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.
15543 Commits
32 Branches
759MB
Tree: d63f529df7
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'd63f529df7'
${ noResults }
FFmpeg/libavcodec/vc1.c

4318 lines
151KB
Raw Blame History 

  1. /*

  2.  * VC-1 and WMV3 decoder

  3.  * Copyright (c) 2006-2007 Konstantin Shishkov

  4.  * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer

  5.  *

  6.  * This file is part of FFmpeg.

  7.  *

  8.  * FFmpeg is free software; you can redistribute it and/or

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

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

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

  12.  *

  13.  * FFmpeg is distributed in the hope that it will be useful,

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

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

  16.  * Lesser General Public License for more details.

  17.  *

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

  19.  * License along with FFmpeg; if not, write to the Free Software

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

  21.  */

  22. 

  23. /**

  24.  * @file vc1.c

  25.  * VC-1 and WMV3 decoder

  26.  *

  27.  */

  28. #include "dsputil.h"

  29. #include "avcodec.h"

  30. #include "mpegvideo.h"

  31. #include "vc1.h"

  32. #include "vc1data.h"

  33. #include "vc1acdata.h"

  34. #include "msmpeg4data.h"

  35. #include "unary.h"

  36. #include "simple_idct.h"

  37. 

  38. #undef NDEBUG

  39. #include <assert.h>

  40. 

  41. #define MB_INTRA_VLC_BITS 9

  42. #define DC_VLC_BITS 9

  43. #define AC_VLC_BITS 9

  44. static const uint16_t table_mb_intra[64][2];

  45. 

  46. 

  47. /**

  48.  * Init VC-1 specific tables and VC1Context members

  49.  * @param v The VC1Context to initialize

  50.  * @return Status

  51.  */

  52. static int vc1_init_common(VC1Context *v)

  53. {

  54.     static int done = 0;

  55.     int i = 0;

  56. 

  57.     v->hrd_rate = v->hrd_buffer = NULL;

  58. 

  59.     /* VLC tables */

  60.     if(!done)

  61.     {

  62.         done = 1;

  63.         init_vlc(&ff_vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,

  64.                  ff_vc1_bfraction_bits, 1, 1,

  65.                  ff_vc1_bfraction_codes, 1, 1, 1);

  66.         init_vlc(&ff_vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,

  67.                  ff_vc1_norm2_bits, 1, 1,

  68.                  ff_vc1_norm2_codes, 1, 1, 1);

  69.         init_vlc(&ff_vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,

  70.                  ff_vc1_norm6_bits, 1, 1,

  71.                  ff_vc1_norm6_codes, 2, 2, 1);

  72.         init_vlc(&ff_vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,

  73.                  ff_vc1_imode_bits, 1, 1,

  74.                  ff_vc1_imode_codes, 1, 1, 1);

  75.         for (i=0; i<3; i++)

  76.         {

  77.             init_vlc(&ff_vc1_ttmb_vlc[i], VC1_TTMB_VLC_BITS, 16,

  78.                      ff_vc1_ttmb_bits[i], 1, 1,

  79.                      ff_vc1_ttmb_codes[i], 2, 2, 1);

  80.             init_vlc(&ff_vc1_ttblk_vlc[i], VC1_TTBLK_VLC_BITS, 8,

  81.                      ff_vc1_ttblk_bits[i], 1, 1,

  82.                      ff_vc1_ttblk_codes[i], 1, 1, 1);

  83.             init_vlc(&ff_vc1_subblkpat_vlc[i], VC1_SUBBLKPAT_VLC_BITS, 15,

  84.                      ff_vc1_subblkpat_bits[i], 1, 1,

  85.                      ff_vc1_subblkpat_codes[i], 1, 1, 1);

  86.         }

  87.         for(i=0; i<4; i++)

  88.         {

  89.             init_vlc(&ff_vc1_4mv_block_pattern_vlc[i], VC1_4MV_BLOCK_PATTERN_VLC_BITS, 16,

  90.                      ff_vc1_4mv_block_pattern_bits[i], 1, 1,

  91.                      ff_vc1_4mv_block_pattern_codes[i], 1, 1, 1);

  92.             init_vlc(&ff_vc1_cbpcy_p_vlc[i], VC1_CBPCY_P_VLC_BITS, 64,

  93.                      ff_vc1_cbpcy_p_bits[i], 1, 1,

  94.                      ff_vc1_cbpcy_p_codes[i], 2, 2, 1);

  95.             init_vlc(&ff_vc1_mv_diff_vlc[i], VC1_MV_DIFF_VLC_BITS, 73,

  96.                      ff_vc1_mv_diff_bits[i], 1, 1,

  97.                      ff_vc1_mv_diff_codes[i], 2, 2, 1);

  98.         }

  99.         for(i=0; i<8; i++)

  100.             init_vlc(&ff_vc1_ac_coeff_table[i], AC_VLC_BITS, vc1_ac_sizes[i],

  101.                      &vc1_ac_tables[i][0][1], 8, 4,

  102.                      &vc1_ac_tables[i][0][0], 8, 4, 1);

  103.         init_vlc(&ff_msmp4_mb_i_vlc, MB_INTRA_VLC_BITS, 64,

  104.                  &ff_msmp4_mb_i_table[0][1], 4, 2,

  105.                  &ff_msmp4_mb_i_table[0][0], 4, 2, 1);

  106.     }

  107. 

  108.     /* Other defaults */

  109.     v->pq = -1;

  110.     v->mvrange = 0; /* 7.1.1.18, p80 */

  111. 

  112.     return 0;

  113. }

  114. 

  115. /***********************************************************************/

  116. /**

  117.  * @defgroup bitplane VC9 Bitplane decoding

  118.  * @see 8.7, p56

  119.  * @{

  120.  */

  121. 

  122. /** @addtogroup bitplane

  123.  * Imode types

  124.  * @{

  125.  */

  126. enum Imode {

  127.     IMODE_RAW,

  128.     IMODE_NORM2,

  129.     IMODE_DIFF2,

  130.     IMODE_NORM6,

  131.     IMODE_DIFF6,

  132.     IMODE_ROWSKIP,

  133.     IMODE_COLSKIP

  134. };

  135. /** @} */ //imode defines

  136. 

  137. /** Decode rows by checking if they are skipped

  138.  * @param plane Buffer to store decoded bits

  139.  * @param[in] width Width of this buffer

  140.  * @param[in] height Height of this buffer

  141.  * @param[in] stride of this buffer

  142.  */

  143. static void decode_rowskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){

  144.     int x, y;

  145. 

  146.     for (y=0; y<height; y++){

  147.         if (!get_bits1(gb)) //rowskip

  148.             memset(plane, 0, width);

  149.         else

  150.             for (x=0; x<width; x++)

  151.                 plane[x] = get_bits1(gb);

  152.         plane += stride;

  153.     }

  154. }

  155. 

  156. /** Decode columns by checking if they are skipped

  157.  * @param plane Buffer to store decoded bits

  158.  * @param[in] width Width of this buffer

  159.  * @param[in] height Height of this buffer

  160.  * @param[in] stride of this buffer

  161.  * @todo FIXME: Optimize

  162.  */

  163. static void decode_colskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){

  164.     int x, y;

  165. 

  166.     for (x=0; x<width; x++){

  167.         if (!get_bits1(gb)) //colskip

  168.             for (y=0; y<height; y++)

  169.                 plane[y*stride] = 0;

  170.         else

  171.             for (y=0; y<height; y++)

  172.                 plane[y*stride] = get_bits1(gb);

  173.         plane ++;

  174.     }

  175. }

  176. 

  177. /** Decode a bitplane's bits

  178.  * @param bp Bitplane where to store the decode bits

  179.  * @param v VC-1 context for bit reading and logging

  180.  * @return Status

  181.  * @todo FIXME: Optimize

  182.  */

  183. static int bitplane_decoding(uint8_t* data, int *raw_flag, VC1Context *v)

  184. {

  185.     GetBitContext *gb = &v->s.gb;

  186. 

  187.     int imode, x, y, code, offset;

  188.     uint8_t invert, *planep = data;

  189.     int width, height, stride;

  190. 

  191.     width = v->s.mb_width;

  192.     height = v->s.mb_height;

  193.     stride = v->s.mb_stride;

  194.     invert = get_bits1(gb);

  195.     imode = get_vlc2(gb, ff_vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);

  196. 

  197.     *raw_flag = 0;

  198.     switch (imode)

  199.     {

  200.     case IMODE_RAW:

  201.         //Data is actually read in the MB layer (same for all tests == "raw")

  202.         *raw_flag = 1; //invert ignored

  203.         return invert;

  204.     case IMODE_DIFF2:

  205.     case IMODE_NORM2:

  206.         if ((height * width) & 1)

  207.         {

  208.             *planep++ = get_bits1(gb);

  209.             offset = 1;

  210.         }

  211.         else offset = 0;

  212.         // decode bitplane as one long line

  213.         for (y = offset; y < height * width; y += 2) {

  214.             code = get_vlc2(gb, ff_vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);

  215.             *planep++ = code & 1;

  216.             offset++;

  217.             if(offset == width) {

  218.                 offset = 0;

  219.                 planep += stride - width;

  220.             }

  221.             *planep++ = code >> 1;

  222.             offset++;

  223.             if(offset == width) {

  224.                 offset = 0;

  225.                 planep += stride - width;

  226.             }

  227.         }

  228.         break;

  229.     case IMODE_DIFF6:

  230.     case IMODE_NORM6:

  231.         if(!(height % 3) && (width % 3)) { // use 2x3 decoding

  232.             for(y = 0; y < height; y+= 3) {

  233.                 for(x = width & 1; x < width; x += 2) {

  234.                     code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);

  235.                     if(code < 0){

  236.                         av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");

  237.                         return -1;

  238.                     }

  239.                     planep[x + 0] = (code >> 0) & 1;

  240.                     planep[x + 1] = (code >> 1) & 1;

  241.                     planep[x + 0 + stride] = (code >> 2) & 1;

  242.                     planep[x + 1 + stride] = (code >> 3) & 1;

  243.                     planep[x + 0 + stride * 2] = (code >> 4) & 1;

  244.                     planep[x + 1 + stride * 2] = (code >> 5) & 1;

  245.                 }

  246.                 planep += stride * 3;

  247.             }

  248.             if(width & 1) decode_colskip(data, 1, height, stride, &v->s.gb);

  249.         } else { // 3x2

  250.             planep += (height & 1) * stride;

  251.             for(y = height & 1; y < height; y += 2) {

  252.                 for(x = width % 3; x < width; x += 3) {

  253.                     code = get_vlc2(gb, ff_vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);

  254.                     if(code < 0){

  255.                         av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");

  256.                         return -1;

  257.                     }

  258.                     planep[x + 0] = (code >> 0) & 1;

  259.                     planep[x + 1] = (code >> 1) & 1;

  260.                     planep[x + 2] = (code >> 2) & 1;

  261.                     planep[x + 0 + stride] = (code >> 3) & 1;

  262.                     planep[x + 1 + stride] = (code >> 4) & 1;

  263.                     planep[x + 2 + stride] = (code >> 5) & 1;

  264.                 }

  265.                 planep += stride * 2;

  266.             }

  267.             x = width % 3;

  268.             if(x) decode_colskip(data  ,             x, height    , stride, &v->s.gb);

  269.             if(height & 1) decode_rowskip(data+x, width - x, 1, stride, &v->s.gb);

  270.         }

  271.         break;

  272.     case IMODE_ROWSKIP:

  273.         decode_rowskip(data, width, height, stride, &v->s.gb);

  274.         break;

  275.     case IMODE_COLSKIP:

  276.         decode_colskip(data, width, height, stride, &v->s.gb);

  277.         break;

  278.     default: break;

  279.     }

  280. 

  281.     /* Applying diff operator */

  282.     if (imode == IMODE_DIFF2 || imode == IMODE_DIFF6)

  283.     {

  284.         planep = data;

  285.         planep[0] ^= invert;

  286.         for (x=1; x<width; x++)

  287.             planep[x] ^= planep[x-1];

  288.         for (y=1; y<height; y++)

  289.         {

  290.             planep += stride;

  291.             planep[0] ^= planep[-stride];

  292.             for (x=1; x<width; x++)

  293.             {

  294.                 if (planep[x-1] != planep[x-stride]) planep[x] ^= invert;

  295.                 else                                 planep[x] ^= planep[x-1];

  296.             }

  297.         }

  298.     }

  299.     else if (invert)

  300.     {

  301.         planep = data;

  302.         for (x=0; x<stride*height; x++) planep[x] = !planep[x]; //FIXME stride

  303.     }

  304.     return (imode<<1) + invert;

  305. }

  306. 

  307. /** @} */ //Bitplane group

  308. 

  309. #define FILTSIGN(a) ((a) >= 0 ? 1 : -1)

  310. /**

  311.  * VC-1 in-loop deblocking filter for one line

  312.  * @param src source block type

  313.  * @param pq block quantizer

  314.  * @return whether other 3 pairs should be filtered or not

  315.  * @see 8.6

  316.  */

  317. static int vc1_filter_line(uint8_t* src, int stride, int pq){

  318.     int a0, a1, a2, a3, d, clip, filt3 = 0;

  319.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  320. 

  321.     a0     = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;

  322.     if(FFABS(a0) < pq){

  323.         a1 = (2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3;

  324.         a2 = (2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3;

  325.         a3 = FFMIN(FFABS(a1), FFABS(a2));

  326.         if(a3 < FFABS(a0)){

  327.             d = 5 * ((a0 >=0 ? a3 : -a3) - a0) / 8;

  328.             clip = (src[-1*stride] - src[ 0*stride])/2;

  329.             if(clip){

  330.                 filt3 = 1;

  331.                 if(clip > 0)

  332.                     d = av_clip(d, 0, clip);

  333.                 else

  334.                     d = av_clip(d, clip, 0);

  335.                 src[-1*stride] = cm[src[-1*stride] - d];

  336.                 src[ 0*stride] = cm[src[ 0*stride] + d];

  337.             }

  338.         }

  339.     }

  340.     return filt3;

  341. }

  342. 

  343. /**

  344.  * VC-1 in-loop deblocking filter

  345.  * @param src source block type

  346.  * @param len edge length to filter (4 or 8 pixels)

  347.  * @param pq block quantizer

  348.  * @see 8.6

  349.  */

  350. static void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)

  351. {

  352.     int i;

  353.     int filt3;

  354. 

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

  356.         filt3 = vc1_filter_line(src + 2*step, stride, pq);

  357.         if(filt3){

  358.             vc1_filter_line(src + 0*step, stride, pq);

  359.             vc1_filter_line(src + 1*step, stride, pq);

  360.             vc1_filter_line(src + 3*step, stride, pq);

  361.         }

  362.         src += step * 4;

  363.     }

  364. }

  365. 

  366. static void vc1_loop_filter_iblk(MpegEncContext *s, int pq)

  367. {

  368.     int i, j;

  369.     if(!s->first_slice_line)

  370.         vc1_loop_filter(s->dest[0], 1, s->linesize, 16, pq);

  371.     vc1_loop_filter(s->dest[0] + 8*s->linesize, 1, s->linesize, 16, pq);

  372.     for(i = !s->mb_x*8; i < 16; i += 8)

  373.         vc1_loop_filter(s->dest[0] + i, s->linesize, 1, 16, pq);

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

  375.         if(!s->first_slice_line)

  376.             vc1_loop_filter(s->dest[j+1], 1, s->uvlinesize, 8, pq);

  377.         if(s->mb_x)

  378.             vc1_loop_filter(s->dest[j+1], s->uvlinesize, 1, 8, pq);

  379.     }

  380. }

  381. 

  382. /***********************************************************************/

  383. /** VOP Dquant decoding

  384.  * @param v VC-1 Context

  385.  */

  386. static int vop_dquant_decoding(VC1Context *v)

  387. {

  388.     GetBitContext *gb = &v->s.gb;

  389.     int pqdiff;

  390. 

  391.     //variable size

  392.     if (v->dquant == 2)

  393.     {

  394.         pqdiff = get_bits(gb, 3);

  395.         if (pqdiff == 7) v->altpq = get_bits(gb, 5);

  396.         else v->altpq = v->pq + pqdiff + 1;

  397.     }

  398.     else

  399.     {

  400.         v->dquantfrm = get_bits1(gb);

  401.         if ( v->dquantfrm )

  402.         {

  403.             v->dqprofile = get_bits(gb, 2);

  404.             switch (v->dqprofile)

  405.             {

  406.             case DQPROFILE_SINGLE_EDGE:

  407.             case DQPROFILE_DOUBLE_EDGES:

  408.                 v->dqsbedge = get_bits(gb, 2);

  409.                 break;

  410.             case DQPROFILE_ALL_MBS:

  411.                 v->dqbilevel = get_bits1(gb);

  412.                 if(!v->dqbilevel)

  413.                     v->halfpq = 0;

  414.             default: break; //Forbidden ?

  415.             }

  416.             if (v->dqbilevel || v->dqprofile != DQPROFILE_ALL_MBS)

  417.             {

  418.                 pqdiff = get_bits(gb, 3);

  419.                 if (pqdiff == 7) v->altpq = get_bits(gb, 5);

  420.                 else v->altpq = v->pq + pqdiff + 1;

  421.             }

  422.         }

  423.     }

  424.     return 0;

  425. }

  426. 

  427. /** Put block onto picture

  428.  */

  429. static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])

  430. {

  431.     uint8_t *Y;

  432.     int ys, us, vs;

  433.     DSPContext *dsp = &v->s.dsp;

  434. 

  435.     if(v->rangeredfrm) {

  436.         int i, j, k;

  437.         for(k = 0; k < 6; k++)

  438.             for(j = 0; j < 8; j++)

  439.                 for(i = 0; i < 8; i++)

  440.                     block[k][i + j*8] = ((block[k][i + j*8] - 128) << 1) + 128;

  441. 

  442.     }

  443.     ys = v->s.current_picture.linesize[0];

  444.     us = v->s.current_picture.linesize[1];

  445.     vs = v->s.current_picture.linesize[2];

  446.     Y = v->s.dest[0];

  447. 

  448.     dsp->put_pixels_clamped(block[0], Y, ys);

  449.     dsp->put_pixels_clamped(block[1], Y + 8, ys);

  450.     Y += ys * 8;

  451.     dsp->put_pixels_clamped(block[2], Y, ys);

  452.     dsp->put_pixels_clamped(block[3], Y + 8, ys);

  453. 

  454.     if(!(v->s.flags & CODEC_FLAG_GRAY)) {

  455.         dsp->put_pixels_clamped(block[4], v->s.dest[1], us);

  456.         dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);

  457.     }

  458. }

  459. 

  460. /** Do motion compensation over 1 macroblock

  461.  * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c

  462.  */

  463. static void vc1_mc_1mv(VC1Context *v, int dir)

  464. {

  465.     MpegEncContext *s = &v->s;

  466.     DSPContext *dsp = &v->s.dsp;

  467.     uint8_t *srcY, *srcU, *srcV;

  468.     int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;

  469. 

  470.     if(!v->s.last_picture.data[0])return;

  471. 

  472.     mx = s->mv[dir][0][0];

  473.     my = s->mv[dir][0][1];

  474. 

