ptr[ 0*step];

  328.         sum_p1p0 += diff_p1p0[i];

  329.         sum_q1q0 += diff_q1q0[i];

  330.     }

  331.     filter_p1 = FFABS(sum_p1p0) < (beta<<2);

  332.     filter_q1 = FFABS(sum_q1q0) < (beta<<2);

  333.     if(!filter_p1 && !filter_q1)

  334.         return;

  335. 

  336.     for(i = 0, ptr = src; i < 4; i++, ptr += stride){

  337.         diff_p1p2[i] = ptr[-2*step] - ptr[-3*step];

  338.         diff_q1q2[i] = ptr[ 1*step] - ptr[ 2*step];

  339.         sum_p1p2 += diff_p1p2[i];

  340.         sum_q1q2 += diff_q1q2[i];

  341.     }

  342. 

  343.     if(edge){

  344.         flag_strong0 = filter_p1 && (FFABS(sum_p1p2) < beta2);

  345.         flag_strong1 = filter_q1 && (FFABS(sum_q1q2) < beta2);

  346.     }else{

  347.         flag_strong0 = flag_strong1 = 0;

  348.     }

  349. 

  350.     lims = filter_p1 + filter_q1 + ((lim_q1 + lim_p1) >> 1) + 1;

  351.     if(flag_strong0 && flag_strong1){ /* strong filtering */

  352.         for(i = 0; i < 4; i++, src += stride){

  353.             int sflag, p0, q0, p1, q1;

  354.             int t = src[0*step] - src[-1*step];

  355. 

  356.             if(!t) continue;

  357.             sflag = (alpha * FFABS(t)) >> 7;

  358.             if(sflag > 1) continue;

  359. 

  360.             p0 = (25*src[-3*step] + 26*src[-2*step]

  361.                 + 26*src[-1*step]

  362.                 + 26*src[ 0*step] + 25*src[ 1*step] + rv40_dither_l[dmode + i]) >> 7;

  363.             q0 = (25*src[-2*step] + 26*src[-1*step]

  364.                 + 26*src[ 0*step]

  365.                 + 26*src[ 1*step] + 25*src[ 2*step] + rv40_dither_r[dmode + i]) >> 7;

  366.             if(sflag){

  367.                 p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims);

  368.                 q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims);

  369.             }

  370.             p1 = (25*src[-4*step] + 26*src[-3*step]

  371.                 + 26*src[-2*step]

  372.                 + 26*p0           + 25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;

  373.             q1 = (25*src[-1*step] + 26*q0

  374.                 + 26*src[ 1*step]

  375.                 + 26*src[ 2*step] + 25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;

  376.             if(sflag){

  377.                 p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims);

  378.                 q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims);

  379.             }

  380.             src[-2*step] = p1;

  381.             src[-1*step] = p0;

  382.             src[ 0*step] = q0;

  383.             src[ 1*step] = q1;

  384.             if(!chroma){

  385.                 src[-3*step] = (25*src[-1*step] + 26*src[-2*step] + 51*src[-3*step] + 26*src[-4*step] + 64) >> 7;

  386.                 src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] + 51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;

  387.             }

  388.         }

  389.     }else if(filter_p1 && filter_q1){

  390.         for(i = 0; i < 4; i++, src += stride)

  391.             rv40_weak_loop_filter(src, step, 1, 1, alpha, beta, lims, lim_q1, lim_p1,

  392.                                   diff_p1p0[i], diff_q1q0[i], diff_p1p2[i], diff_q1q2[i]);

  393.     }else{

  394.         for(i = 0; i < 4; i++, src += stride)

  395.             rv40_weak_loop_filter(src, step, filter_p1, filter_q1,

  396.                                   alpha, beta, lims>>1, lim_q1>>1, lim_p1>>1,

  397.                                   diff_p1p0[i], diff_q1q0[i], diff_p1p2[i], diff_q1q2[i]);

  398.     }

  399. }

  400. 