  475.     // store motion vectors for further use in B frames

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

  477.         s->current_picture.motion_val[1][s->block_index[0]][0] = mx;

  478.         s->current_picture.motion_val[1][s->block_index[0]][1] = my;

  479.     }

  480.     uvmx = (mx + ((mx & 3) == 3)) >> 1;

  481.     uvmy = (my + ((my & 3) == 3)) >> 1;

  482.     if(v->fastuvmc) {

  483.         uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));

  484.         uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));

  485.     }

  486.     if(!dir) {

  487.         srcY = s->last_picture.data[0];

  488.         srcU = s->last_picture.data[1];

  489.         srcV = s->last_picture.data[2];

  490.     } else {

  491.         srcY = s->next_picture.data[0];

  492.         srcU = s->next_picture.data[1];

  493.         srcV = s->next_picture.data[2];

  494.     }

  495. 

  496.     src_x = s->mb_x * 16 + (mx >> 2);

  497.     src_y = s->mb_y * 16 + (my >> 2);

  498.     uvsrc_x = s->mb_x * 8 + (uvmx >> 2);

  499.     uvsrc_y = s->mb_y * 8 + (uvmy >> 2);

  500. 

  501.     if(v->profile != PROFILE_ADVANCED){

  502.         src_x   = av_clip(  src_x, -16, s->mb_width  * 16);

  503.         src_y   = av_clip(  src_y, -16, s->mb_height * 16);

  504.         uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);

  505.         uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);

  506.     }else{

  507.         src_x   = av_clip(  src_x, -17, s->avctx->coded_width);

  508.         src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);

  509.         uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);

  510.         uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);

  511.     }

  512. 

  513.     srcY += src_y * s->linesize + src_x;

  514.     srcU += uvsrc_y * s->uvlinesize + uvsrc_x;

  515.     srcV += uvsrc_y * s->uvlinesize + uvsrc_x;

  516. 

  517.     /* for grayscale we should not try to read from unknown area */

  518.     if(s->flags & CODEC_FLAG_GRAY) {

  519.         srcU = s->edge_emu_buffer + 18 * s->linesize;

  520.         srcV = s->edge_emu_buffer + 18 * s->linesize;

  521.     }

  522. 

  523.     if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)

  524.        || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel*3

  525.        || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 16 - s->mspel*3){

  526.         uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;

  527. 

  528.         srcY -= s->mspel * (1 + s->linesize);

  529.         ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,

  530.                             src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);

  531.         srcY = s->edge_emu_buffer;

  532.         ff_emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8+1, 8+1,

  533.                             uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);

  534.         ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,

  535.                             uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);

  536.         srcU = uvbuf;

  537.         srcV = uvbuf + 16;

  538.         /* if we deal with range reduction we need to scale source blocks */

  539.         if(v->rangeredfrm) {

  540.             int i, j;

  541.             uint8_t *src, *src2;

  542. 

  543.             src = srcY;

  544.             for(j = 0; j < 17 + s->mspel*2; j++) {

  545.                 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;

  546.                 src += s->linesize;

  547.             }

  548.             src = srcU; src2 = srcV;

  549.             for(j = 0; j < 9; j++) {

  550.                 for(i = 0; i < 9; i++) {

  551.                     src[i] = ((src[i] - 128) >> 1) + 128;

  552.                     src2[i] = ((src2[i] - 128) >> 1) + 128;

  553.                 }

  554.                 src += s->uvlinesize;

  555.                 src2 += s->uvlinesize;

  556.             }

  557.         }

  558.         /* if we deal with intensity compensation we need to scale source blocks */

  559.         if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {

  560.             int i, j;

  561.             uint8_t *src, *src2;

  562. 

  563.             src = srcY;

  564.             for(j = 0; j < 17 + s->mspel*2; j++) {

  565.                 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = v->luty[src[i]];

  566.                 src += s->linesize;

  567.             }

  568.             src = srcU; src2 = srcV;

  569.             for(j = 0; j < 9; j++) {

  570.                 for(i = 0; i < 9; i++) {

  571.                     src[i] = v->lutuv[src[i]];

  572.                     src2[i] = v->lutuv[src2[i]];

  573.                 }

  574.                 src += s->uvlinesize;

  575.                 src2 += s->uvlinesize;

  576.             }

  577.         }

  578.         srcY += s->mspel * (1 + s->linesize);

  579.     }

  580. 

  581.     if(s->mspel) {

  582.         dxy = ((my & 3) << 2) | (mx & 3);

  583.         dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0]    , srcY    , s->linesize, v->rnd);

  584.         dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);

  585.         srcY += s->linesize * 8;

  586.         dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize    , srcY    , s->linesize, v->rnd);

  587.         dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);

  588.     } else { // hpel mc - always used for luma

  589.         dxy = (my & 2) | ((mx & 2) >> 1);

  590. 

  591.         if(!v->rnd)

  592.             dsp->put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);

  593.         else

  594.             dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);

  595.     }

  596. 

  597.     if(s->flags & CODEC_FLAG_GRAY) return;

  598.     /* Chroma MC always uses qpel bilinear */

  599.     uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);

  600.     uvmx = (uvmx&3)<<1;

  601.     uvmy = (uvmy&3)<<1;

  602.     if(!v->rnd){

  603.         dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);

  604.         dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);

  605.     }else{

  606.         dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);

  607.         dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);

  608.     }

  609. }

  610. 

  611. /** Do motion compensation for 4-MV macroblock - luminance block

  612.  */

  613. static void vc1_mc_4mv_luma(VC1Context *v, int n)

  614. {

  615.     MpegEncContext *s = &v->s;

  616.     DSPContext *dsp = &v->s.dsp;

  617.     uint8_t *srcY;

  618.     int dxy, mx, my, src_x, src_y;

  619.     int off;

  620. 

  621.     if(!v->s.last_picture.data[0])return;

  622.     mx = s->mv[0][n][0];

  623.     my = s->mv[0][n][1];

  624.     srcY = s->last_picture.data[0];

  625. 

  626.     off = s->linesize * 4 * (n&2) + (n&1) * 8;

  627. 

  628.     src_x = s->mb_x * 16 + (n&1) * 8 + (mx >> 2);

  629.     src_y = s->mb_y * 16 + (n&2) * 4 + (my >> 2);

  630. 

  631.     if(v->profile != PROFILE_ADVANCED){

  632.         src_x   = av_clip(  src_x, -16, s->mb_width  * 16);

  633.         src_y   = av_clip(  src_y, -16, s->mb_height * 16);

  634.     }else{

  635.         src_x   = av_clip(  src_x, -17, s->avctx->coded_width);

  636.         src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);

  637.     }

  638. 

  639.     srcY += src_y * s->linesize + src_x;

  640. 

  641.     if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)

  642.        || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 8 - s->mspel*2

  643.        || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 8 - s->mspel*2){

  644.         srcY -= s->mspel * (1 + s->linesize);

  645.         ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 9+s->mspel*2, 9+s->mspel*2,

  646.                             src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);

  647.         srcY = s->edge_emu_buffer;

  648.         /* if we deal with range reduction we need to scale source blocks */

  649.         if(v->rangeredfrm) {

  650.             int i, j;

  651.             uint8_t *src;

  652. 

  653.             src = srcY;

  654.             for(j = 0; j < 9 + s->mspel*2; j++) {

  655.                 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;

  656.                 src += s->linesize;

  657.             }

  658.         }

  659.         /* if we deal with intensity compensation we need to scale source blocks */

  660.         if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {

  661.             int i, j;

  662.             uint8_t *src;

  663. 

  664.             src = srcY;

  665.             for(j = 0; j < 9 + s->mspel*2; j++) {

  666.                 for(i = 0; i < 9 + s->mspel*2; i++) src[i] = v->luty[src[i]];

  667.                 src += s->linesize;

  668.             }

  669.         }

  670.         srcY += s->mspel * (1 + s->linesize);

  671.     }

  672. 

  673.     if(s->mspel) {

  674.         dxy = ((my & 3) << 2) | (mx & 3);

  675.         dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, v->rnd);

  676.     } else { // hpel mc - always used for luma

  677.         dxy = (my & 2) | ((mx & 2) >> 1);

  678.         if(!v->rnd)

  679.             dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);

  680.         else

  681.             dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);

  682.     }

  683. }

  684. 

  685. static inline int median4(int a, int b, int c, int d)

  686. {

  687.     if(a < b) {

  688.         if(c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;

  689.         else      return (FFMIN(b, c) + FFMAX(a, d)) / 2;

  690.     } else {

  691.         if(c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;

  692.         else      return (FFMIN(a, c) + FFMAX(b, d)) / 2;

  693.     }

  694. }

  695. 

  696. 

  697. /** Do motion compensation for 4-MV macroblock - both chroma blocks

  698.  */

  699. static void vc1_mc_4mv_chroma(VC1Context *v)

  700. {

  701.     MpegEncContext *s = &v->s;

  702.     DSPContext *dsp = &v->s.dsp;

  703.     uint8_t *srcU, *srcV;

  704.     int uvdxy, uvmx, uvmy, uvsrc_x, uvsrc_y;

  705.     int i, idx, tx = 0, ty = 0;

  706.     int mvx[4], mvy[4], intra[4];

  707.     static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};

  708. 

  709.     if(!v->s.last_picture.data[0])return;

  710.     if(s->flags & CODEC_FLAG_GRAY) return;

  711. 

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

  713.         mvx[i] = s->mv[0][i][0];

  714.         mvy[i] = s->mv[0][i][1];

  715.         intra[i] = v->mb_type[0][s->block_index[i]];

  716.     }

  717. 

  718.     /* calculate chroma MV vector from four luma MVs */

  719.     idx = (intra[3] << 3) | (intra[2] << 2) | (intra[1] << 1) | intra[0];

  720.     if(!idx) { // all blocks are inter

  721.         tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);

  722.         ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);

  723.     } else if(count[idx] == 1) { // 3 inter blocks

  724.         switch(idx) {

  725.         case 0x1:

  726.             tx = mid_pred(mvx[1], mvx[2], mvx[3]);

  727.             ty = mid_pred(mvy[1], mvy[2], mvy[3]);

  728.             break;

  729.         case 0x2:

  730.             tx = mid_pred(mvx[0], mvx[2], mvx[3]);

  731.             ty = mid_pred(mvy[0], mvy[2], mvy[3]);

  732.             break;

  733.         case 0x4:

  734.             tx = mid_pred(mvx[0], mvx[1], mvx[3]);

  735.             ty = mid_pred(mvy[0], mvy[1], mvy[3]);

  736.             break;

  737.         case 0x8:

  738.             tx = mid_pred(mvx[0], mvx[1], mvx[2]);

  739.             ty = mid_pred(mvy[0], mvy[1], mvy[2]);

  740.             break;

  741.         }

  742.     } else if(count[idx] == 2) {

  743.         int t1 = 0, t2 = 0;

  744.         for(i=0; i<3;i++) if(!intra[i]) {t1 = i; break;}

  745.         for(i= t1+1; i<4; i++)if(!intra[i]) {t2 = i; break;}

  746.         tx = (mvx[t1] + mvx[t2]) / 2;

  747.         ty = (mvy[t1] + mvy[t2]) / 2;

  748.     } else {

  749.         s->current_picture.motion_val[1][s->block_index[0]][0] = 0;

  750.         s->current_picture.motion_val[1][s->block_index[0]][1] = 0;

  751.         return; //no need to do MC for inter blocks

  752.     }

  753. 

  754.     s->current_picture.motion_val[1][s->block_index[0]][0] = tx;

  755.     s->current_picture.motion_val[1][s->block_index[0]][1] = ty;

  756.     uvmx = (tx + ((tx&3) == 3)) >> 1;

  757.     uvmy = (ty + ((ty&3) == 3)) >> 1;

  758.     if(v->fastuvmc) {

  759.         uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));

  760.         uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));

  761.     }

  762. 

  763.     uvsrc_x = s->mb_x * 8 + (uvmx >> 2);

  764.     uvsrc_y = s->mb_y * 8 + (uvmy >> 2);

  765. 

  766.     if(v->profile != PROFILE_ADVANCED){

  767.         uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);

  768.         uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);

  769.     }else{

  770.         uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);

  771.         uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);

  772.     }

  773. 

  774.     srcU = s->last_picture.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;

  775.     srcV = s->last_picture.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;

  776.     if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)

  777.        || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9

  778.        || (unsigned)uvsrc_y > (s->v_edge_pos >> 1) - 9){

  779.         ff_emulated_edge_mc(s->edge_emu_buffer     , srcU, s->uvlinesize, 8+1, 8+1,

  780.                             uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);

  781.         ff_emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, 8+1, 8+1,

  782.                             uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);

  783.         srcU = s->edge_emu_buffer;

  784.         srcV = s->edge_emu_buffer + 16;

  785. 

  786.         /* if we deal with range reduction we need to scale source blocks */

  787.         if(v->rangeredfrm) {

  788.             int i, j;

  789.             uint8_t *src, *src2;

  790. 

  791.             src = srcU; src2 = srcV;

  792.             for(j = 0; j < 9; j++) {

  793.                 for(i = 0; i < 9; i++) {

  794.                     src[i] = ((src[i] - 128) >> 1) + 128;

  795.                     src2[i] = ((src2[i] - 128) >> 1) + 128;

  796.                 }

  797.                 src += s->uvlinesize;

  798.                 src2 += s->uvlinesize;

  799.             }

  800.         }

  801.         /* if we deal with intensity compensation we need to scale source blocks */

  802.         if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {

  803.             int i, j;

  804.             uint8_t *src, *src2;

  805. 

  806.             src = srcU; src2 = srcV;

  807.             for(j = 0; j < 9; j++) {

  808.                 for(i = 0; i < 9; i++) {

  809.                     src[i] = v->lutuv[src[i]];

  810.                     src2[i] = v->lutuv[src2[i]];

  811.                 }

  812.                 src += s->uvlinesize;

  813.                 src2 += s->uvlinesize;

  814.             }

  815.         }

  816.     }

  817. 

  818.     /* Chroma MC always uses qpel bilinear */

  819.     uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);

  820.     uvmx = (uvmx&3)<<1;

  821.     uvmy = (uvmy&3)<<1;

  822.     if(!v->rnd){

  823.         dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);

  824.         dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);

  825.     }else{

  826.         dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);

  827.         dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);

  828.     }

  829. }

  830. 

  831. static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb);

  832. 

  833. /**

  834.  * Decode Simple/Main Profiles sequence header

  835.  * @see Figure 7-8, p16-17

  836.  * @param avctx Codec context

  837.  * @param gb GetBit context initialized from Codec context extra_data

  838.  * @return Status

  839.  */

  840. static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb)

  841. {

  842.     VC1Context *v = avctx->priv_data;

  843. 

  844.     av_log(avctx, AV_LOG_DEBUG, "Header: %0X\n", show_bits(gb, 32));

  845.     v->profile = get_bits(gb, 2);

  846.     if (v->profile == PROFILE_COMPLEX)

  847.     {

  848.         av_log(avctx, AV_LOG_ERROR, "WMV3 Complex Profile is not fully supported\n");

  849.     }

  850. 

  851.     if (v->profile == PROFILE_ADVANCED)

  852.     {

  853.         v->zz_8x4 = ff_vc1_adv_progressive_8x4_zz;

  854.         v->zz_4x8 = ff_vc1_adv_progressive_4x8_zz;

  855.         return decode_sequence_header_adv(v, gb);

  856.     }

  857.     else

  858.     {

  859.         v->zz_8x4 = wmv2_scantableA;

  860.         v->zz_4x8 = wmv2_scantableB;

  861.         v->res_sm = get_bits(gb, 2); //reserved

  862.         if (v->res_sm)

  863.         {

  864.             av_log(avctx, AV_LOG_ERROR,

  865.                    "Reserved RES_SM=%i is forbidden\n", v->res_sm);

  866.             return -1;

  867.         }

  868.     }

  869. 

  870.     // (fps-2)/4 (->30)

  871.     v->frmrtq_postproc = get_bits(gb, 3); //common

  872.     // (bitrate-32kbps)/64kbps

  873.     v->bitrtq_postproc = get_bits(gb, 5); //common

  874.     v->s.loop_filter = get_bits1(gb); //common

  875.     if(v->s.loop_filter == 1 && v->profile == PROFILE_SIMPLE)

  876.     {

  877.         av_log(avctx, AV_LOG_ERROR,

  878.                "LOOPFILTER shell not be enabled in simple profile\n");

  879.     }

  880. 

  881.     v->res_x8 = get_bits1(gb); //reserved

  882.     v->multires = get_bits1(gb);

  883.     v->res_fasttx = get_bits1(gb);

  884.     if (!v->res_fasttx)

  885.     {

  886.         v->s.dsp.vc1_inv_trans_8x8 = ff_simple_idct;

  887.         v->s.dsp.vc1_inv_trans_8x4 = ff_simple_idct84_add;

  888.         v->s.dsp.vc1_inv_trans_4x8 = ff_simple_idct48_add;

  889.         v->s.dsp.vc1_inv_trans_4x4 = ff_simple_idct44_add;

  890.     }

  891. 

  892.     v->fastuvmc =  get_bits1(gb); //common

  893.     if (!v->profile && !v->fastuvmc)

  894.     {

  895.         av_log(avctx, AV_LOG_ERROR,

  896.                "FASTUVMC unavailable in Simple Profile\n");

  897.         return -1;

  898.     }

  899.     v->extended_mv =  get_bits1(gb); //common

  900.     if (!v->profile && v->extended_mv)

  901.     {

  902.         av_log(avctx, AV_LOG_ERROR,

  903.                "Extended MVs unavailable in Simple Profile\n");

  904.         return -1;

  905.     }

  906.     v->dquant =  get_bits(gb, 2); //common

  907.     v->vstransform =  get_bits1(gb); //common

  908. 

  909.     v->res_transtab = get_bits1(gb);

  910.     if (v->res_transtab)

  911.     {

  912.         av_log(avctx, AV_LOG_ERROR,

  913.                "1 for reserved RES_TRANSTAB is forbidden\n");

  914.         return -1;

  915.     }

  916. 

  917.     v->overlap = get_bits1(gb); //common

  918. 

  919.     v->s.resync_marker = get_bits1(gb);

  920.     v->rangered = get_bits1(gb);

  921.     if (v->rangered && v->profile == PROFILE_SIMPLE)

  922.     {

  923.         av_log(avctx, AV_LOG_INFO,

  924.                "RANGERED should be set to 0 in simple profile\n");

  925.     }

  926. 

  927.     v->s.max_b_frames = avctx->max_b_frames = get_bits(gb, 3); //common

  928.     v->quantizer_mode = get_bits(gb, 2); //common

  929. 

  930.     v->finterpflag = get_bits1(gb); //common

  931.     v->res_rtm_flag = get_bits1(gb); //reserved

  932.     if (!v->res_rtm_flag)

  933.     {

  934. //            av_log(avctx, AV_LOG_ERROR,

  935. //                   "0 for reserved RES_RTM_FLAG is forbidden\n");

  936.         av_log(avctx, AV_LOG_ERROR,

  937.                "Old WMV3 version detected, only I-frames will be decoded\n");

  938.         //return -1;

  939.     }

  940.     //TODO: figure out what they mean (always 0x402F)

  941.     if(!v->res_fasttx) skip_bits(gb, 16);

  942.     av_log(avctx, AV_LOG_DEBUG,

  943.                "Profile %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"

  944.                "LoopFilter=%i, MultiRes=%i, FastUVMC=%i, Extended MV=%i\n"

  945.                "Rangered=%i, VSTransform=%i, Overlap=%i, SyncMarker=%i\n"

  946.                "DQuant=%i, Quantizer mode=%i, Max B frames=%i\n",

  947.                v->profile, v->frmrtq_postproc, v->bitrtq_postproc,

  948.                v->s.loop_filter, v->multires, v->fastuvmc, v->extended_mv,

  949.                v->rangered, v->vstransform, v->overlap, v->s.resync_marker,

  950.                v->dquant, v->quantizer_mode, avctx->max_b_frames

  951.                );

  952.     return 0;

  953. }

  954. 

  955. static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb)

  956. {

  957.     v->res_rtm_flag = 1;

  958.     v->level = get_bits(gb, 3);

  959.     if(v->level >= 5)

  960.     {

  961.         av_log(v->s.avctx, AV_LOG_ERROR, "Reserved LEVEL %i\n",v->level);

  962.     }

  963.     v->chromaformat = get_bits(gb, 2);

  964.     if (v->chromaformat != 1)

  965.     {

  966.         av_log(v->s.avctx, AV_LOG_ERROR,

  967.                "Only 4:2:0 chroma format supported\n");

  968.         return -1;

  969.     }

  970. 

  971.     // (fps-2)/4 (->30)

  972.     v->frmrtq_postproc = get_bits(gb, 3); //common

  973.     // (bitrate-32kbps)/64kbps

  974.     v->bitrtq_postproc = get_bits(gb, 5); //common

  975.     v->postprocflag = get_bits1(gb); //common

  976. 

  977.     v->s.avctx->coded_width = (get_bits(gb, 12) + 1) << 1;

  978.     v->s.avctx->coded_height = (get_bits(gb, 12) + 1) << 1;

  979.     v->s.avctx->width = v->s.avctx->coded_width;

  980.     v->s.avctx->height = v->s.avctx->coded_height;

  981.     v->broadcast = get_bits1(gb);

  982.     v->interlace = get_bits1(gb);

  983.     v->tfcntrflag = get_bits1(gb);

  984.     v->finterpflag = get_bits1(gb);

  985.     skip_bits1(gb); // reserved

  986. 

  987.     v->s.h_edge_pos = v->s.avctx->coded_width;

  988.     v->s.v_edge_pos = v->s.avctx->coded_height;

  989. 

  990.     av_log(v->s.avctx, AV_LOG_DEBUG,

  991.                "Advanced Profile level %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"

  992.                "LoopFilter=%i, ChromaFormat=%i, Pulldown=%i, Interlace: %i\n"

  993.                "TFCTRflag=%i, FINTERPflag=%i\n",

  994.                v->level, v->frmrtq_postproc, v->bitrtq_postproc,

  995.                v->s.loop_filter, v->chromaformat, v->broadcast, v->interlace,

  996.                v->tfcntrflag, v->finterpflag

  997.                );

  998. 

  999.     v->psf = get_bits1(gb);

  1000.     if(v->psf) { //PsF, 6.1.13

  1001.         av_log(v->s.avctx, AV_LOG_ERROR, "Progressive Segmented Frame mode: not supported (yet)\n");

  1002.         return -1;

  1003.     }

  1004.     v->s.max_b_frames = v->s.avctx->max_b_frames = 7;

  1005.     if(get_bits1(gb)) { //Display Info - decoding is not affected by it

  1006.         int w, h, ar = 0;

  1007.         av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");

  1008.         v->s.avctx->width  = v->s.width  = w = get_bits(gb, 14) + 1;

  1009.         v->s.avctx->height = v->s.height = h = get_bits(gb, 14) + 1;

  1010.         av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h);

  1011.         if(get_bits1(gb))

  1012.             ar = get_bits(gb, 4);

  1013.         if(ar && ar < 14){

  1014.             v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];

  1015.         }else if(ar == 15){

  1016.             w = get_bits(gb, 8);

  1017.             h = get_bits(gb, 8);

  1018.             v->s.avctx->sample_aspect_ratio = (AVRational){w, h};

  1019.         }

  1020. 

  1021.         if(get_bits1(gb)){ //framerate stuff

  1022.             if(get_bits1(gb)) {

  1023.                 v->s.avctx->time_base.num = 32;

  1024.                 v->s.avctx->time_base.den = get_bits(gb, 16) + 1;

  1025.             } else {

  1026.                 int nr, dr;

  1027.                 nr = get_bits(gb, 8);

  1028.                 dr = get_bits(gb, 4);

  1029.                 if(nr && nr < 8 && dr && dr < 3){

  1030.                     v->s.avctx->time_base.num = ff_vc1_fps_dr[dr - 1];

  1031.                     v->s.avctx->time_base.den = ff_vc1_fps_nr[nr - 1] * 1000;

  1032.                 }

  1033.             }

  1034.         }

  1035. 

  1036.         if(get_bits1(gb)){

  1037.             v->color_prim = get_bits(gb, 8);

  1038.             v->transfer_char = get_bits(gb, 8);

  1039.             v->matrix_coef = get_bits(gb, 8);

  1040.         }

  1041.     }

  1042. 

  1043.     v->hrd_param_flag = get_bits1(gb);

  1044.     if(v->hrd_param_flag) {

  1045.         int i;

  1046.         v->hrd_num_leaky_buckets = get_bits(gb, 5);

  1047.         skip_bits(gb, 4); //bitrate exponent

  1048.         skip_bits(gb, 4); //buffer size exponent

  1049.         for(i = 0; i < v->hrd_num_leaky_buckets; i++) {

  1050.             skip_bits(gb, 16); //hrd_rate[n]

  1051.             skip_bits(gb, 16); //hrd_buffer[n]

  1052.         }

  1053.     }

  1054.     return 0;

  1055. }

  1056. 

  1057. static int decode_entry_point(AVCodecContext *avctx, GetBitContext *gb)

  1058. {

  1059.     VC1Context *v = avctx->priv_data;

  1060.     int i, blink, clentry, refdist;

  1061. 

  1062.     av_log(avctx, AV_LOG_DEBUG, "Entry point: %08X\n", show_bits_long(gb, 32));

  1063.     blink = get_bits1(gb); // broken link

  1064.     clentry = get_bits1(gb); // closed entry

  1065.     v->panscanflag = get_bits1(gb);

  1066.     refdist = get_bits1(gb); // refdist flag

  1067.     v->s.loop_filter = get_bits1(gb);

  1068.     v->fastuvmc = get_bits1(gb);

  1069.     v->extended_mv = get_bits1(gb);

  1070.     v->dquant = get_bits(gb, 2);

  1071.     v->vstransform = get_bits1(gb);

  1072.     v->overlap = get_bits1(gb);

  1073.     v->quantizer_mode = get_bits(gb, 2);

  1074. 

  1075.     if(v->hrd_param_flag){

  1076.         for(i = 0; i < v->hrd_num_leaky_buckets; i++) {

  1077.             skip_bits(gb, 8); //hrd_full[n]

  1078.         }

  1079.     }

  1080. 

  1081.     if(get_bits1(gb)){

  1082.         avctx->coded_width = (get_bits(gb, 12)+1)<<1;

  1083.         avctx->coded_height = (get_bits(gb, 12)+1)<<1;

  1084.     }

  1085.     if(v->extended_mv)

  1086.         v->extended_dmv = get_bits1(gb);

  1087.     if(get_bits1(gb)) {

  1088.         av_log(avctx, AV_LOG_ERROR, "Luma scaling is not supported, expect wrong picture\n");

  1089.         skip_bits(gb, 3); // Y range, ignored for now

  1090.     }

  1091.     if(get_bits1(gb)) {

  1092.         av_log(avctx, AV_LOG_ERROR, "Chroma scaling is not supported, expect wrong picture\n");

  1093.         skip_bits(gb, 3); // UV range, ignored for now

  1094.     }

  1095. 

  1096.     av_log(avctx, AV_LOG_DEBUG, "Entry point info:\n"

  1097.         "BrokenLink=%i, ClosedEntry=%i, PanscanFlag=%i\n"

  1098.         "RefDist=%i, Postproc=%i, FastUVMC=%i, ExtMV=%i\n"

  1099.         "DQuant=%i, VSTransform=%i, Overlap=%i, Qmode=%i\n",

  1100.         blink, clentry, v->panscanflag, refdist, v->s.loop_filter,

  1101.         v->fastuvmc, v->extended_mv, v->dquant, v->vstransform, v->overlap, v->quantizer_mode);

  1102. 

  1103.     return 0;

  1104. }

  1105. 

  1106. static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb)

  1107. {

  1108.     int pqindex, lowquant, status;

  1109. 

  1110.     if(v->finterpflag) v->interpfrm = get_bits1(gb);

  1111.     skip_bits(gb, 2); //framecnt unused

  1112.     v->rangeredfrm = 0;

  1113.     if (v->rangered) v->rangeredfrm = get_bits1(gb);

  1114.     v->s.pict_type = get_bits1(gb);

  1115.     if (v->s.avctx->max_b_frames) {

  1116.         if (!v->s.pict_type) {

  1117.             if (get_bits1(gb)) v->s.pict_type = FF_I_TYPE;

  1118.             else v->s.pict_type = FF_B_TYPE;

  1119.         } else v->s.pict_type = FF_P_TYPE;

  1120.     } else v->s.pict_type = v->s.pict_type ? FF_P_TYPE : FF_I_TYPE;

  1121. 

  1122.     v->bi_type = 0;

  1123.     if(v->s.pict_type == FF_B_TYPE) {

  1124.         v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);

  1125.         v->bfraction = ff_vc1_bfraction_lut[v->bfraction];

  1126.         if(v->bfraction == 0) {

  1127.             v->s.pict_type = FF_BI_TYPE;

  1128.         }

  1129.     }

  1130.     if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)

  1131.         skip_bits(gb, 7); // skip buffer fullness

  1132. 

  1133.     /* calculate RND */

  1134.     if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)

  1135.         v->rnd = 1;

  1136.     if(v->s.pict_type == FF_P_TYPE)

  1137.         v->rnd ^= 1;

  1138. 

  1139.     /* Quantizer stuff */

  1140.     pqindex = get_bits(gb, 5);

  1141.     if(!pqindex) return -1;

  1142.     if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)

  1143.         v->pq = ff_vc1_pquant_table[0][pqindex];

  1144.     else

  1145.         v->pq = ff_vc1_pquant_table[1][pqindex];

  1146. 

  1147.     v->pquantizer = 1;

  1148.     if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)

  1149.         v->pquantizer = pqindex < 9;

  1150.     if (v->quantizer_mode == QUANT_NON_UNIFORM)

  1151.         v->pquantizer = 0;

  1152.     v->pqindex = pqindex;

  1153.     if (pqindex < 9) v->halfpq = get_bits1(gb);

  1154.     else v->halfpq = 0;

  1155.     if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)

  1156.         v->pquantizer = get_bits1(gb);

  1157.     v->dquantfrm = 0;

  1158.     if (v->extended_mv == 1) v->mvrange = get_unary(gb, 0, 3);

  1159.     v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13

  1160.     v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11

  1161.     v->range_x = 1 << (v->k_x - 1);

  1162.     v->range_y = 1 << (v->k_y - 1);

  1163.     if (v->profile == PROFILE_ADVANCED)

  1164.     {

  1165.         if (v->postprocflag) v->postproc = get_bits1(gb);

  1166.     }

  1167.     else

  1168.         if (v->multires && v->s.pict_type != FF_B_TYPE) v->respic = get_bits(gb, 2);

  1169. 

  1170.     if(v->res_x8 && (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)){

  1171.         v->x8_type = get_bits1(gb);

  1172.     }else v->x8_type = 0;

  1173. //av_log(v->s.avctx, AV_LOG_INFO, "%c Frame: QP=[%i]%i (+%i/2) %i\n",

  1174. //        (v->s.pict_type == FF_P_TYPE) ? 'P' : ((v->s.pict_type == FF_I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm);

  1175. 

  1176.     if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;

  1177. 

  1178.     switch(v->s.pict_type) {

  1179.     case FF_P_TYPE:

  1180.         if (v->pq < 5) v->tt_index = 0;

  1181.         else if(v->pq < 13) v->tt_index = 1;

  1182.         else v->tt_index = 2;

  1183. 

  1184.         lowquant = (v->pq > 12) ? 0 : 1;

  1185.         v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];

  1186.         if (v->mv_mode == MV_PMODE_INTENSITY_COMP)

  1187.         {

  1188.             int scale, shift, i;

  1189.             v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];

  1190.             v->lumscale = get_bits(gb, 6);

  1191.             v->lumshift = get_bits(gb, 6);

  1192.             v->use_ic = 1;

  1193.             /* fill lookup tables for intensity compensation */

  1194.             if(!v->lumscale) {

  1195.                 scale = -64;

  1196.                 shift = (255 - v->lumshift * 2) << 6;

  1197.                 if(v->lumshift > 31)

  1198.                     shift += 128 << 6;

  1199.             } else {

  1200.                 scale = v->lumscale + 32;

  1201.                 if(v->lumshift > 31)

  1202.                     shift = (v->lumshift - 64) << 6;

  1203.                 else

  1204.                     shift = v->lumshift << 6;

  1205.             }

  1206.             for(i = 0; i < 256; i++) {

  1207.                 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);

  1208.                 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);

  1209.             }

  1210.         }

  1211.         if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)

  1212.             v->s.quarter_sample = 0;

  1213.         else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {

  1214.             if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)

  1215.                 v->s.quarter_sample = 0;

  1216.             else

  1217.                 v->s.quarter_sample = 1;

  1218.         } else

  1219.             v->s.quarter_sample = 1;

  1220.         v->s.mspel = !(v->mv_mode == MV_PMODE_1MV_HPEL_BILIN || (v->mv_mode == MV_PMODE_INTENSITY_COMP && v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN));

  1221. 

  1222.         if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&

  1223.                  v->mv_mode2 == MV_PMODE_MIXED_MV)

  1224.                 || v->mv_mode == MV_PMODE_MIXED_MV)

  1225.         {

  1226.             status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);

  1227.             if (status < 0) return -1;

  1228.             av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "

  1229.                    "Imode: %i, Invert: %i\n", status>>1, status&1);

  1230.         } else {

  1231.             v->mv_type_is_raw = 0;

  1232.             memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);

  1233.         }

  1234.         status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);

  1235.         if (status < 0) return -1;

  1236.         av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "

  1237.                "Imode: %i, Invert: %i\n", status>>1, status&1);

  1238. 

  1239.         /* Hopefully this is correct for P frames */

  1240.         v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables

  1241.         v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];

  1242. 

  1243.         if (v->dquant)

  1244.         {

  1245.             av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");

  1246.             vop_dquant_decoding(v);

  1247.         }

  1248. 

  1249.         v->ttfrm = 0; //FIXME Is that so ?

  1250.         if (v->vstransform)

  1251.         {

  1252.             v->ttmbf = get_bits1(gb);

  1253.             if (v->ttmbf)

  1254.             {

  1255.                 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];

  1256.             }

  1257.         } else {

  1258.             v->ttmbf = 1;

  1259.             v->ttfrm = TT_8X8;

  1260.         }

  1261.         break;

  1262.     case FF_B_TYPE:

  1263.         if (v->pq < 5) v->tt_index = 0;

  1264.         else if(v->pq < 13) v->tt_index = 1;

  1265.         else v->tt_index = 2;

  1266. 

  1267.         lowquant = (v->pq > 12) ? 0 : 1;

  1268.         v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;

  1269.         v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);

  1270.         v->s.mspel = v->s.quarter_sample;

  1271. 

  1272.         status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);

  1273.         if (status < 0) return -1;

  1274.         av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "

  1275.                "Imode: %i, Invert: %i\n", status>>1, status&1);

  1276.         status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);

  1277.         if (status < 0) return -1;

  1278.         av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "

  1279.                "Imode: %i, Invert: %i\n", status>>1, status&1);

  1280. 

  1281.         v->s.mv_table_index = get_bits(gb, 2);

  1282.         v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];

  1283. 

  1284.         if (v->dquant)

  1285.         {

  1286.             av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");

  1287.             vop_dquant_decoding(v);

  1288.         }

  1289. 

  1290.         v->ttfrm = 0;

  1291.         if (v->vstransform)

  1292.         {

  1293.             v->ttmbf = get_bits1(gb);

  1294.             if (v->ttmbf)

  1295.             {

  1296.                 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];

  1297.             }

  1298.         } else {

  1299.             v->ttmbf = 1;

  1300.             v->ttfrm = TT_8X8;

  1301.         }

  1302.         break;

  1303.     }

  1304. 

  1305.     if(!v->x8_type)

  1306.     {

  1307.         /* AC Syntax */

  1308.         v->c_ac_table_index = decode012(gb);

  1309.         if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)

  1310.         {

  1311.             v->y_ac_table_index = decode012(gb);

  1312.         }

  1313.         /* DC Syntax */

  1314.         v->s.dc_table_index = get_bits1(gb);

  1315.     }

  1316. 

  1317.     if(v->s.pict_type == FF_BI_TYPE) {

  1318.         v->s.pict_type = FF_B_TYPE;

  1319.         v->bi_type = 1;

  1320.     }

  1321.     return 0;

  1322. }

  1323. 

  1324. static int vc1_parse_frame_header_adv(VC1Context *v, GetBitContext* gb)

  1325. {

  1326.     int pqindex, lowquant;

  1327.     int status;

  1328. 

  1329.     v->p_frame_skipped = 0;

  1330. 

  1331.     if(v->interlace){

  1332.         v->fcm = decode012(gb);

  1333.         if(v->fcm) return -1; // interlaced frames/fields are not implemented

  1334.     }

  1335.     switch(get_unary(gb, 0, 4)) {

  1336.     case 0:

  1337.         v->s.pict_type = FF_P_TYPE;

  1338.         break;

  1339.     case 1:

  1340.         v->s.pict_type = FF_B_TYPE;

  1341.         break;

  1342.     case 2:

  1343.         v->s.pict_type = FF_I_TYPE;

  1344.         break;

  1345.     case 3:

  1346.         v->s.pict_type = FF_BI_TYPE;

  1347.         break;

  1348.     case 4:

  1349.         v->s.pict_type = FF_P_TYPE; // skipped pic

  1350.         v->p_frame_skipped = 1;

  1351.         return 0;

  1352.     }

  1353.     if(v->tfcntrflag)

  1354.         skip_bits(gb, 8);

  1355.     if(v->broadcast) {

  1356.         if(!v->interlace || v->psf) {

  1357.             v->rptfrm = get_bits(gb, 2);

  1358.         } else {

  1359.             v->tff = get_bits1(gb);

  1360.             v->rptfrm = get_bits1(gb);

  1361.         }

  1362.     }

  1363.     if(v->panscanflag) {

  1364.         //...

  1365.     }

  1366.     v->rnd = get_bits1(gb);

  1367.     if(v->interlace)

  1368.         v->uvsamp = get_bits1(gb);

  1369.     if(v->finterpflag) v->interpfrm = get_bits1(gb);

  1370.     if(v->s.pict_type == FF_B_TYPE) {

  1371.         v->bfraction = get_vlc2(gb, ff_vc1_bfraction_vlc.table, VC1_BFRACTION_VLC_BITS, 1);

  1372.         v->bfraction = ff_vc1_bfraction_lut[v->bfraction];

  1373.         if(v->bfraction == 0) {

  1374.             v->s.pict_type = FF_BI_TYPE; /* XXX: should not happen here */

  1375.         }

  1376.     }

  1377.     pqindex = get_bits(gb, 5);

  1378.     if(!pqindex) return -1;

  1379.     v->pqindex = pqindex;

  1380.     if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)

  1381.         v->pq = ff_vc1_pquant_table[0][pqindex];

  1382.     else

  1383.         v->pq = ff_vc1_pquant_table[1][pqindex];

  1384. 