  401. static void rv40_v_loop_filter(uint8_t *src, int stride, int dmode,

  402.                                int lim_q1, int lim_p1,

  403.                                int alpha, int beta, int beta2, int chroma, int edge){

  404.     rv40_adaptive_loop_filter(src, 1, stride, dmode, lim_q1, lim_p1,

  405.                               alpha, beta, beta2, chroma, edge);

  406. }

  407. static void rv40_h_loop_filter(uint8_t *src, int stride, int dmode,

  408.                                int lim_q1, int lim_p1,

  409.                                int alpha, int beta, int beta2, int chroma, int edge){

  410.     rv40_adaptive_loop_filter(src, stride, 1, dmode, lim_q1, lim_p1,

  411.                               alpha, beta, beta2, chroma, edge);

  412. }

  413. 

  414. enum RV40BlockPos{

  415.     POS_CUR,

  416.     POS_TOP,

  417.     POS_LEFT,

  418.     POS_BOTTOM,

  419. };

  420. 

  421. #define MASK_CUR          0x0001

  422. #define MASK_RIGHT        0x0008

  423. #define MASK_BOTTOM       0x0010

  424. #define MASK_TOP          0x1000

  425. #define MASK_Y_TOP_ROW    0x000F

  426. #define MASK_Y_LAST_ROW   0xF000

  427. #define MASK_Y_LEFT_COL   0x1111

  428. #define MASK_Y_RIGHT_COL  0x8888

  429. #define MASK_C_TOP_ROW    0x0003

  430. #define MASK_C_LAST_ROW   0x000C

  431. #define MASK_C_LEFT_COL   0x0005

  432. #define MASK_C_RIGHT_COL  0x000A

  433. 

  434. static const int neighbour_offs_x[4] = { 0,  0, -1, 0 };

  435. static const int neighbour_offs_y[4] = { 0, -1,  0, 1 };

  436. 

  437. /**

  438.  * RV40 loop filtering function

  439.  */

  440. static void rv40_loop_filter(RV34DecContext *r, int row)

  441. {

  442.     MpegEncContext *s = &r->s;

  443.     int mb_pos, mb_x;

  444.     int i, j, k;

  445.     uint8_t *Y, *C;

  446.     int alpha, beta, betaY, betaC;

  447.     int q;

  448.     int mbtype[4];   ///< current macroblock and its neighbours types

  449.     /**

  450.      * flags indicating that macroblock can be filtered with strong filter

  451.      * it is set only for intra coded MB and MB with DCs coded separately

  452.      */

  453.     int mb_strong[4];

  454.     int clip[4];     ///< MB filter clipping value calculated from filtering strength

  455.     /**

  456.      * coded block patterns for luma part of current macroblock and its neighbours

  457.      * Format:

  458.      * LSB corresponds to the top left block,

  459.      * each nibble represents one row of subblocks.

  460.      */

  461.     int cbp[4];

  462.     /**

  463.      * coded block patterns for chroma part of current macroblock and its neighbours

  464.      * Format is the same as for luma with two subblocks in a row.

  465.      */

  466.     int uvcbp[4][2];

  467.     /**

  468.      * This mask represents the pattern of luma subblocks that should be filtered

  469.      * in addition to the coded ones because because they lie at the edge of

  470.      * 8x8 block with different enough motion vectors

  471.      */

  472.     int mvmasks[4];

  473. 

  474.     mb_pos = row * s->mb_stride;

  475.     for(mb_x = 0; mb_x < s->mb_width; mb_x++, mb_pos++){

  476.         int mbtype = s->current_picture_ptr->mb_type[mb_pos];

  477.         if(IS_INTRA(mbtype) || IS_SEPARATE_DC(mbtype))

  478.             r->cbp_luma  [mb_pos] = r->deblock_coefs[mb_pos] = 0xFFFF;

  479.         if(IS_INTRA(mbtype))

  480.             r->cbp_chroma[mb_pos] = 0xFF;

  481.     }

  482.     mb_pos = row * s->mb_stride;

  483.     for(mb_x = 0; mb_x < s->mb_width; mb_x++, mb_pos++){

  484.         int y_h_deblock, y_v_deblock;

  485.         int c_v_deblock[2], c_h_deblock[2];

  486.         int clip_left;

  487.         int avail[4];

  488.         int y_to_deblock, c_to_deblock[2];

  489. 