  1385.     v->pquantizer = 1;

  1386.     if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)

  1387.         v->pquantizer = pqindex < 9;

  1388.     if (v->quantizer_mode == QUANT_NON_UNIFORM)

  1389.         v->pquantizer = 0;

  1390.     v->pqindex = pqindex;

  1391.     if (pqindex < 9) v->halfpq = get_bits1(gb);

  1392.     else v->halfpq = 0;

  1393.     if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)

  1394.         v->pquantizer = get_bits1(gb);

  1395. 

  1396.     if(v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_P_TYPE) v->use_ic = 0;

  1397. 

  1398.     switch(v->s.pict_type) {

  1399.     case FF_I_TYPE:

  1400.     case FF_BI_TYPE:

  1401.         status = bitplane_decoding(v->acpred_plane, &v->acpred_is_raw, v);

  1402.         if (status < 0) return -1;

  1403.         av_log(v->s.avctx, AV_LOG_DEBUG, "ACPRED plane encoding: "

  1404.                 "Imode: %i, Invert: %i\n", status>>1, status&1);

  1405.         v->condover = CONDOVER_NONE;

  1406.         if(v->overlap && v->pq <= 8) {

  1407.             v->condover = decode012(gb);

  1408.             if(v->condover == CONDOVER_SELECT) {

  1409.                 status = bitplane_decoding(v->over_flags_plane, &v->overflg_is_raw, v);

  1410.                 if (status < 0) return -1;

  1411.                 av_log(v->s.avctx, AV_LOG_DEBUG, "CONDOVER plane encoding: "

  1412.                         "Imode: %i, Invert: %i\n", status>>1, status&1);

  1413.             }

  1414.         }

  1415.         break;

  1416.     case FF_P_TYPE:

  1417.         if(v->postprocflag)

  1418.             v->postproc = get_bits1(gb);

  1419.         if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);

  1420.         else v->mvrange = 0;

  1421.         v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13

  1422.         v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11

  1423.         v->range_x = 1 << (v->k_x - 1);

  1424.         v->range_y = 1 << (v->k_y - 1);

  1425. 

  1426.         if (v->pq < 5) v->tt_index = 0;

  1427.         else if(v->pq < 13) v->tt_index = 1;

  1428.         else v->tt_index = 2;

  1429. 

  1430.         lowquant = (v->pq > 12) ? 0 : 1;

  1431.         v->mv_mode = ff_vc1_mv_pmode_table[lowquant][get_unary(gb, 1, 4)];

  1432.         if (v->mv_mode == MV_PMODE_INTENSITY_COMP)

  1433.         {

  1434.             int scale, shift, i;

  1435.             v->mv_mode2 = ff_vc1_mv_pmode_table2[lowquant][get_unary(gb, 1, 3)];

  1436.             v->lumscale = get_bits(gb, 6);

  1437.             v->lumshift = get_bits(gb, 6);

  1438.             /* fill lookup tables for intensity compensation */

  1439.             if(!v->lumscale) {

  1440.                 scale = -64;

  1441.                 shift = (255 - v->lumshift * 2) << 6;

  1442.                 if(v->lumshift > 31)

  1443.                     shift += 128 << 6;

  1444.             } else {

  1445.                 scale = v->lumscale + 32;

  1446.                 if(v->lumshift > 31)

  1447.                     shift = (v->lumshift - 64) << 6;

  1448.                 else

  1449.                     shift = v->lumshift << 6;

  1450.             }

  1451.             for(i = 0; i < 256; i++) {

  1452.                 v->luty[i] = av_clip_uint8((scale * i + shift + 32) >> 6);

  1453.                 v->lutuv[i] = av_clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);

  1454.             }

  1455.             v->use_ic = 1;

  1456.         }

  1457.         if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)

  1458.             v->s.quarter_sample = 0;

  1459.         else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {

  1460.             if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)

  1461.                 v->s.quarter_sample = 0;

  1462.             else

  1463.                 v->s.quarter_sample = 1;

  1464.         } else

  1465.             v->s.quarter_sample = 1;

  1466.         v->s.mspel = !(v->mv_mode == MV_PMODE_1MV_HPEL_BILIN || (v->mv_mode == MV_PMODE_INTENSITY_COMP && v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN));

  1467. 

  1468.         if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&

  1469.                  v->mv_mode2 == MV_PMODE_MIXED_MV)

  1470.                 || v->mv_mode == MV_PMODE_MIXED_MV)

  1471.         {

  1472.             status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);

  1473.             if (status < 0) return -1;

  1474.             av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "

  1475.                    "Imode: %i, Invert: %i\n", status>>1, status&1);

  1476.         } else {

  1477.             v->mv_type_is_raw = 0;

  1478.             memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);

  1479.         }

  1480.         status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);

  1481.         if (status < 0) return -1;

  1482.         av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "

  1483.                "Imode: %i, Invert: %i\n", status>>1, status&1);

  1484. 

  1485.         /* Hopefully this is correct for P frames */

  1486.         v->s.mv_table_index = get_bits(gb, 2); //but using ff_vc1_ tables

  1487.         v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];

  1488.         if (v->dquant)

  1489.         {

  1490.             av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");

  1491.             vop_dquant_decoding(v);

  1492.         }

  1493. 

  1494.         v->ttfrm = 0; //FIXME Is that so ?

  1495.         if (v->vstransform)

  1496.         {

  1497.             v->ttmbf = get_bits1(gb);

  1498.             if (v->ttmbf)

  1499.             {

  1500.                 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];

  1501.             }

  1502.         } else {

  1503.             v->ttmbf = 1;

  1504.             v->ttfrm = TT_8X8;

  1505.         }

  1506.         break;

  1507.     case FF_B_TYPE:

  1508.         if(v->postprocflag)

  1509.             v->postproc = get_bits1(gb);

  1510.         if (v->extended_mv) v->mvrange = get_unary(gb, 0, 3);

  1511.         else v->mvrange = 0;

  1512.         v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13

  1513.         v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11

  1514.         v->range_x = 1 << (v->k_x - 1);

  1515.         v->range_y = 1 << (v->k_y - 1);

  1516. 

  1517.         if (v->pq < 5) v->tt_index = 0;

  1518.         else if(v->pq < 13) v->tt_index = 1;

  1519.         else v->tt_index = 2;

  1520. 

  1521.         lowquant = (v->pq > 12) ? 0 : 1;

  1522.         v->mv_mode = get_bits1(gb) ? MV_PMODE_1MV : MV_PMODE_1MV_HPEL_BILIN;

  1523.         v->s.quarter_sample = (v->mv_mode == MV_PMODE_1MV);

  1524.         v->s.mspel = v->s.quarter_sample;

  1525. 

  1526.         status = bitplane_decoding(v->direct_mb_plane, &v->dmb_is_raw, v);

  1527.         if (status < 0) return -1;

  1528.         av_log(v->s.avctx, AV_LOG_DEBUG, "MB Direct Type plane encoding: "

  1529.                "Imode: %i, Invert: %i\n", status>>1, status&1);

  1530.         status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);

  1531.         if (status < 0) return -1;

  1532.         av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "

  1533.                "Imode: %i, Invert: %i\n", status>>1, status&1);

  1534. 

  1535.         v->s.mv_table_index = get_bits(gb, 2);

  1536.         v->cbpcy_vlc = &ff_vc1_cbpcy_p_vlc[get_bits(gb, 2)];

  1537. 

  1538.         if (v->dquant)

  1539.         {

  1540.             av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");

  1541.             vop_dquant_decoding(v);

  1542.         }

  1543. 

  1544.         v->ttfrm = 0;

  1545.         if (v->vstransform)

  1546.         {

  1547.             v->ttmbf = get_bits1(gb);

  1548.             if (v->ttmbf)

  1549.             {

  1550.                 v->ttfrm = ff_vc1_ttfrm_to_tt[get_bits(gb, 2)];

  1551.             }

  1552.         } else {

  1553.             v->ttmbf = 1;

  1554.             v->ttfrm = TT_8X8;

  1555.         }

  1556.         break;

  1557.     }

  1558. 

  1559.     /* AC Syntax */

  1560.     v->c_ac_table_index = decode012(gb);

  1561.     if (v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE)

  1562.     {

  1563.         v->y_ac_table_index = decode012(gb);

  1564.     }

  1565.     /* DC Syntax */

  1566.     v->s.dc_table_index = get_bits1(gb);

  1567.     if ((v->s.pict_type == FF_I_TYPE || v->s.pict_type == FF_BI_TYPE) && v->dquant) {

  1568.         av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");

  1569.         vop_dquant_decoding(v);

  1570.     }

  1571. 

  1572.     v->bi_type = 0;

  1573.     if(v->s.pict_type == FF_BI_TYPE) {

  1574.         v->s.pict_type = FF_B_TYPE;

  1575.         v->bi_type = 1;

  1576.     }

  1577.     return 0;

  1578. }

  1579. 

  1580. /***********************************************************************/

  1581. /**

  1582.  * @defgroup block VC-1 Block-level functions

  1583.  * @see 7.1.4, p91 and 8.1.1.7, p(1)04

  1584.  * @{

  1585.  */

  1586. 

  1587. /**

  1588.  * @def GET_MQUANT

  1589.  * @brief Get macroblock-level quantizer scale

  1590.  */

  1591. #define GET_MQUANT()                                           \

  1592.   if (v->dquantfrm)                                            \

  1593.   {                                                            \

  1594.     int edges = 0;                                             \

  1595.     if (v->dqprofile == DQPROFILE_ALL_MBS)                     \

  1596.     {                                                          \

  1597.       if (v->dqbilevel)                                        \

  1598.       {                                                        \

  1599.         mquant = (get_bits1(gb)) ? v->altpq : v->pq;           \

  1600.       }                                                        \

  1601.       else                                                     \

  1602.       {                                                        \

  1603.         mqdiff = get_bits(gb, 3);                              \

  1604.         if (mqdiff != 7) mquant = v->pq + mqdiff;              \

  1605.         else mquant = get_bits(gb, 5);                         \

  1606.       }                                                        \

  1607.     }                                                          \

  1608.     if(v->dqprofile == DQPROFILE_SINGLE_EDGE)                  \

  1609.         edges = 1 << v->dqsbedge;                              \

  1610.     else if(v->dqprofile == DQPROFILE_DOUBLE_EDGES)            \

  1611.         edges = (3 << v->dqsbedge) % 15;                       \

  1612.     else if(v->dqprofile == DQPROFILE_FOUR_EDGES)              \

  1613.         edges = 15;                                            \

  1614.     if((edges&1) && !s->mb_x)                                  \

  1615.         mquant = v->altpq;                                     \

  1616.     if((edges&2) && s->first_slice_line)                       \

  1617.         mquant = v->altpq;                                     \

  1618.     if((edges&4) && s->mb_x == (s->mb_width - 1))              \

  1619.         mquant = v->altpq;                                     \

  1620.     if((edges&8) && s->mb_y == (s->mb_height - 1))             \

  1621.         mquant = v->altpq;                                     \

  1622.   }

  1623. 

  1624. /**

  1625.  * @def GET_MVDATA(_dmv_x, _dmv_y)

  1626.  * @brief Get MV differentials

  1627.  * @see MVDATA decoding from 8.3.5.2, p(1)20

  1628.  * @param _dmv_x Horizontal differential for decoded MV

  1629.  * @param _dmv_y Vertical differential for decoded MV

  1630.  */

  1631. #define GET_MVDATA(_dmv_x, _dmv_y)                                  \

  1632.   index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table,\

  1633.                        VC1_MV_DIFF_VLC_BITS, 2);                    \

  1634.   if (index > 36)                                                   \

  1635.   {                                                                 \

  1636.     mb_has_coeffs = 1;                                              \

  1637.     index -= 37;                                                    \

  1638.   }                                                                 \

  1639.   else mb_has_coeffs = 0;                                           \

  1640.   s->mb_intra = 0;                                                  \

  1641.   if (!index) { _dmv_x = _dmv_y = 0; }                              \

  1642.   else if (index == 35)                                             \

  1643.   {                                                                 \

  1644.     _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample);          \

  1645.     _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample);          \

  1646.   }                                                                 \

  1647.   else if (index == 36)                                             \

  1648.   {                                                                 \

  1649.     _dmv_x = 0;                                                     \

  1650.     _dmv_y = 0;                                                     \

  1651.     s->mb_intra = 1;                                                \

  1652.   }                                                                 \

  1653.   else                                                              \

  1654.   {                                                                 \

  1655.     index1 = index%6;                                               \

  1656.     if (!s->quarter_sample && index1 == 5) val = 1;                 \

  1657.     else                                   val = 0;                 \

  1658.     if(size_table[index1] - val > 0)                                \

  1659.         val = get_bits(gb, size_table[index1] - val);               \

  1660.     else                                   val = 0;                 \

  1661.     sign = 0 - (val&1);                                             \

  1662.     _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \

  1663.                                                                     \

  1664.     index1 = index/6;                                               \

  1665.     if (!s->quarter_sample && index1 == 5) val = 1;                 \

  1666.     else                                   val = 0;                 \

  1667.     if(size_table[index1] - val > 0)                                \

  1668.         val = get_bits(gb, size_table[index1] - val);               \

  1669.     else                                   val = 0;                 \

  1670.     sign = 0 - (val&1);                                             \

  1671.     _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \

  1672.   }

  1673. 

  1674. /** Predict and set motion vector

  1675.  */

  1676. static inline void vc1_pred_mv(MpegEncContext *s, int n, int dmv_x, int dmv_y, int mv1, int r_x, int r_y, uint8_t* is_intra)

  1677. {

  1678.     int xy, wrap, off = 0;

  1679.     int16_t *A, *B, *C;

  1680.     int px, py;

  1681.     int sum;

  1682. 

  1683.     /* scale MV difference to be quad-pel */

  1684.     dmv_x <<= 1 - s->quarter_sample;

  1685.     dmv_y <<= 1 - s->quarter_sample;

  1686. 

  1687.     wrap = s->b8_stride;

  1688.     xy = s->block_index[n];

  1689. 

  1690.     if(s->mb_intra){

  1691.         s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;

  1692.         s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;

  1693.         s->current_picture.motion_val[1][xy][0] = 0;

  1694.         s->current_picture.motion_val[1][xy][1] = 0;

  1695.         if(mv1) { /* duplicate motion data for 1-MV block */

  1696.             s->current_picture.motion_val[0][xy + 1][0] = 0;

  1697.             s->current_picture.motion_val[0][xy + 1][1] = 0;

  1698.             s->current_picture.motion_val[0][xy + wrap][0] = 0;

  1699.             s->current_picture.motion_val[0][xy + wrap][1] = 0;

  1700.             s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;

  1701.             s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;

  1702.             s->current_picture.motion_val[1][xy + 1][0] = 0;

  1703.             s->current_picture.motion_val[1][xy + 1][1] = 0;

  1704.             s->current_picture.motion_val[1][xy + wrap][0] = 0;

  1705.             s->current_picture.motion_val[1][xy + wrap][1] = 0;

  1706.             s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;

  1707.             s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;

  1708.         }

  1709.         return;

  1710.     }

  1711. 

  1712.     C = s->current_picture.motion_val[0][xy - 1];

  1713.     A = s->current_picture.motion_val[0][xy - wrap];

  1714.     if(mv1)

  1715.         off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;

  1716.     else {

  1717.         //in 4-MV mode different blocks have different B predictor position

  1718.         switch(n){

  1719.         case 0:

  1720.             off = (s->mb_x > 0) ? -1 : 1;

  1721.             break;

  1722.         case 1:

  1723.             off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;

  1724.             break;

  1725.         case 2:

  1726.             off = 1;

  1727.             break;

  1728.         case 3:

  1729.             off = -1;

  1730.         }

  1731.     }

  1732.     B = s->current_picture.motion_val[0][xy - wrap + off];

  1733. 

  1734.     if(!s->first_slice_line || (n==2 || n==3)) { // predictor A is not out of bounds

  1735.         if(s->mb_width == 1) {

  1736.             px = A[0];

  1737.             py = A[1];

  1738.         } else {

  1739.             px = mid_pred(A[0], B[0], C[0]);

  1740.             py = mid_pred(A[1], B[1], C[1]);

  1741.         }

  1742.     } else if(s->mb_x || (n==1 || n==3)) { // predictor C is not out of bounds

  1743.         px = C[0];

  1744.         py = C[1];

  1745.     } else {

  1746.         px = py = 0;

  1747.     }

  1748.     /* Pullback MV as specified in 8.3.5.3.4 */

  1749.     {

  1750.         int qx, qy, X, Y;

  1751.         qx = (s->mb_x << 6) + ((n==1 || n==3) ? 32 : 0);

  1752.         qy = (s->mb_y << 6) + ((n==2 || n==3) ? 32 : 0);

  1753.         X = (s->mb_width << 6) - 4;

  1754.         Y = (s->mb_height << 6) - 4;

  1755.         if(mv1) {

  1756.             if(qx + px < -60) px = -60 - qx;

  1757.             if(qy + py < -60) py = -60 - qy;

  1758.         } else {

  1759.             if(qx + px < -28) px = -28 - qx;

  1760.             if(qy + py < -28) py = -28 - qy;

  1761.         }

  1762.         if(qx + px > X) px = X - qx;

  1763.         if(qy + py > Y) py = Y - qy;

  1764.     }

  1765.     /* Calculate hybrid prediction as specified in 8.3.5.3.5 */

  1766.     if((!s->first_slice_line || (n==2 || n==3)) && (s->mb_x || (n==1 || n==3))) {

  1767.         if(is_intra[xy - wrap])

  1768.             sum = FFABS(px) + FFABS(py);

  1769.         else

  1770.             sum = FFABS(px - A[0]) + FFABS(py - A[1]);

  1771.         if(sum > 32) {

  1772.             if(get_bits1(&s->gb)) {

  1773.                 px = A[0];

  1774.                 py = A[1];

  1775.             } else {

  1776.                 px = C[0];

  1777.                 py = C[1];

  1778.             }

  1779.         } else {

  1780.             if(is_intra[xy - 1])

  1781.                 sum = FFABS(px) + FFABS(py);

  1782.             else

  1783.                 sum = FFABS(px - C[0]) + FFABS(py - C[1]);

  1784.             if(sum > 32) {

  1785.                 if(get_bits1(&s->gb)) {

  1786.                     px = A[0];

  1787.                     py = A[1];

  1788.                 } else {

  1789.                     px = C[0];

  1790.                     py = C[1];

  1791.                 }

  1792.             }

  1793.         }

  1794.     }

  1795.     /* store MV using signed modulus of MV range defined in 4.11 */

  1796.     s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;

  1797.     s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;

  1798.     if(mv1) { /* duplicate motion data for 1-MV block */

  1799.         s->current_picture.motion_val[0][xy + 1][0] = s->current_picture.motion_val[0][xy][0];

  1800.         s->current_picture.motion_val[0][xy + 1][1] = s->current_picture.motion_val[0][xy][1];

  1801.         s->current_picture.motion_val[0][xy + wrap][0] = s->current_picture.motion_val[0][xy][0];

  1802.         s->current_picture.motion_val[0][xy + wrap][1] = s->current_picture.motion_val[0][xy][1];

  1803.         s->current_picture.motion_val[0][xy + wrap + 1][0] = s->current_picture.motion_val[0][xy][0];

  1804.         s->current_picture.motion_val[0][xy + wrap + 1][1] = s->current_picture.motion_val[0][xy][1];

  1805.     }

  1806. }

  1807. 

  1808. /** Motion compensation for direct or interpolated blocks in B-frames

  1809.  */

  1810. static void vc1_interp_mc(VC1Context *v)

  1811. {

  1812.     MpegEncContext *s = &v->s;

  1813.     DSPContext *dsp = &v->s.dsp;

  1814.     uint8_t *srcY, *srcU, *srcV;

  1815.     int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;

  1816. 

  1817.     if(!v->s.next_picture.data[0])return;

  1818. 

  1819.     mx = s->mv[1][0][0];

  1820.     my = s->mv[1][0][1];

  1821.     uvmx = (mx + ((mx & 3) == 3)) >> 1;

  1822.     uvmy = (my + ((my & 3) == 3)) >> 1;

  1823.     if(v->fastuvmc) {

  1824.         uvmx = uvmx + ((uvmx<0)?-(uvmx&1):(uvmx&1));

  1825.         uvmy = uvmy + ((uvmy<0)?-(uvmy&1):(uvmy&1));

  1826.     }

  1827.     srcY = s->next_picture.data[0];

  1828.     srcU = s->next_picture.data[1];

  1829.     srcV = s->next_picture.data[2];

  1830. 

  1831.     src_x = s->mb_x * 16 + (mx >> 2);

  1832.     src_y = s->mb_y * 16 + (my >> 2);

  1833.     uvsrc_x = s->mb_x * 8 + (uvmx >> 2);

  1834.     uvsrc_y = s->mb_y * 8 + (uvmy >> 2);

  1835. 

  1836.     if(v->profile != PROFILE_ADVANCED){

  1837.         src_x   = av_clip(  src_x, -16, s->mb_width  * 16);

  1838.         src_y   = av_clip(  src_y, -16, s->mb_height * 16);

  1839.         uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);

  1840.         uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);

  1841.     }else{

  1842.         src_x   = av_clip(  src_x, -17, s->avctx->coded_width);

  1843.         src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);

  1844.         uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);

  1845.         uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);

  1846.     }

  1847. 

  1848.     srcY += src_y * s->linesize + src_x;

  1849.     srcU += uvsrc_y * s->uvlinesize + uvsrc_x;

  1850.     srcV += uvsrc_y * s->uvlinesize + uvsrc_x;

  1851. 

  1852.     /* for grayscale we should not try to read from unknown area */

  1853.     if(s->flags & CODEC_FLAG_GRAY) {

  1854.         srcU = s->edge_emu_buffer + 18 * s->linesize;

  1855.         srcV = s->edge_emu_buffer + 18 * s->linesize;

  1856.     }

  1857. 

  1858.     if(v->rangeredfrm

  1859.        || (unsigned)src_x > s->h_edge_pos - (mx&3) - 16

  1860.        || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){

  1861.         uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;

  1862. 

  1863.         srcY -= s->mspel * (1 + s->linesize);

  1864.         ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,

  1865.                             src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);

  1866.         srcY = s->edge_emu_buffer;

  1867.         ff_emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8+1, 8+1,

  1868.                             uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);

  1869.         ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,

  1870.                             uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);

  1871.         srcU = uvbuf;

  1872.         srcV = uvbuf + 16;

  1873.         /* if we deal with range reduction we need to scale source blocks */

  1874.         if(v->rangeredfrm) {

  1875.             int i, j;

  1876.             uint8_t *src, *src2;

  1877. 

  1878.             src = srcY;

  1879.             for(j = 0; j < 17 + s->mspel*2; j++) {

  1880.                 for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;

  1881.                 src += s->linesize;

  1882.             }

  1883.             src = srcU; src2 = srcV;

  1884.             for(j = 0; j < 9; j++) {

  1885.                 for(i = 0; i < 9; i++) {

  1886.                     src[i] = ((src[i] - 128) >> 1) + 128;

  1887.                     src2[i] = ((src2[i] - 128) >> 1) + 128;

  1888.                 }

  1889.                 src += s->uvlinesize;

  1890.                 src2 += s->uvlinesize;

  1891.             }

  1892.         }

  1893.         srcY += s->mspel * (1 + s->linesize);

  1894.     }

  1895. 

  1896.     mx >>= 1;

  1897.     my >>= 1;

  1898.     dxy = ((my & 1) << 1) | (mx & 1);

  1899. 

  1900.     dsp->avg_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);

  1901. 

  1902.     if(s->flags & CODEC_FLAG_GRAY) return;

  1903.     /* Chroma MC always uses qpel blilinear */

  1904.     uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);

  1905.     uvmx = (uvmx&3)<<1;

  1906.     uvmy = (uvmy&3)<<1;

  1907.     dsp->avg_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);

  1908.     dsp->avg_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);

  1909. }

  1910. 

  1911. static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)

  1912. {

  1913.     int n = bfrac;

  1914. 

  1915. #if B_FRACTION_DEN==256

  1916.     if(inv)

  1917.         n -= 256;

  1918.     if(!qs)

  1919.         return 2 * ((value * n + 255) >> 9);

  1920.     return (value * n + 128) >> 8;

  1921. #else

  1922.     if(inv)

  1923.         n -= B_FRACTION_DEN;

  1924.     if(!qs)

  1925.         return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));

  1926.     return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;

  1927. #endif

  1928. }

  1929. 

  1930. /** Reconstruct motion vector for B-frame and do motion compensation

  1931.  */

  1932. static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mode)

  1933. {

  1934.     if(v->use_ic) {

  1935.         v->mv_mode2 = v->mv_mode;

  1936.         v->mv_mode = MV_PMODE_INTENSITY_COMP;

  1937.     }

  1938.     if(direct) {

  1939.         vc1_mc_1mv(v, 0);

  1940.         vc1_interp_mc(v);

  1941.         if(v->use_ic) v->mv_mode = v->mv_mode2;

  1942.         return;

  1943.     }

  1944.     if(mode == BMV_TYPE_INTERPOLATED) {

  1945.         vc1_mc_1mv(v, 0);

  1946.         vc1_interp_mc(v);

  1947.         if(v->use_ic) v->mv_mode = v->mv_mode2;

  1948.         return;

  1949.     }

  1950. 

  1951.     if(v->use_ic && (mode == BMV_TYPE_BACKWARD)) v->mv_mode = v->mv_mode2;

  1952.     vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));

  1953.     if(v->use_ic) v->mv_mode = v->mv_mode2;

  1954. }

  1955. 

  1956. static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mvtype)

  1957. {

  1958.     MpegEncContext *s = &v->s;

  1959.     int xy, wrap, off = 0;

  1960.     int16_t *A, *B, *C;

  1961.     int px, py;

  1962.     int sum;

  1963.     int r_x, r_y;

  1964.     const uint8_t *is_intra = v->mb_type[0];

  1965. 

  1966.     r_x = v->range_x;

  1967.     r_y = v->range_y;

  1968.     /* scale MV difference to be quad-pel */

  1969.     dmv_x[0] <<= 1 - s->quarter_sample;

  1970.     dmv_y[0] <<= 1 - s->quarter_sample;

  1971.     dmv_x[1] <<= 1 - s->quarter_sample;

  1972.     dmv_y[1] <<= 1 - s->quarter_sample;

  1973. 

  1974.     wrap = s->b8_stride;

  1975.     xy = s->block_index[0];

  1976. 

  1977.     if(s->mb_intra) {

  1978.         s->current_picture.motion_val[0][xy][0] =

  1979.         s->current_picture.motion_val[0][xy][1] =

  1980.         s->current_picture.motion_val[1][xy][0] =

  1981.         s->current_picture.motion_val[1][xy][1] = 0;

  1982.         return;

  1983.     }

  1984.     s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);

  1985.     s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);

  1986.     s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);

  1987.     s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);

  1988. 

  1989.     /* Pullback predicted motion vectors as specified in 8.4.5.4 */

  1990.     s->mv[0][0][0] = av_clip(s->mv[0][0][0], -60 - (s->mb_x << 6), (s->mb_width  << 6) - 4 - (s->mb_x << 6));

  1991.     s->mv[0][0][1] = av_clip(s->mv[0][0][1], -60 - (s->mb_y << 6), (s->mb_height << 6) - 4 - (s->mb_y << 6));

  1992.     s->mv[1][0][0] = av_clip(s->mv[1][0][0], -60 - (s->mb_x << 6), (s->mb_width  << 6) - 4 - (s->mb_x << 6));

  1993.     s->mv[1][0][1] = av_clip(s->mv[1][0][1], -60 - (s->mb_y << 6), (s->mb_height << 6) - 4 - (s->mb_y << 6));

  1994.     if(direct) {

  1995.         s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];

  1996.         s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];

  1997.         s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];

  1998.         s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];

  1999.         return;

  2000.     }

  2001. 

  2002.     if((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {

  2003.         C = s->current_picture.motion_val[0][xy - 2];

  2004.         A = s->current_picture.motion_val[0][xy - wrap*2];

  2005.         off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  2006.         B = s->current_picture.motion_val[0][xy - wrap*2 + off];

  2007. 

  2008.         if(!s->mb_x) C[0] = C[1] = 0;

  2009.         if(!s->first_slice_line) { // predictor A is not out of bounds

  2010.             if(s->mb_width == 1) {

  2011.                 px = A[0];

  2012.                 py = A[1];

  2013.             } else {

  2014.                 px = mid_pred(A[0], B[0], C[0]);

  2015.                 py = mid_pred(A[1], B[1], C[1]);

  2016.             }

  2017.         } else if(s->mb_x) { // predictor C is not out of bounds

  2018.             px = C[0];

  2019.             py = C[1];

  2020.         } else {

  2021.             px = py = 0;

  2022.         }

  2023.         /* Pullback MV as specified in 8.3.5.3.4 */

  2024.         {

  2025.             int qx, qy, X, Y;

  2026.             if(v->profile < PROFILE_ADVANCED) {

  2027.                 qx = (s->mb_x << 5);

  2028.                 qy = (s->mb_y << 5);

  2029.                 X = (s->mb_width << 5) - 4;

  2030.                 Y = (s->mb_height << 5) - 4;

  2031.                 if(qx + px < -28) px = -28 - qx;

  2032.                 if(qy + py < -28) py = -28 - qy;

  2033.                 if(qx + px > X) px = X - qx;

  2034.                 if(qy + py > Y) py = Y - qy;

  2035.             } else {

  2036.                 qx = (s->mb_x << 6);

  2037.                 qy = (s->mb_y << 6);

  2038.                 X = (s->mb_width << 6) - 4;

  2039.                 Y = (s->mb_height << 6) - 4;

  2040.                 if(qx + px < -60) px = -60 - qx;

  2041.                 if(qy + py < -60) py = -60 - qy;

  2042.                 if(qx + px > X) px = X - qx;

  2043.                 if(qy + py > Y) py = Y - qy;

  2044.             }

  2045.         }

  2046.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 */

  2047.         if(0 && !s->first_slice_line && s->mb_x) {

  2048.             if(is_intra[xy - wrap])

  2049.                 sum = FFABS(px) + FFABS(py);

  2050.             else

  2051.                 sum = FFABS(px - A[0]) + FFABS(py - A[1]);

  2052.             if(sum > 32) {

  2053.                 if(get_bits1(&s->gb)) {

  2054.                     px = A[0];

  2055.                     py = A[1];

  2056.                 } else {

  2057.                     px = C[0];

  2058.                     py = C[1];

  2059.                 }

  2060.             } else {

  2061.                 if(is_intra[xy - 2])

  2062.                     sum = FFABS(px) + FFABS(py);

  2063.                 else

  2064.                     sum = FFABS(px - C[0]) + FFABS(py - C[1]);

  2065.                 if(sum > 32) {

  2066.                     if(get_bits1(&s->gb)) {

  2067.                         px = A[0];

  2068.                         py = A[1];

  2069.                     } else {

  2070.                         px = C[0];

  2071.                         py = C[1];

  2072.                     }

  2073.                 }

  2074.             }

  2075.         }

  2076.         /* store MV using signed modulus of MV range defined in 4.11 */

  2077.         s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;

  2078.         s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;

  2079.     }

  2080.     if((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {

  2081.         C = s->current_picture.motion_val[1][xy - 2];

  2082.         A = s->current_picture.motion_val[1][xy - wrap*2];

  2083.         off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  2084.         B = s->current_picture.motion_val[1][xy - wrap*2 + off];

  2085. 

  2086.         if(!s->mb_x) C[0] = C[1] = 0;

  2087.         if(!s->first_slice_line) { // predictor A is not out of bounds

  2088.             if(s->mb_width == 1) {

  2089.                 px = A[0];

  2090.                 py = A[1];

  2091.             } else {

  2092.                 px = mid_pred(A[0], B[0], C[0]);

  2093.                 py = mid_pred(A[1], B[1], C[1]);

  2094.             }

  2095.         } else if(s->mb_x) { // predictor C is not out of bounds

  2096.             px = C[0];

  2097.             py = C[1];

  2098.         } else {

  2099.             px = py = 0;

  2100.         }

  2101.         /* Pullback MV as specified in 8.3.5.3.4 */

  2102.         {

  2103.             int qx, qy, X, Y;

  2104.             if(v->profile < PROFILE_ADVANCED) {

  2105.                 qx = (s->mb_x << 5);

  2106.                 qy = (s->mb_y << 5);

  2107.                 X = (s->mb_width << 5) - 4;

  2108.                 Y = (s->mb_height << 5) - 4;

  2109.                 if(qx + px < -28) px = -28 - qx;

  2110.                 if(qy + py < -28) py = -28 - qy;

  2111.                 if(qx + px > X) px = X - qx;

  2112.                 if(qy + py > Y) py = Y - qy;

  2113.             } else {

  2114.                 qx = (s->mb_x << 6);

  2115.                 qy = (s->mb_y << 6);

  2116.                 X = (s->mb_width << 6) - 4;

  2117.                 Y = (s->mb_height << 6) - 4;

  2118.                 if(qx + px < -60) px = -60 - qx;

  2119.                 if(qy + py < -60) py = -60 - qy;

  2120.                 if(qx + px > X) px = X - qx;

  2121.                 if(qy + py > Y) py = Y - qy;

  2122.             }

  2123.         }

  2124.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 */

  2125.         if(0 && !s->first_slice_line && s->mb_x) {

  2126.             if(is_intra[xy - wrap])

  2127.                 sum = FFABS(px) + FFABS(py);

  2128.             else

  2129.                 sum = FFABS(px - A[0]) + FFABS(py - A[1]);

  2130.             if(sum > 32) {

  2131.                 if(get_bits1(&s->gb)) {

  2132.                     px = A[0];

  2133.                     py = A[1];

  2134.                 } else {

  2135.                     px = C[0];

  2136.                     py = C[1];

  2137.                 }

  2138.             } else {

  2139.                 if(is_intra[xy - 2])

  2140.                     sum = FFABS(px) + FFABS(py);

  2141.                 else

  2142.                     sum = FFABS(px - C[0]) + FFABS(py - C[1]);

  2143.                 if(sum > 32) {

  2144.                     if(get_bits1(&s->gb)) {

  2145.                         px = A[0];

  2146.                         py = A[1];

  2147.                     } else {

  2148.                         px = C[0];

  2149.                         py = C[1];

  2150.                     }

  2151.                 }

  2152.             }

  2153.         }

  2154.         /* store MV using signed modulus of MV range defined in 4.11 */

  2155. 

  2156.         s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;

  2157.         s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;

  2158.     }

  2159.     s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];

  2160.     s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];

  2161.     s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];

  2162.     s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];

  2163. }

  2164. 

  2165. /** Get predicted DC value for I-frames only

  2166.  * prediction dir: left=0, top=1

  2167.  * @param s MpegEncContext

  2168.  * @param[in] n block index in the current MB

  2169.  * @param dc_val_ptr Pointer to DC predictor

  2170.  * @param dir_ptr Prediction direction for use in AC prediction

  2171.  */

  2172. static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,

  2173.                               int16_t **dc_val_ptr, int *dir_ptr)

  2174. {

  2175.     int a, b, c, wrap, pred, scale;

  2176.     int16_t *dc_val;

  2177.     static const uint16_t dcpred[32] = {

  2178.     -1, 1024,  512,  341,  256,  205,  171,  146,  128,

  2179.          114,  102,   93,   85,   79,   73,   68,   64,

  2180.           60,   57,   54,   51,   49,   47,   45,   43,

  2181.           41,   39,   38,   37,   35,   34,   33

  2182.     };

  2183. 

  2184.     /* find prediction - wmv3_dc_scale always used here in fact */

  2185.     if (n < 4)     scale = s->y_dc_scale;

  2186.     else           scale = s->c_dc_scale;

  2187. 

  2188.     wrap = s->block_wrap[n];

  2189.     dc_val= s->dc_val[0] + s->block_index[n];

  2190. 

  2191.     /* B A

  2192.      * C X

  2193.      */

  2194.     c = dc_val[ - 1];

  2195.     b = dc_val[ - 1 - wrap];

  2196.     a = dc_val[ - wrap];

  2197. 

  2198.     if (pq < 9 || !overlap)

  2199.     {

  2200.         /* Set outer values */

  2201.         if (s->first_slice_line && (n!=2 && n!=3)) b=a=dcpred[scale];

  2202.         if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=dcpred[scale];

  2203.     }

  2204.     else

  2205.     {

  2206.         /* Set outer values */

  2207.         if (s->first_slice_line && (n!=2 && n!=3)) b=a=0;

  2208.         if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=0;

  2209.     }

  2210. 

  2211.     if (abs(a - b) <= abs(b - c)) {

  2212.         pred = c;

  2213.         *dir_ptr = 1;//left

  2214.     } else {

  2215.         pred = a;

  2216.         *dir_ptr = 0;//top

  2217.     }

  2218. 

  2219.     /* update predictor */

  2220.     *dc_val_ptr = &dc_val[0];

  2221.     return pred;

  2222. }

  2223. 

  2224. 

  2225. /** Get predicted DC value

  2226.  * prediction dir: left=0, top=1

  2227.  * @param s MpegEncContext

  2228.  * @param[in] n block index in the current MB

  2229.  * @param dc_val_ptr Pointer to DC predictor

  2230.  * @param dir_ptr Prediction direction for use in AC prediction

  2231.  */

  2232. static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,

  2233.                               int a_avail, int c_avail,

  2234.                               int16_t **dc_val_ptr, int *dir_ptr)

  2235. {

  2236.     int a, b, c, wrap, pred, scale;

  2237.     int16_t *dc_val;

  2238.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2239.     int q1, q2 = 0;

  2240. 

  2241.     /* find prediction - wmv3_dc_scale always used here in fact */

  2242.     if (n < 4)     scale = s->y_dc_scale;

  2243.     else           scale = s->c_dc_scale;

  2244. 

  2245.     wrap = s->block_wrap[n];

  2246.     dc_val= s->dc_val[0] + s->block_index[n];

  2247. 