  490.         q = s->current_picture_ptr->qscale_table[mb_pos];

  491.         alpha = rv40_alpha_tab[q];

  492.         beta  = rv40_beta_tab [q];

  493.         betaY = betaC = beta * 3;

  494.         if(s->width * s->height <= 176*144)

  495.             betaY += beta;

  496. 

  497.         avail[0] = 1;

  498.         avail[1] = row;

  499.         avail[2] = mb_x;

  500.         avail[3] = row < s->mb_height - 1;

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

  502.             if(avail[i]){

  503.                 int pos = mb_pos + neighbour_offs_x[i] + neighbour_offs_y[i]*s->mb_stride;

  504.                 mvmasks[i] = r->deblock_coefs[pos];

  505.                 mbtype [i] = s->current_picture_ptr->mb_type[pos];

  506.                 cbp    [i] = r->cbp_luma[pos];

  507.                 uvcbp[i][0] = r->cbp_chroma[pos] & 0xF;

  508.                 uvcbp[i][1] = r->cbp_chroma[pos] >> 4;

  509.             }else{

  510.                 mvmasks[i] = 0;

  511.                 mbtype [i] = mbtype[0];

  512.                 cbp    [i] = 0;

  513.                 uvcbp[i][0] = uvcbp[i][1] = 0;

  514.             }

  515.             mb_strong[i] = IS_INTRA(mbtype[i]) || IS_SEPARATE_DC(mbtype[i]);

  516.             clip[i] = rv40_filter_clip_tbl[mb_strong[i] + 1][q];

  517.         }

  518.         y_to_deblock =  mvmasks[POS_CUR]

  519.                      | (mvmasks[POS_BOTTOM] << 16);

  520.         /* This pattern contains bits signalling that horizontal edges of

  521.          * the current block can be filtered.

  522.          * That happens when either of adjacent subblocks is coded or lies on

  523.          * the edge of 8x8 blocks with motion vectors differing by more than

  524.          * 3/4 pel in any component (any edge orientation for some reason).

  525.          */

  526.         y_h_deblock =   y_to_deblock

  527.                     | ((cbp[POS_CUR]                           <<  4) & ~MASK_Y_TOP_ROW)

  528.                     | ((cbp[POS_TOP]        & MASK_Y_LAST_ROW) >> 12);

  529.         /* This pattern contains bits signalling that vertical edges of

  530.          * the current block can be filtered.

  531.          * That happens when either of adjacent subblocks is coded or lies on

  532.          * the edge of 8x8 blocks with motion vectors differing by more than

  533.          * 3/4 pel in any component (any edge orientation for some reason).

  534.          */

  535.         y_v_deblock =   y_to_deblock

  536.                     | ((cbp[POS_CUR]                      << 1) & ~MASK_Y_LEFT_COL)

  537.                     | ((cbp[POS_LEFT] & MASK_Y_RIGHT_COL) >> 3);

  538.         if(!mb_x)

  539.             y_v_deblock &= ~MASK_Y_LEFT_COL;

  540.         if(!row)

  541.             y_h_deblock &= ~MASK_Y_TOP_ROW;

  542.         if(row == s->mb_height - 1 || (mb_strong[POS_CUR] || mb_strong[POS_BOTTOM]))

  543.             y_h_deblock &= ~(MASK_Y_TOP_ROW << 16);

  544.         /* Calculating chroma patterns is similar and easier since there is

  545.          * no motion vector pattern for them.

  546.          */

  547.         for(i = 0; i < 2; i++){

  548.             c_to_deblock[i] = (uvcbp[POS_BOTTOM][i] << 4) | uvcbp[POS_CUR][i];

  549.             c_v_deblock[i] =   c_to_deblock[i]

  550.                            | ((uvcbp[POS_CUR] [i]                       << 1) & ~MASK_C_LEFT_COL)

  551.                            | ((uvcbp[POS_LEFT][i]   & MASK_C_RIGHT_COL) >> 1);

  552.             c_h_deblock[i] =   c_to_deblock[i]

  553.                            | ((uvcbp[POS_TOP][i]    & MASK_C_LAST_ROW)  >> 2)

  554.                            |  (uvcbp[POS_CUR][i]                        << 2);

  555.             if(!mb_x)

  556.                 c_v_deblock[i] &= ~MASK_C_LEFT_COL;

  557.             if(!row)

  558.                 c_h_deblock[i] &= ~MASK_C_TOP_ROW;

  559.             if(row == s->mb_height - 1 || mb_strong[POS_CUR] || mb_strong[POS_BOTTOM])

  560.                 c_h_deblock[i] &= ~(MASK_C_TOP_ROW << 4);

  561.         }

  562. 

  563.         for(j = 0; j < 16; j += 4){

  564.             Y = s->current_picture_ptr->data[0] + mb_x*16 + (row*16 + j) * s->linesize;

  565.             for(i = 0; i < 4; i++, Y += 4){

  566.                 int ij = i + j;

  567.                 int clip_cur = y_to_deblock & (MASK_CUR << ij) ? clip[POS_CUR] : 0;

  568.                 int dither = j ? ij : i*4;

  569. 

  570.                 // if bottom block is coded then we can filter its top edge

  571.                 // (or bottom edge of this block, which is the same)

  572.                 if(y_h_deblock & (MASK_BOTTOM << ij)){

  573.                     rv40_h_loop_filter(Y+4*s->linesize, s->linesize, dither,

  574.                                        y_to_deblock & (MASK_BOTTOM << ij) ? clip[POS_CUR] : 0,

  575.                                        clip_cur,

  576.                                        alpha, beta, betaY, 0, 0);

  577.                 }

  578.                 // filter left block edge in ordinary mode (with low filtering strength)

  579.                 if(y_v_deblock & (MASK_CUR << ij) && (i || !(mb_strong[POS_CUR] || mb_strong[POS_LEFT]))){

  580.                     if(!i)

  581.                         clip_left = mvmasks[POS_LEFT] & (MASK_RIGHT << j) ? clip[POS_LEFT] : 0;

  582.                     else

  583.                         clip_left = y_to_deblock & (MASK_CUR << (ij-1)) ? clip[POS_CUR] : 0;

  584.                     rv40_v_loop_filter(Y, s->linesize, dither,

  585.                                        clip_cur,

  586.                                        clip_left,

  587.                                        alpha, beta, betaY, 0, 0);

  588.                 }

  589.                 // filter top edge of the current macroblock when filtering strength is high

  590.                 if(!j && y_h_deblock & (MASK_CUR << i) && (mb_strong[POS_CUR] || mb_strong[POS_TOP])){

  591.                     rv40_h_loop_filter(Y, s->linesize, dither,

  592.                                        clip_cur,

  593.                                        mvmasks[POS_TOP] & (MASK_TOP << i) ? clip[POS_TOP] : 0,

  594.                                        alpha, beta, betaY, 0, 1);

  595.                 }

  596.                 // filter left block edge in edge mode (with high filtering strength)

  597.                 if(y_v_deblock & (MASK_CUR << ij) && !i && (mb_strong[POS_CUR] || mb_strong[POS_LEFT])){

  598.                     clip_left = mvmasks[POS_LEFT] & (MASK_RIGHT << j) ? clip[POS_LEFT] : 0;

  599.                     rv40_v_loop_filter(Y, s->linesize, dither,

  600.                                        clip_cur,

  601.                                        clip_left,

  602.                                        alpha, beta, betaY, 0, 1);

  603.                 }

  604.             }

  605.         }

  606.         for(k = 0; k < 2; k++){

  607.             for(j = 0; j < 2; j++){

  608.                 C = s->current_picture_ptr->data[k+1] + mb_x*8 + (row*8 + j*4) * s->uvlinesize;

  609.                 for(i = 0; i < 2; i++, C += 4){

  610.                     int ij = i + j*2;