  2248.     /* B A

  2249.      * C X

  2250.      */

  2251.     c = dc_val[ - 1];

  2252.     b = dc_val[ - 1 - wrap];

  2253.     a = dc_val[ - wrap];

  2254.     /* scale predictors if needed */

  2255.     q1 = s->current_picture.qscale_table[mb_pos];

  2256.     if(c_avail && (n!= 1 && n!=3)) {

  2257.         q2 = s->current_picture.qscale_table[mb_pos - 1];

  2258.         if(q2 && q2 != q1)

  2259.             c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;

  2260.     }

  2261.     if(a_avail && (n!= 2 && n!=3)) {

  2262.         q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];

  2263.         if(q2 && q2 != q1)

  2264.             a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;

  2265.     }

  2266.     if(a_avail && c_avail && (n!=3)) {

  2267.         int off = mb_pos;

  2268.         if(n != 1) off--;

  2269.         if(n != 2) off -= s->mb_stride;

  2270.         q2 = s->current_picture.qscale_table[off];

  2271.         if(q2 && q2 != q1)

  2272.             b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;

  2273.     }

  2274. 

  2275.     if(a_avail && c_avail) {

  2276.         if(abs(a - b) <= abs(b - c)) {

  2277.             pred = c;

  2278.             *dir_ptr = 1;//left

  2279.         } else {

  2280.             pred = a;

  2281.             *dir_ptr = 0;//top

  2282.         }

  2283.     } else if(a_avail) {

  2284.         pred = a;

  2285.         *dir_ptr = 0;//top

  2286.     } else if(c_avail) {

  2287.         pred = c;

  2288.         *dir_ptr = 1;//left

  2289.     } else {

  2290.         pred = 0;

  2291.         *dir_ptr = 1;//left

  2292.     }

  2293. 

  2294.     /* update predictor */

  2295.     *dc_val_ptr = &dc_val[0];

  2296.     return pred;

  2297. }

  2298. 

  2299. 

  2300. /**

  2301.  * @defgroup std_mb VC1 Macroblock-level functions in Simple/Main Profiles

  2302.  * @see 7.1.4, p91 and 8.1.1.7, p(1)04

  2303.  * @{

  2304.  */

  2305. 

  2306. static inline int vc1_coded_block_pred(MpegEncContext * s, int n, uint8_t **coded_block_ptr)

  2307. {

  2308.     int xy, wrap, pred, a, b, c;

  2309. 

  2310.     xy = s->block_index[n];

  2311.     wrap = s->b8_stride;

  2312. 

  2313.     /* B C

  2314.      * A X

  2315.      */

  2316.     a = s->coded_block[xy - 1       ];

  2317.     b = s->coded_block[xy - 1 - wrap];

  2318.     c = s->coded_block[xy     - wrap];

  2319. 

  2320.     if (b == c) {

  2321.         pred = a;

  2322.     } else {

  2323.         pred = c;

  2324.     }

  2325. 

  2326.     /* store value */

  2327.     *coded_block_ptr = &s->coded_block[xy];

  2328. 

  2329.     return pred;

  2330. }

  2331. 

  2332. /**

  2333.  * Decode one AC coefficient

  2334.  * @param v The VC1 context

  2335.  * @param last Last coefficient

  2336.  * @param skip How much zero coefficients to skip

  2337.  * @param value Decoded AC coefficient value

  2338.  * @see 8.1.3.4

  2339.  */

  2340. static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, int codingset)

  2341. {

  2342.     GetBitContext *gb = &v->s.gb;

  2343.     int index, escape, run = 0, level = 0, lst = 0;

  2344. 

  2345.     index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);

  2346.     if (index != vc1_ac_sizes[codingset] - 1) {

  2347.         run = vc1_index_decode_table[codingset][index][0];

  2348.         level = vc1_index_decode_table[codingset][index][1];

  2349.         lst = index >= vc1_last_decode_table[codingset];

  2350.         if(get_bits1(gb))

  2351.             level = -level;

  2352.     } else {

  2353.         escape = decode210(gb);

  2354.         if (escape != 2) {

  2355.             index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);

  2356.             run = vc1_index_decode_table[codingset][index][0];

  2357.             level = vc1_index_decode_table[codingset][index][1];

  2358.             lst = index >= vc1_last_decode_table[codingset];

  2359.             if(escape == 0) {

  2360.                 if(lst)

  2361.                     level += vc1_last_delta_level_table[codingset][run];

  2362.                 else

  2363.                     level += vc1_delta_level_table[codingset][run];

  2364.             } else {

  2365.                 if(lst)

  2366.                     run += vc1_last_delta_run_table[codingset][level] + 1;

  2367.                 else

  2368.                     run += vc1_delta_run_table[codingset][level] + 1;

  2369.             }

  2370.             if(get_bits1(gb))

  2371.                 level = -level;

  2372.         } else {

  2373.             int sign;

  2374.             lst = get_bits1(gb);

  2375.             if(v->s.esc3_level_length == 0) {

  2376.                 if(v->pq < 8 || v->dquantfrm) { // table 59

  2377.                     v->s.esc3_level_length = get_bits(gb, 3);

  2378.                     if(!v->s.esc3_level_length)

  2379.                         v->s.esc3_level_length = get_bits(gb, 2) + 8;

  2380.                 } else { //table 60

  2381.                     v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;

  2382.                 }

  2383.                 v->s.esc3_run_length = 3 + get_bits(gb, 2);

  2384.             }

  2385.             run = get_bits(gb, v->s.esc3_run_length);

  2386.             sign = get_bits1(gb);

  2387.             level = get_bits(gb, v->s.esc3_level_length);

  2388.             if(sign)

  2389.                 level = -level;

  2390.         }

  2391.     }

  2392. 

  2393.     *last = lst;

  2394.     *skip = run;

  2395.     *value = level;

  2396. }

  2397. 

  2398. /** Decode intra block in intra frames - should be faster than decode_intra_block

  2399.  * @param v VC1Context

  2400.  * @param block block to decode

  2401.  * @param coded are AC coeffs present or not

  2402.  * @param codingset set of VLC to decode data

  2403.  */

  2404. static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset)

  2405. {

  2406.     GetBitContext *gb = &v->s.gb;

  2407.     MpegEncContext *s = &v->s;

  2408.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2409.     int run_diff, i;

  2410.     int16_t *dc_val;

  2411.     int16_t *ac_val, *ac_val2;

  2412.     int dcdiff;

  2413. 

  2414.     /* Get DC differential */

  2415.     if (n < 4) {

  2416.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2417.     } else {

  2418.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2419.     }

  2420.     if (dcdiff < 0){

  2421.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2422.         return -1;

  2423.     }

  2424.     if (dcdiff)

  2425.     {

  2426.         if (dcdiff == 119 /* ESC index value */)

  2427.         {

  2428.             /* TODO: Optimize */

  2429.             if (v->pq == 1) dcdiff = get_bits(gb, 10);

  2430.             else if (v->pq == 2) dcdiff = get_bits(gb, 9);

  2431.             else dcdiff = get_bits(gb, 8);

  2432.         }

  2433.         else

  2434.         {

  2435.             if (v->pq == 1)

  2436.                 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;

  2437.             else if (v->pq == 2)

  2438.                 dcdiff = (dcdiff<<1) + get_bits1(gb)   - 1;

  2439.         }

  2440.         if (get_bits1(gb))

  2441.             dcdiff = -dcdiff;

  2442.     }

  2443. 

  2444.     /* Prediction */

  2445.     dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);

  2446.     *dc_val = dcdiff;

  2447. 

  2448.     /* Store the quantized DC coeff, used for prediction */

  2449.     if (n < 4) {

  2450.         block[0] = dcdiff * s->y_dc_scale;

  2451.     } else {

  2452.         block[0] = dcdiff * s->c_dc_scale;

  2453.     }

  2454.     /* Skip ? */

  2455.     run_diff = 0;

  2456.     i = 0;

  2457.     if (!coded) {

  2458.         goto not_coded;

  2459.     }

  2460. 

  2461.     //AC Decoding

  2462.     i = 1;

  2463. 

  2464.     {

  2465.         int last = 0, skip, value;

  2466.         const int8_t *zz_table;

  2467.         int scale;

  2468.         int k;

  2469. 

  2470.         scale = v->pq * 2 + v->halfpq;

  2471. 

  2472.         if(v->s.ac_pred) {

  2473.             if(!dc_pred_dir)

  2474.                 zz_table = wmv1_scantable[2];

  2475.             else

  2476.                 zz_table = wmv1_scantable[3];

  2477.         } else

  2478.             zz_table = wmv1_scantable[1];

  2479. 

  2480.         ac_val = s->ac_val[0][0] + s->block_index[n] * 16;

  2481.         ac_val2 = ac_val;

  2482.         if(dc_pred_dir) //left

  2483.             ac_val -= 16;

  2484.         else //top

  2485.             ac_val -= 16 * s->block_wrap[n];

  2486. 

  2487.         while (!last) {

  2488.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2489.             i += skip;

  2490.             if(i > 63)

  2491.                 break;

  2492.             block[zz_table[i++]] = value;

  2493.         }

  2494. 

  2495.         /* apply AC prediction if needed */

  2496.         if(s->ac_pred) {

  2497.             if(dc_pred_dir) { //left

  2498.                 for(k = 1; k < 8; k++)

  2499.                     block[k << 3] += ac_val[k];

  2500.             } else { //top

  2501.                 for(k = 1; k < 8; k++)

  2502.                     block[k] += ac_val[k + 8];

  2503.             }

  2504.         }

  2505.         /* save AC coeffs for further prediction */

  2506.         for(k = 1; k < 8; k++) {

  2507.             ac_val2[k] = block[k << 3];

  2508.             ac_val2[k + 8] = block[k];

  2509.         }

  2510. 

  2511.         /* scale AC coeffs */

  2512.         for(k = 1; k < 64; k++)

  2513.             if(block[k]) {

  2514.                 block[k] *= scale;

  2515.                 if(!v->pquantizer)

  2516.                     block[k] += (block[k] < 0) ? -v->pq : v->pq;

  2517.             }

  2518. 

  2519.         if(s->ac_pred) i = 63;

  2520.     }

  2521. 

  2522. not_coded:

  2523.     if(!coded) {

  2524.         int k, scale;

  2525.         ac_val = s->ac_val[0][0] + s->block_index[n] * 16;

  2526.         ac_val2 = ac_val;

  2527. 

  2528.         scale = v->pq * 2 + v->halfpq;

  2529.         memset(ac_val2, 0, 16 * 2);

  2530.         if(dc_pred_dir) {//left

  2531.             ac_val -= 16;

  2532.             if(s->ac_pred)

  2533.                 memcpy(ac_val2, ac_val, 8 * 2);

  2534.         } else {//top

  2535.             ac_val -= 16 * s->block_wrap[n];

  2536.             if(s->ac_pred)

  2537.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  2538.         }

  2539. 

  2540.         /* apply AC prediction if needed */

  2541.         if(s->ac_pred) {

  2542.             if(dc_pred_dir) { //left

  2543.                 for(k = 1; k < 8; k++) {

  2544.                     block[k << 3] = ac_val[k] * scale;

  2545.                     if(!v->pquantizer && block[k << 3])

  2546.                         block[k << 3] += (block[k << 3] < 0) ? -v->pq : v->pq;

  2547.                 }

  2548.             } else { //top

  2549.                 for(k = 1; k < 8; k++) {

  2550.                     block[k] = ac_val[k + 8] * scale;

  2551.                     if(!v->pquantizer && block[k])

  2552.                         block[k] += (block[k] < 0) ? -v->pq : v->pq;

  2553.                 }

  2554.             }

  2555.             i = 63;

  2556.         }

  2557.     }

  2558.     s->block_last_index[n] = i;

  2559. 

  2560.     return 0;

  2561. }

  2562. 

  2563. /** Decode intra block in intra frames - should be faster than decode_intra_block

  2564.  * @param v VC1Context

  2565.  * @param block block to decode

  2566.  * @param coded are AC coeffs present or not

  2567.  * @param codingset set of VLC to decode data

  2568.  */

  2569. static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset, int mquant)

  2570. {

  2571.     GetBitContext *gb = &v->s.gb;

  2572.     MpegEncContext *s = &v->s;

  2573.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2574.     int run_diff, i;

  2575.     int16_t *dc_val;

  2576.     int16_t *ac_val, *ac_val2;

  2577.     int dcdiff;

  2578.     int a_avail = v->a_avail, c_avail = v->c_avail;

  2579.     int use_pred = s->ac_pred;

  2580.     int scale;

  2581.     int q1, q2 = 0;

  2582.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2583. 

  2584.     /* Get DC differential */

  2585.     if (n < 4) {

  2586.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2587.     } else {

  2588.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2589.     }

  2590.     if (dcdiff < 0){

  2591.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2592.         return -1;

  2593.     }

  2594.     if (dcdiff)

  2595.     {

  2596.         if (dcdiff == 119 /* ESC index value */)

  2597.         {

  2598.             /* TODO: Optimize */

  2599.             if (mquant == 1) dcdiff = get_bits(gb, 10);

  2600.             else if (mquant == 2) dcdiff = get_bits(gb, 9);

  2601.             else dcdiff = get_bits(gb, 8);

  2602.         }

  2603.         else

  2604.         {

  2605.             if (mquant == 1)

  2606.                 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;

  2607.             else if (mquant == 2)

  2608.                 dcdiff = (dcdiff<<1) + get_bits1(gb)   - 1;

  2609.         }

  2610.         if (get_bits1(gb))

  2611.             dcdiff = -dcdiff;

  2612.     }

  2613. 

  2614.     /* Prediction */

  2615.     dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);

  2616.     *dc_val = dcdiff;

  2617. 

  2618.     /* Store the quantized DC coeff, used for prediction */

  2619.     if (n < 4) {

  2620.         block[0] = dcdiff * s->y_dc_scale;

  2621.     } else {

  2622.         block[0] = dcdiff * s->c_dc_scale;

  2623.     }

  2624.     /* Skip ? */

  2625.     run_diff = 0;

  2626.     i = 0;

  2627. 

  2628.     //AC Decoding

  2629.     i = 1;

  2630. 

  2631.     /* check if AC is needed at all */

  2632.     if(!a_avail && !c_avail) use_pred = 0;

  2633.     ac_val = s->ac_val[0][0] + s->block_index[n] * 16;

  2634.     ac_val2 = ac_val;

  2635. 

  2636.     scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);

  2637. 

  2638.     if(dc_pred_dir) //left

  2639.         ac_val -= 16;

  2640.     else //top

  2641.         ac_val -= 16 * s->block_wrap[n];

  2642. 

  2643.     q1 = s->current_picture.qscale_table[mb_pos];

  2644.     if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];

  2645.     if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];

  2646.     if(dc_pred_dir && n==1) q2 = q1;

  2647.     if(!dc_pred_dir && n==2) q2 = q1;

  2648.     if(n==3) q2 = q1;

  2649. 

  2650.     if(coded) {

  2651.         int last = 0, skip, value;

  2652.         const int8_t *zz_table;

  2653.         int k;

  2654. 

  2655.         if(v->s.ac_pred) {

  2656.             if(!dc_pred_dir)

  2657.                 zz_table = wmv1_scantable[2];

  2658.             else

  2659.                 zz_table = wmv1_scantable[3];

  2660.         } else

  2661.             zz_table = wmv1_scantable[1];

  2662. 

  2663.         while (!last) {

  2664.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2665.             i += skip;

  2666.             if(i > 63)

  2667.                 break;

  2668.             block[zz_table[i++]] = value;

  2669.         }

  2670. 

  2671.         /* apply AC prediction if needed */

  2672.         if(use_pred) {

  2673.             /* scale predictors if needed*/

  2674.             if(q2 && q1!=q2) {

  2675.                 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2676.                 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2677. 

  2678.                 if(dc_pred_dir) { //left

  2679.                     for(k = 1; k < 8; k++)

  2680.                         block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2681.                 } else { //top

  2682.                     for(k = 1; k < 8; k++)

  2683.                         block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2684.                 }

  2685.             } else {

  2686.                 if(dc_pred_dir) { //left

  2687.                     for(k = 1; k < 8; k++)

  2688.                         block[k << 3] += ac_val[k];

  2689.                 } else { //top

  2690.                     for(k = 1; k < 8; k++)

  2691.                         block[k] += ac_val[k + 8];

  2692.                 }

  2693.             }

  2694.         }

  2695.         /* save AC coeffs for further prediction */

  2696.         for(k = 1; k < 8; k++) {

  2697.             ac_val2[k] = block[k << 3];

  2698.             ac_val2[k + 8] = block[k];

  2699.         }

  2700. 

  2701.         /* scale AC coeffs */

  2702.         for(k = 1; k < 64; k++)

  2703.             if(block[k]) {

  2704.                 block[k] *= scale;

  2705.                 if(!v->pquantizer)

  2706.                     block[k] += (block[k] < 0) ? -mquant : mquant;

  2707.             }

  2708. 

  2709.         if(use_pred) i = 63;

  2710.     } else { // no AC coeffs

  2711.         int k;

  2712. 

  2713.         memset(ac_val2, 0, 16 * 2);

  2714.         if(dc_pred_dir) {//left

  2715.             if(use_pred) {

  2716.                 memcpy(ac_val2, ac_val, 8 * 2);

  2717.                 if(q2 && q1!=q2) {

  2718.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2719.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2720.                     for(k = 1; k < 8; k++)

  2721.                         ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2722.                 }

  2723.             }

  2724.         } else {//top

  2725.             if(use_pred) {

  2726.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  2727.                 if(q2 && q1!=q2) {

  2728.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2729.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2730.                     for(k = 1; k < 8; k++)

  2731.                         ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2732.                 }

  2733.             }

  2734.         }

  2735. 

  2736.         /* apply AC prediction if needed */

  2737.         if(use_pred) {

  2738.             if(dc_pred_dir) { //left

  2739.                 for(k = 1; k < 8; k++) {

  2740.                     block[k << 3] = ac_val2[k] * scale;

  2741.                     if(!v->pquantizer && block[k << 3])

  2742.                         block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;

  2743.                 }

  2744.             } else { //top

  2745.                 for(k = 1; k < 8; k++) {

  2746.                     block[k] = ac_val2[k + 8] * scale;

  2747.                     if(!v->pquantizer && block[k])

  2748.                         block[k] += (block[k] < 0) ? -mquant : mquant;

  2749.                 }

  2750.             }

  2751.             i = 63;

  2752.         }

  2753.     }

  2754.     s->block_last_index[n] = i;

  2755. 

  2756.     return 0;

  2757. }

  2758. 

  2759. /** Decode intra block in inter frames - more generic version than vc1_decode_i_block

  2760.  * @param v VC1Context

  2761.  * @param block block to decode

  2762.  * @param coded are AC coeffs present or not

  2763.  * @param mquant block quantizer

  2764.  * @param codingset set of VLC to decode data

  2765.  */

  2766. static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int coded, int mquant, int codingset)

  2767. {

  2768.     GetBitContext *gb = &v->s.gb;

  2769.     MpegEncContext *s = &v->s;

  2770.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2771.     int run_diff, i;

  2772.     int16_t *dc_val;

  2773.     int16_t *ac_val, *ac_val2;

  2774.     int dcdiff;

  2775.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2776.     int a_avail = v->a_avail, c_avail = v->c_avail;

  2777.     int use_pred = s->ac_pred;

  2778.     int scale;

  2779.     int q1, q2 = 0;

  2780. 

  2781.     /* XXX: Guard against dumb values of mquant */

  2782.     mquant = (mquant < 1) ? 0 : ( (mquant>31) ? 31 : mquant );

  2783. 

  2784.     /* Set DC scale - y and c use the same */

  2785.     s->y_dc_scale = s->y_dc_scale_table[mquant];

  2786.     s->c_dc_scale = s->c_dc_scale_table[mquant];

  2787. 

  2788.     /* Get DC differential */

  2789.     if (n < 4) {

  2790.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2791.     } else {

  2792.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2793.     }

  2794.     if (dcdiff < 0){

  2795.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2796.         return -1;

  2797.     }

  2798.     if (dcdiff)

  2799.     {

  2800.         if (dcdiff == 119 /* ESC index value */)

  2801.         {

  2802.             /* TODO: Optimize */

  2803.             if (mquant == 1) dcdiff = get_bits(gb, 10);

  2804.             else if (mquant == 2) dcdiff = get_bits(gb, 9);

  2805.             else dcdiff = get_bits(gb, 8);

  2806.         }

  2807.         else

  2808.         {

  2809.             if (mquant == 1)

  2810.                 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;

  2811.             else if (mquant == 2)

  2812.                 dcdiff = (dcdiff<<1) + get_bits1(gb)   - 1;

  2813.         }

  2814.         if (get_bits1(gb))

  2815.             dcdiff = -dcdiff;

  2816.     }

  2817. 

  2818.     /* Prediction */

  2819.     dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);

  2820.     *dc_val = dcdiff;

  2821. 

  2822.     /* Store the quantized DC coeff, used for prediction */

  2823. 

  2824.     if (n < 4) {

  2825.         block[0] = dcdiff * s->y_dc_scale;

  2826.     } else {

  2827.         block[0] = dcdiff * s->c_dc_scale;

  2828.     }

  2829.     /* Skip ? */

  2830.     run_diff = 0;

  2831.     i = 0;

  2832. 

  2833.     //AC Decoding

  2834.     i = 1;

  2835. 

  2836.     /* check if AC is needed at all and adjust direction if needed */

  2837.     if(!a_avail) dc_pred_dir = 1;

  2838.     if(!c_avail) dc_pred_dir = 0;

  2839.     if(!a_avail && !c_avail) use_pred = 0;

  2840.     ac_val = s->ac_val[0][0] + s->block_index[n] * 16;

  2841.     ac_val2 = ac_val;

  2842. 

  2843.     scale = mquant * 2 + v->halfpq;

  2844. 

  2845.     if(dc_pred_dir) //left

  2846.         ac_val -= 16;

  2847.     else //top

  2848.         ac_val -= 16 * s->block_wrap[n];

  2849. 

  2850.     q1 = s->current_picture.qscale_table[mb_pos];

  2851.     if(dc_pred_dir && c_avail && mb_pos) q2 = s->current_picture.qscale_table[mb_pos - 1];

  2852.     if(!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];

  2853.     if(dc_pred_dir && n==1) q2 = q1;

  2854.     if(!dc_pred_dir && n==2) q2 = q1;

  2855.     if(n==3) q2 = q1;

  2856. 

  2857.     if(coded) {

  2858.         int last = 0, skip, value;

  2859.         const int8_t *zz_table;

  2860.         int k;

  2861. 

  2862.         zz_table = wmv1_scantable[0];

  2863. 

  2864.         while (!last) {

  2865.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2866.             i += skip;

  2867.             if(i > 63)

  2868.                 break;

  2869.             block[zz_table[i++]] = value;

  2870.         }

  2871. 

  2872.         /* apply AC prediction if needed */

  2873.         if(use_pred) {

  2874.             /* scale predictors if needed*/

  2875.             if(q2 && q1!=q2) {

  2876.                 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2877.                 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2878. 

  2879.                 if(dc_pred_dir) { //left

  2880.                     for(k = 1; k < 8; k++)

  2881.                         block[k << 3] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2882.                 } else { //top

  2883.                     for(k = 1; k < 8; k++)

  2884.                         block[k] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2885.                 }

  2886.             } else {

  2887.                 if(dc_pred_dir) { //left

  2888.                     for(k = 1; k < 8; k++)

  2889.                         block[k << 3] += ac_val[k];

  2890.                 } else { //top

  2891.                     for(k = 1; k < 8; k++)

  2892.                         block[k] += ac_val[k + 8];

  2893.                 }

  2894.             }

  2895.         }

  2896.         /* save AC coeffs for further prediction */

  2897.         for(k = 1; k < 8; k++) {

  2898.             ac_val2[k] = block[k << 3];

  2899.             ac_val2[k + 8] = block[k];

  2900.         }

  2901. 

  2902.         /* scale AC coeffs */

  2903.         for(k = 1; k < 64; k++)

  2904.             if(block[k]) {

  2905.                 block[k] *= scale;

  2906.                 if(!v->pquantizer)

  2907.                     block[k] += (block[k] < 0) ? -mquant : mquant;

  2908.             }

  2909. 

  2910.         if(use_pred) i = 63;

  2911.     } else { // no AC coeffs

  2912.         int k;

  2913. 

  2914.         memset(ac_val2, 0, 16 * 2);

  2915.         if(dc_pred_dir) {//left

  2916.             if(use_pred) {

  2917.                 memcpy(ac_val2, ac_val, 8 * 2);

  2918.                 if(q2 && q1!=q2) {

  2919.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2920.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2921.                     for(k = 1; k < 8; k++)

  2922.                         ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2923.                 }

  2924.             }

  2925.         } else {//top

  2926.             if(use_pred) {

  2927.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  2928.                 if(q2 && q1!=q2) {

  2929.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2930.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2931.                     for(k = 1; k < 8; k++)

  2932.                         ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2933.                 }

  2934.             }

  2935.         }

  2936. 

  2937.         /* apply AC prediction if needed */

  2938.         if(use_pred) {

  2939.             if(dc_pred_dir) { //left

  2940.                 for(k = 1; k < 8; k++) {

  2941.                     block[k << 3] = ac_val2[k] * scale;

  2942.                     if(!v->pquantizer && block[k << 3])

  2943.                         block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;

  2944.                 }

  2945.             } else { //top

  2946.                 for(k = 1; k < 8; k++) {

  2947.                     block[k] = ac_val2[k + 8] * scale;

  2948.                     if(!v->pquantizer && block[k])

  2949.                         block[k] += (block[k] < 0) ? -mquant : mquant;

  2950.                 }

  2951.             }

  2952.             i = 63;

  2953.         }

  2954.     }

  2955.     s->block_last_index[n] = i;

  2956. 

  2957.     return 0;

  2958. }

  2959. 

  2960. /** Decode P block

  2961.  */

  2962. static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block,

  2963.                               uint8_t *dst, int linesize, int skip_block, int apply_filter, int cbp_top, int cbp_left)

  2964. {

  2965.     MpegEncContext *s = &v->s;

  2966.     GetBitContext *gb = &s->gb;

  2967.     int i, j;

  2968.     int subblkpat = 0;

  2969.     int scale, off, idx, last, skip, value;

  2970.     int ttblk = ttmb & 7;

  2971.     int pat = 0;

  2972. 

  2973.     if(ttmb == -1) {

  2974.         ttblk = ff_vc1_ttblk_to_tt[v->tt_index][get_vlc2(gb, ff_vc1_ttblk_vlc[v->tt_index].table, VC1_TTBLK_VLC_BITS, 1)];

  2975.     }

  2976.     if(ttblk == TT_4X4) {

  2977.         subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);

  2978.     }

  2979.     if((ttblk != TT_8X8 && ttblk != TT_4X4) && (v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))) {

  2980.         subblkpat = decode012(gb);

  2981.         if(subblkpat) subblkpat ^= 3; //swap decoded pattern bits

  2982.         if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) ttblk = TT_8X4;

  2983.         if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) ttblk = TT_4X8;

  2984.     }

  2985.     scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);

  2986. 

  2987.     // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT

  2988.     if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {

  2989.         subblkpat = 2 - (ttblk == TT_8X4_TOP);

  2990.         ttblk = TT_8X4;

  2991.     }

  2992.     if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {

  2993.         subblkpat = 2 - (ttblk == TT_4X8_LEFT);

  2994.         ttblk = TT_4X8;

  2995.     }

  2996.     switch(ttblk) {

  2997.     case TT_8X8:

  2998.         pat = 0xF;

  2999.         i = 0;

  3000.         last = 0;

  3001.         while (!last) {

  3002.             vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3003.             i += skip;

  3004.             if(i > 63)

  3005.                 break;

  3006.             idx = wmv1_scantable[0][i++];

  3007.             block[idx] = value * scale;

  3008.             if(!v->pquantizer)

  3009.                 block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3010.         }

  3011.         if(!skip_block){

  3012.             s->dsp.vc1_inv_trans_8x8(block);

  3013.             s->dsp.add_pixels_clamped(block, dst, linesize);

  3014.             if(apply_filter && cbp_top  & 0xC)

  3015.                 vc1_loop_filter(dst, 1, linesize, 8, mquant);

  3016.             if(apply_filter && cbp_left & 0xA)

  3017.                 vc1_loop_filter(dst, linesize, 1, 8, mquant);

  3018.         }

  3019.         break;

  3020.     case TT_4X4:

  3021.         pat = ~subblkpat & 0xF;

  3022.         for(j = 0; j < 4; j++) {

  3023.             last = subblkpat & (1 << (3 - j));

  3024.             i = 0;

  3025.             off = (j & 1) * 4 + (j & 2) * 16;

  3026.             while (!last) {

  3027.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3028.                 i += skip;

  3029.                 if(i > 15)

  3030.                     break;

  3031.                 idx = ff_vc1_simple_progressive_4x4_zz[i++];

  3032.                 block[idx + off] = value * scale;

  3033.                 if(!v->pquantizer)

  3034.                     block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;

  3035.             }

  3036.             if(!(subblkpat & (1 << (3 - j))) && !skip_block){

  3037.                 s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off);

  3038.                 if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2)))))

  3039.                     vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, 1, linesize, 4, mquant);

  3040.                 if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1)))))

  3041.                     vc1_loop_filter(dst + (j&1)*4 + (j&2)*2*linesize, linesize, 1, 4, mquant);

  3042.             }

  3043.         }

  3044.         break;

  3045.     case TT_8X4:

  3046.         pat = ~((subblkpat & 2)*6 + (subblkpat & 1)*3) & 0xF;

  3047.         for(j = 0; j < 2; j++) {

  3048.             last = subblkpat & (1 << (1 - j));

  3049.             i = 0;

  3050.             off = j * 32;

  3051.             while (!last) {

  3052.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3053.                 i += skip;

  3054.                 if(i > 31)

  3055.                     break;

  3056.                 idx = v->zz_8x4[i++]+off;

  3057.                 block[idx] = value * scale;

  3058.                 if(!v->pquantizer)

  3059.                     block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3060.             }

  3061.             if(!(subblkpat & (1 << (1 - j))) && !skip_block){

  3062.                 s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off);

  3063.                 if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC))

  3064.                     vc1_loop_filter(dst + j*4*linesize, 1, linesize, 8, mquant);

  3065.                 if(apply_filter && cbp_left & (2 << j))

  3066.                     vc1_loop_filter(dst + j*4*linesize, linesize, 1, 4, mquant);

  3067.             }

  3068.         }

  3069.         break;

  3070.     case TT_4X8:

  3071.         pat = ~(subblkpat*5) & 0xF;

  3072.         for(j = 0; j < 2; j++) {

  3073.             last = subblkpat & (1 << (1 - j));

  3074.             i = 0;

  3075.             off = j * 4;

  3076.             while (!last) {

  3077.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3078.                 i += skip;

  3079.                 if(i > 31)

  3080.                     break;

  3081.                 idx = v->zz_4x8[i++]+off;

  3082.                 block[idx] = value * scale;

  3083.                 if(!v->pquantizer)

  3084.                     block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3085.             }

  3086.             if(!(subblkpat & (1 << (1 - j))) && !skip_block){

  3087.                 s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);

  3088.                 if(apply_filter && cbp_top & (2 << j))

  3089.                     vc1_loop_filter(dst + j*4, 1, linesize, 4, mquant);

  3090.                 if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA))

  3091.                     vc1_loop_filter(dst + j*4, linesize, 1, 8, mquant);

  3092.             }

  3093.         }

  3094.         break;

  3095.     }

  3096.     return pat;

  3097. }

  3098. 

  3099. 

  3100. /** Decode one P-frame MB (in Simple/Main profile)

  3101.  */

  3102. static int vc1_decode_p_mb(VC1Context *v)

  3103. {

  3104.     MpegEncContext *s = &v->s;

  3105.     GetBitContext *gb = &s->gb;

  3106.     int i, j;

  3107.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3108.     int cbp; /* cbp decoding stuff */

  3109.     int mqdiff, mquant; /* MB quantization */

  3110.     int ttmb = v->ttfrm; /* MB Transform type */

  3111. 

  3112.     static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },

  3113.       offset_table[6] = { 0, 1, 3, 7, 15, 31 };

  3114.     int mb_has_coeffs = 1; /* last_flag */

  3115.     int dmv_x, dmv_y; /* Differential MV components */

  3116.     int index, index1; /* LUT indexes */

  3117.     int val, sign; /* temp values */

  3118.     int first_block = 1;

  3119.     int dst_idx, off;

  3120.     int skipped, fourmv;

  3121.     int block_cbp = 0, pat;

  3122. 

  3123.     mquant = v->pq; /* Loosy initialization */

  3124. 

  3125.     if (v->mv_type_is_raw)

  3126.         fourmv = get_bits1(gb);

  3127.     else

  3128.         fourmv = v->mv_type_mb_plane[mb_pos];

  3129.     if (v->skip_is_raw)

  3130.         skipped = get_bits1(gb);

  3131.     else

  3132.         skipped = v->s.mbskip_table[mb_pos];

  3133. 

  3134.     s->dsp.clear_blocks(s->block[0]);

  3135. 

  3136.     if (!fourmv) /* 1MV mode */

  3137.     {

  3138.         if (!skipped)

  3139.         {

  3140.             GET_MVDATA(dmv_x, dmv_y);

  3141. 

  3142.             if (s->mb_intra) {

  3143.                 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;

  3144.                 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;

  3145.             }

  3146.             s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;

  3147.             vc1_pred_mv(s, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);

  3148. 

  3149.             /* FIXME Set DC val for inter block ? */

  3150.             if (s->mb_intra && !mb_has_coeffs)

  3151.             {

  3152.                 GET_MQUANT();

  3153.                 s->ac_pred = get_bits1(gb);

  3154.                 cbp = 0;

  3155.             }

  3156.             else if (mb_has_coeffs)

  3157.             {

  3158.                 if (s->mb_intra) s->ac_pred = get_bits1(gb);

  3159.                 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3160.                 GET_MQUANT();

  3161.             }

  3162.             else

  3163.             {

  3164.                 mquant = v->pq;

  3165.                 cbp = 0;

  3166.             }

  3167.             s->current_picture.qscale_table[mb_pos] = mquant;

  3168. 

  3169.             if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)

  3170.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,

  3171.                                 VC1_TTMB_VLC_BITS, 2);

  3172.             if(!s->mb_intra) vc1_mc_1mv(v, 0);

  3173.             dst_idx = 0;

  3174.             for (i=0; i<6; i++)

  3175.             {

  3176.                 s->dc_val[0][s->block_index[i]] = 0;

  3177.                 dst_idx += i >> 2;

  3178.                 val = ((cbp >> (5 - i)) & 1);

  3179.                 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3180.                 v->mb_type[0][s->block_index[i]] = s->mb_intra;

  3181.                 if(s->mb_intra) {

  3182.                     /* check if prediction blocks A and C are available */

  3183.                     v->a_avail = v->c_avail = 0;

  3184.                     if(i == 2 || i == 3 || !s->first_slice_line)

  3185.                         v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3186.                     if(i == 1 || i == 3 || s->mb_x)

  3187.                         v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3188. 

  3189.                     vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);

  3190.                     if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;

  3191.                     s->dsp.vc1_inv_trans_8x8(s->block[i]);

  3192.                     if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;

  3193.                     s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  3194.                     if(v->pq >= 9 && v->overlap) {

  3195.                         if(v->c_avail)

  3196.                             s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  3197.                         if(v->a_avail)

  3198.                             s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  3199.                     }

  3200.                     if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){

  3201.                         int left_cbp, top_cbp;

  3202.                         if(i & 4){

  3203.                             left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);

  3204.                             top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);

  3205.                         }else{

  3206.                             left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));

  3207.                             top_cbp  = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));

  3208.                         }

  3209.                         if(left_cbp & 0xC)

  3210.                             vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);

  3211.                         if(top_cbp  & 0xA)

  3212.                             vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);

  3213.                     }

  3214.                     block_cbp |= 0xF << (i << 2);

  3215.                 } else if(val) {

  3216.                     int left_cbp = 0, top_cbp = 0, filter = 0;

  3217.                     if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){

  3218.                         filter = 1;

  3219.                         if(i & 4){

  3220.                             left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);

  3221.                             top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);

  3222.                         }else{

  3223.                             left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));

  3224.                             top_cbp  = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));

  3225.                         }

  3226.                     }

  3227.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp);

  3228.                     block_cbp |= pat << (i << 2);

  3229.                     if(!v->ttmbf && ttmb < 8) ttmb = -1;

  3230.                     first_block = 0;

  3231.                 }

  3232.             }

  3233.         }

  3234.         else //Skipped

  3235.         {

  3236.             s->mb_intra = 0;

  3237.             for(i = 0; i < 6; i++) {

  3238.                 v->mb_type[0][s->block_index[i]] = 0;

  3239.                 s->dc_val[0][s->block_index[i]] = 0;

  3240.             }

  3241.             s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP;

  3242.             s->current_picture.qscale_table[mb_pos] = 0;

  3243.             vc1_pred_mv(s, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);

  3244.             vc1_mc_1mv(v, 0);

  3245.             return 0;

  3246.         }

  3247.     } //1MV mode

  3248.     else //4MV mode

  3249.     {

  3250.         if (!skipped /* unskipped MB */)

  3251.         {

  3252.             int intra_count = 0, coded_inter = 0;

  3253.             int is_intra[6], is_coded[6];

  3254.             /* Get CBPCY */

  3255.             cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3256.             for (i=0; i<6; i++)

  3257.             {

  3258.                 val = ((cbp >> (5 - i)) & 1);

  3259.                 s->dc_val[0][s->block_index[i]] = 0;

  3260.                 s->mb_intra = 0;

  3261.                 if(i < 4) {

  3262.                     dmv_x = dmv_y = 0;

  3263.                     s->mb_intra = 0;

  3264.                     mb_has_coeffs = 0;

  3265.                     if(val) {

  3266.                         GET_MVDATA(dmv_x, dmv_y);

  3267.                     }

  3268.                     vc1_pred_mv(s, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);

  3269.                     if(!s->mb_intra) vc1_mc_4mv_luma(v, i);

  3270.                     intra_count += s->mb_intra;

  3271.                     is_intra[i] = s->mb_intra;

  3272.                     is_coded[i] = mb_has_coeffs;

  3273.                 }

  3274.                 if(i&4){

  3275.                     is_intra[i] = (intra_count >= 3);

  3276.                     is_coded[i] = val;

  3277.                 }

  3278.                 if(i == 4) vc1_mc_4mv_chroma(v);

  3279.                 v->mb_type[0][s->block_index[i]] = is_intra[i];

  3280.                 if(!coded_inter) coded_inter = !is_intra[i] & is_coded[i];

  3281.             }

  3282.             // if there are no coded blocks then don't do anything more

  3283.             if(!intra_count && !coded_inter) return 0;

  3284.             dst_idx = 0;

  3285.             GET_MQUANT();

  3286.             s->current_picture.qscale_table[mb_pos] = mquant;

  3287.             /* test if block is intra and has pred */

  3288.             {

  3289.                 int intrapred = 0;

  3290.                 for(i=0; i<6; i++)

  3291.                     if(is_intra[i]) {

  3292.                         if(((!s->first_slice_line || (i==2 || i==3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])

  3293.                             || ((s->mb_x || (i==1 || i==3)) && v->mb_type[0][s->block_index[i] - 1])) {

  3294.                             intrapred = 1;

  3295.                             break;

  3296.                         }

  3297.                     }

  3298.                 if(intrapred)s->ac_pred = get_bits1(gb);

  3299.                 else s->ac_pred = 0;

  3300.             }

  3301.             if (!v->ttmbf && coded_inter)

  3302.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  3303.             for (i=0; i<6; i++)

  3304.             {

  3305.                 dst_idx += i >> 2;

  3306.                 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3307.                 s->mb_intra = is_intra[i];

  3308.                 if (is_intra[i]) {

  3309.                     /* check if prediction blocks A and C are available */

  3310.                     v->a_avail = v->c_avail = 0;

  3311.                     if(i == 2 || i == 3 || !s->first_slice_line)

  3312.                         v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3313.                     if(i == 1 || i == 3 || s->mb_x)

  3314.                         v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3315. 