  611.                     int clip_cur = c_to_deblock[k] & (MASK_CUR << ij) ? clip[POS_CUR] : 0;

  612.                     if(c_h_deblock[k] & (MASK_CUR << (ij+2))){

  613.                         int clip_bot = c_to_deblock[k] & (MASK_CUR << (ij+2)) ? clip[POS_CUR] : 0;

  614.                         rv40_h_loop_filter(C+4*s->uvlinesize, s->uvlinesize, i*8,

  615.                                            clip_bot,

  616.                                            clip_cur,

  617.                                            alpha, beta, betaC, 1, 0);

  618.                     }

  619.                     if((c_v_deblock[k] & (MASK_CUR << ij)) && (i || !(mb_strong[POS_CUR] || mb_strong[POS_LEFT]))){

  620.                         if(!i)

  621.                             clip_left = uvcbp[POS_LEFT][k] & (MASK_CUR << (2*j+1)) ? clip[POS_LEFT] : 0;

  622.                         else

  623.                             clip_left = c_to_deblock[k]    & (MASK_CUR << (ij-1))  ? clip[POS_CUR]  : 0;

  624.                         rv40_v_loop_filter(C, s->uvlinesize, j*8,

  625.                                            clip_cur,

  626.                                            clip_left,

  627.                                            alpha, beta, betaC, 1, 0);

  628.                     }

  629.                     if(!j && c_h_deblock[k] & (MASK_CUR << ij) && (mb_strong[POS_CUR] || mb_strong[POS_TOP])){

  630.                         int clip_top = uvcbp[POS_TOP][k] & (MASK_CUR << (ij+2)) ? clip[POS_TOP] : 0;

  631.                         rv40_h_loop_filter(C, s->uvlinesize, i*8,

  632.                                            clip_cur,

  633.                                            clip_top,

  634.                                            alpha, beta, betaC, 1, 1);

  635.                     }

  636.                     if(c_v_deblock[k] & (MASK_CUR << ij) && !i && (mb_strong[POS_CUR] || mb_strong[POS_LEFT])){

  637.                         clip_left = uvcbp[POS_LEFT][k] & (MASK_CUR << (2*j+1)) ? clip[POS_LEFT] : 0;

  638.                         rv40_v_loop_filter(C, s->uvlinesize, j*8,

  639.                                            clip_cur,

  640.                                            clip_left,

  641.                                            alpha, beta, betaC, 1, 1);

  642.                     }

  643.                 }

  644.             }

  645.         }

  646.     }

  647. }

  648. 

  649. /**

  650.  * Initialize decoder.

  651.  */

  652. static av_cold int rv40_decode_init(AVCodecContext *avctx)

  653. {

  654.     RV34DecContext *r = avctx->priv_data;

  655. 

  656.     r->rv30 = 0;

  657.     ff_rv34_decode_init(avctx);

  658.     if(!aic_top_vlc.bits)

  659.         rv40_init_tables();

  660.     r->parse_slice_header = rv40_parse_slice_header;

  661.     r->decode_intra_types = rv40_decode_intra_types;

  662.     r->decode_mb_info     = rv40_decode_mb_info;

  663.     r->loop_filter        = rv40_loop_filter;

  664.     r->luma_dc_quant_i = rv40_luma_dc_quant[0];

  665.     r->luma_dc_quant_p = rv40_luma_dc_quant[1];

  666.     return 0;

  667. }

  668. 

  669. AVCodec rv40_decoder = {

  670.     "rv40",

  671.     CODEC_TYPE_VIDEO,

  672.     CODEC_ID_RV40,

  673.     sizeof(RV34DecContext),

  674.     rv40_decode_init,

  675.     NULL,

  676.     ff_rv34_decode_end,

  677.     ff_rv34_decode_frame,

  678.     CODEC_CAP_DR1 | CODEC_CAP_DELAY,

  679.     .flush = ff_mpeg_flush,

  680.     .long_name = NULL_IF_CONFIG_SMALL("RealVideo 4.0"),

  681.     .pix_fmts= ff_pixfmt_list_420,

  682. };