  3316.                     vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant, (i&4)?v->codingset2:v->codingset);

  3317.                     if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;

  3318.                     s->dsp.vc1_inv_trans_8x8(s->block[i]);

  3319.                     if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;

  3320.                     s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);

  3321.                     if(v->pq >= 9 && v->overlap) {

  3322.                         if(v->c_avail)

  3323.                             s->dsp.vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  3324.                         if(v->a_avail)

  3325.                             s->dsp.vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  3326.                     }

  3327.                     if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){

  3328.                         int left_cbp, top_cbp;

  3329.                         if(i & 4){

  3330.                             left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);

  3331.                             top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);

  3332.                         }else{

  3333.                             left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));

  3334.                             top_cbp  = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));

  3335.                         }

  3336.                         if(left_cbp & 0xC)

  3337.                             vc1_loop_filter(s->dest[dst_idx] + off, 1, i & 4 ? s->uvlinesize : s->linesize, 8, mquant);

  3338.                         if(top_cbp  & 0xA)

  3339.                             vc1_loop_filter(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize, 1, 8, mquant);

  3340.                     }

  3341.                     block_cbp |= 0xF << (i << 2);

  3342.                 } else if(is_coded[i]) {

  3343.                     int left_cbp = 0, top_cbp = 0, filter = 0;

  3344.                     if(v->s.loop_filter && s->mb_x && s->mb_x != (s->mb_width - 1) && s->mb_y && s->mb_y != (s->mb_height - 1)){

  3345.                         filter = 1;

  3346.                         if(i & 4){

  3347.                             left_cbp = v->cbp[s->mb_x - 1]            >> (i * 4);

  3348.                             top_cbp  = v->cbp[s->mb_x - s->mb_stride] >> (i * 4);

  3349.                         }else{

  3350.                             left_cbp = (i & 1) ? (pat >> ((i-1)*4)) : (v->cbp[s->mb_x - 1]           >> ((i+1)*4));

  3351.                             top_cbp  = (i & 2) ? (pat >> ((i-2)*4)) : (v->cbp[s->mb_x - s->mb_stride] >> ((i+2)*4));

  3352.                         }

  3353.                     }

  3354.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), filter, left_cbp, top_cbp);

  3355.                     block_cbp |= pat << (i << 2);

  3356.                     if(!v->ttmbf && ttmb < 8) ttmb = -1;

  3357.                     first_block = 0;

  3358.                 }

  3359.             }

  3360.             return 0;

  3361.         }

  3362.         else //Skipped MB

  3363.         {

  3364.             s->mb_intra = 0;

  3365.             s->current_picture.qscale_table[mb_pos] = 0;

  3366.             for (i=0; i<6; i++) {

  3367.                 v->mb_type[0][s->block_index[i]] = 0;

  3368.                 s->dc_val[0][s->block_index[i]] = 0;

  3369.             }

  3370.             for (i=0; i<4; i++)

  3371.             {

  3372.                 vc1_pred_mv(s, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);

  3373.                 vc1_mc_4mv_luma(v, i);

  3374.             }

  3375.             vc1_mc_4mv_chroma(v);

  3376.             s->current_picture.qscale_table[mb_pos] = 0;

  3377.             return 0;

  3378.         }

  3379.     }

  3380.     v->cbp[s->mb_x] = block_cbp;

  3381. 

  3382.     /* Should never happen */

  3383.     return -1;

  3384. }

  3385. 

  3386. /** Decode one B-frame MB (in Main profile)

  3387.  */

  3388. static void vc1_decode_b_mb(VC1Context *v)

  3389. {

  3390.     MpegEncContext *s = &v->s;

  3391.     GetBitContext *gb = &s->gb;

  3392.     int i, j;

  3393.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3394.     int cbp = 0; /* cbp decoding stuff */

  3395.     int mqdiff, mquant; /* MB quantization */

  3396.     int ttmb = v->ttfrm; /* MB Transform type */

  3397. 

  3398.     static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },

  3399.       offset_table[6] = { 0, 1, 3, 7, 15, 31 };

  3400.     int mb_has_coeffs = 0; /* last_flag */

  3401.     int index, index1; /* LUT indexes */

  3402.     int val, sign; /* temp values */

  3403.     int first_block = 1;

  3404.     int dst_idx, off;

  3405.     int skipped, direct;

  3406.     int dmv_x[2], dmv_y[2];

  3407.     int bmvtype = BMV_TYPE_BACKWARD;

  3408. 

  3409.     mquant = v->pq; /* Loosy initialization */

  3410.     s->mb_intra = 0;

  3411. 

  3412.     if (v->dmb_is_raw)

  3413.         direct = get_bits1(gb);

  3414.     else

  3415.         direct = v->direct_mb_plane[mb_pos];

  3416.     if (v->skip_is_raw)

  3417.         skipped = get_bits1(gb);

  3418.     else

  3419.         skipped = v->s.mbskip_table[mb_pos];

  3420. 

  3421.     s->dsp.clear_blocks(s->block[0]);

  3422.     dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;

  3423.     for(i = 0; i < 6; i++) {

  3424.         v->mb_type[0][s->block_index[i]] = 0;

  3425.         s->dc_val[0][s->block_index[i]] = 0;

  3426.     }

  3427.     s->current_picture.qscale_table[mb_pos] = 0;

  3428. 

  3429.     if (!direct) {

  3430.         if (!skipped) {

  3431.             GET_MVDATA(dmv_x[0], dmv_y[0]);

  3432.             dmv_x[1] = dmv_x[0];

  3433.             dmv_y[1] = dmv_y[0];

  3434.         }

  3435.         if(skipped || !s->mb_intra) {

  3436.             bmvtype = decode012(gb);

  3437.             switch(bmvtype) {

  3438.             case 0:

  3439.                 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;

  3440.                 break;

  3441.             case 1:

  3442.                 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;

  3443.                 break;

  3444.             case 2:

  3445.                 bmvtype = BMV_TYPE_INTERPOLATED;

  3446.                 dmv_x[0] = dmv_y[0] = 0;

  3447.             }

  3448.         }

  3449.     }

  3450.     for(i = 0; i < 6; i++)

  3451.         v->mb_type[0][s->block_index[i]] = s->mb_intra;

  3452. 

  3453.     if (skipped) {

  3454.         if(direct) bmvtype = BMV_TYPE_INTERPOLATED;

  3455.         vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  3456.         vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  3457.         return;

  3458.     }

  3459.     if (direct) {

  3460.         cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3461.         GET_MQUANT();

  3462.         s->mb_intra = 0;

  3463.         mb_has_coeffs = 0;

  3464.         s->current_picture.qscale_table[mb_pos] = mquant;

  3465.         if(!v->ttmbf)

  3466.             ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  3467.         dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;

  3468.         vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  3469.         vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  3470.     } else {

  3471.         if(!mb_has_coeffs && !s->mb_intra) {

  3472.             /* no coded blocks - effectively skipped */

  3473.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  3474.             vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  3475.             return;

  3476.         }

  3477.         if(s->mb_intra && !mb_has_coeffs) {

  3478.             GET_MQUANT();

  3479.             s->current_picture.qscale_table[mb_pos] = mquant;

  3480.             s->ac_pred = get_bits1(gb);

  3481.             cbp = 0;

  3482.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  3483.         } else {

  3484.             if(bmvtype == BMV_TYPE_INTERPOLATED) {

  3485.                 GET_MVDATA(dmv_x[0], dmv_y[0]);

  3486.                 if(!mb_has_coeffs) {

  3487.                     /* interpolated skipped block */

  3488.                     vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  3489.                     vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  3490.                     return;

  3491.                 }

  3492.             }

  3493.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  3494.             if(!s->mb_intra) {

  3495.                 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  3496.             }

  3497.             if(s->mb_intra)

  3498.                 s->ac_pred = get_bits1(gb);

  3499.             cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3500.             GET_MQUANT();

  3501.             s->current_picture.qscale_table[mb_pos] = mquant;

  3502.             if(!v->ttmbf && !s->mb_intra && mb_has_coeffs)

  3503.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  3504.         }

  3505.     }

  3506.     dst_idx = 0;

  3507.     for (i=0; i<6; i++)

  3508.     {

  3509.         s->dc_val[0][s->block_index[i]] = 0;

  3510.         dst_idx += i >> 2;

  3511.         val = ((cbp >> (5 - i)) & 1);

  3512.         off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3513.         v->mb_type[0][s->block_index[i]] = s->mb_intra;

  3514.         if(s->mb_intra) {

  3515.             /* check if prediction blocks A and C are available */

  3516.             v->a_avail = v->c_avail = 0;

  3517.             if(i == 2 || i == 3 || !s->first_slice_line)

  3518.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3519.             if(i == 1 || i == 3 || s->mb_x)

  3520.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3521. 

  3522.             vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);

  3523.             if((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;

  3524.             s->dsp.vc1_inv_trans_8x8(s->block[i]);

  3525.             if(v->rangeredfrm) for(j = 0; j < 64; j++) s->block[i][j] <<= 1;

  3526.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  3527.         } else if(val) {

  3528.             vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize, (i&4) && (s->flags & CODEC_FLAG_GRAY), 0, 0, 0);

  3529.             if(!v->ttmbf && ttmb < 8) ttmb = -1;

  3530.             first_block = 0;

  3531.         }

  3532.     }

  3533. }

  3534. 

  3535. /** Decode blocks of I-frame

  3536.  */

  3537. static void vc1_decode_i_blocks(VC1Context *v)

  3538. {

  3539.     int k, j;

  3540.     MpegEncContext *s = &v->s;

  3541.     int cbp, val;

  3542.     uint8_t *coded_val;

  3543.     int mb_pos;

  3544. 

  3545.     /* select codingmode used for VLC tables selection */

  3546.     switch(v->y_ac_table_index){

  3547.     case 0:

  3548.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  3549.         break;

  3550.     case 1:

  3551.         v->codingset = CS_HIGH_MOT_INTRA;

  3552.         break;

  3553.     case 2:

  3554.         v->codingset = CS_MID_RATE_INTRA;

  3555.         break;

  3556.     }

  3557. 

  3558.     switch(v->c_ac_table_index){

  3559.     case 0:

  3560.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  3561.         break;

  3562.     case 1:

  3563.         v->codingset2 = CS_HIGH_MOT_INTER;

  3564.         break;

  3565.     case 2:

  3566.         v->codingset2 = CS_MID_RATE_INTER;

  3567.         break;

  3568.     }

  3569. 

  3570.     /* Set DC scale - y and c use the same */

  3571.     s->y_dc_scale = s->y_dc_scale_table[v->pq];

  3572.     s->c_dc_scale = s->c_dc_scale_table[v->pq];

  3573. 

  3574.     //do frame decode

  3575.     s->mb_x = s->mb_y = 0;

  3576.     s->mb_intra = 1;

  3577.     s->first_slice_line = 1;

  3578.     for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {

  3579.         for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {

  3580.             ff_init_block_index(s);

  3581.             ff_update_block_index(s);

  3582.             s->dsp.clear_blocks(s->block[0]);

  3583.             mb_pos = s->mb_x + s->mb_y * s->mb_width;

  3584.             s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;

  3585.             s->current_picture.qscale_table[mb_pos] = v->pq;

  3586.             s->current_picture.motion_val[1][s->block_index[0]][0] = 0;

  3587.             s->current_picture.motion_val[1][s->block_index[0]][1] = 0;

  3588. 

  3589.             // do actual MB decoding and displaying

  3590.             cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);

  3591.             v->s.ac_pred = get_bits1(&v->s.gb);

  3592. 

  3593.             for(k = 0; k < 6; k++) {

  3594.                 val = ((cbp >> (5 - k)) & 1);

  3595. 

  3596.                 if (k < 4) {

  3597.                     int pred = vc1_coded_block_pred(&v->s, k, &coded_val);

  3598.                     val = val ^ pred;

  3599.                     *coded_val = val;

  3600.                 }

  3601.                 cbp |= val << (5 - k);

  3602. 

  3603.                 vc1_decode_i_block(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2);

  3604. 

  3605.                 s->dsp.vc1_inv_trans_8x8(s->block[k]);

  3606.                 if(v->pq >= 9 && v->overlap) {

  3607.                     for(j = 0; j < 64; j++) s->block[k][j] += 128;

  3608.                 }

  3609.             }

  3610. 

  3611.             vc1_put_block(v, s->block);

  3612.             if(v->pq >= 9 && v->overlap) {

  3613.                 if(s->mb_x) {

  3614.                     s->dsp.vc1_h_overlap(s->dest[0], s->linesize);

  3615.                     s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  3616.                     if(!(s->flags & CODEC_FLAG_GRAY)) {

  3617.                         s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);

  3618.                         s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);

  3619.                     }

  3620.                 }

  3621.                 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);

  3622.                 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  3623.                 if(!s->first_slice_line) {

  3624.                     s->dsp.vc1_v_overlap(s->dest[0], s->linesize);

  3625.                     s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);

  3626.                     if(!(s->flags & CODEC_FLAG_GRAY)) {

  3627.                         s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);

  3628.                         s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);

  3629.                     }

  3630.                 }

  3631.                 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  3632.                 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  3633.             }

  3634.             if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);

  3635. 

  3636.             if(get_bits_count(&s->gb) > v->bits) {

  3637.                 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));

  3638.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);

  3639.                 return;

  3640.             }

  3641.         }

  3642.         ff_draw_horiz_band(s, s->mb_y * 16, 16);

  3643.         s->first_slice_line = 0;

  3644.     }

  3645.     ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));

  3646. }

  3647. 

  3648. /** Decode blocks of I-frame for advanced profile

  3649.  */

  3650. static void vc1_decode_i_blocks_adv(VC1Context *v)

  3651. {

  3652.     int k, j;

  3653.     MpegEncContext *s = &v->s;

  3654.     int cbp, val;

  3655.     uint8_t *coded_val;

  3656.     int mb_pos;

  3657.     int mquant = v->pq;

  3658.     int mqdiff;

  3659.     int overlap;

  3660.     GetBitContext *gb = &s->gb;

  3661. 

  3662.     /* select codingmode used for VLC tables selection */

  3663.     switch(v->y_ac_table_index){

  3664.     case 0:

  3665.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  3666.         break;

  3667.     case 1:

  3668.         v->codingset = CS_HIGH_MOT_INTRA;

  3669.         break;

  3670.     case 2:

  3671.         v->codingset = CS_MID_RATE_INTRA;

  3672.         break;

  3673.     }

  3674. 

  3675.     switch(v->c_ac_table_index){

  3676.     case 0:

  3677.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  3678.         break;

  3679.     case 1:

  3680.         v->codingset2 = CS_HIGH_MOT_INTER;

  3681.         break;

  3682.     case 2:

  3683.         v->codingset2 = CS_MID_RATE_INTER;

  3684.         break;

  3685.     }

  3686. 

  3687.     //do frame decode

  3688.     s->mb_x = s->mb_y = 0;

  3689.     s->mb_intra = 1;

  3690.     s->first_slice_line = 1;

  3691.     for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {

  3692.         for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {

  3693.             ff_init_block_index(s);

  3694.             ff_update_block_index(s);

  3695.             s->dsp.clear_blocks(s->block[0]);

  3696.             mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3697.             s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;

  3698.             s->current_picture.motion_val[1][s->block_index[0]][0] = 0;

  3699.             s->current_picture.motion_val[1][s->block_index[0]][1] = 0;

  3700. 

  3701.             // do actual MB decoding and displaying

  3702.             cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);

  3703.             if(v->acpred_is_raw)

  3704.                 v->s.ac_pred = get_bits1(&v->s.gb);

  3705.             else

  3706.                 v->s.ac_pred = v->acpred_plane[mb_pos];

  3707. 

  3708.             if(v->condover == CONDOVER_SELECT) {

  3709.                 if(v->overflg_is_raw)

  3710.                     overlap = get_bits1(&v->s.gb);

  3711.                 else

  3712.                     overlap = v->over_flags_plane[mb_pos];

  3713.             } else

  3714.                 overlap = (v->condover == CONDOVER_ALL);

  3715. 

  3716.             GET_MQUANT();

  3717. 

  3718.             s->current_picture.qscale_table[mb_pos] = mquant;

  3719.             /* Set DC scale - y and c use the same */

  3720.             s->y_dc_scale = s->y_dc_scale_table[mquant];

  3721.             s->c_dc_scale = s->c_dc_scale_table[mquant];

  3722. 

  3723.             for(k = 0; k < 6; k++) {

  3724.                 val = ((cbp >> (5 - k)) & 1);

  3725. 

  3726.                 if (k < 4) {

  3727.                     int pred = vc1_coded_block_pred(&v->s, k, &coded_val);

  3728.                     val = val ^ pred;

  3729.                     *coded_val = val;

  3730.                 }

  3731.                 cbp |= val << (5 - k);

  3732. 

  3733.                 v->a_avail = !s->first_slice_line || (k==2 || k==3);

  3734.                 v->c_avail = !!s->mb_x || (k==1 || k==3);

  3735. 

  3736.                 vc1_decode_i_block_adv(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2, mquant);

  3737. 

  3738.                 s->dsp.vc1_inv_trans_8x8(s->block[k]);

  3739.                 for(j = 0; j < 64; j++) s->block[k][j] += 128;

  3740.             }

  3741. 

  3742.             vc1_put_block(v, s->block);

  3743.             if(overlap) {

  3744.                 if(s->mb_x) {

  3745.                     s->dsp.vc1_h_overlap(s->dest[0], s->linesize);

  3746.                     s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  3747.                     if(!(s->flags & CODEC_FLAG_GRAY)) {

  3748.                         s->dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);

  3749.                         s->dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);

  3750.                     }

  3751.                 }

  3752.                 s->dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);

  3753.                 s->dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  3754.                 if(!s->first_slice_line) {

  3755.                     s->dsp.vc1_v_overlap(s->dest[0], s->linesize);

  3756.                     s->dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);

  3757.                     if(!(s->flags & CODEC_FLAG_GRAY)) {

  3758.                         s->dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);

  3759.                         s->dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);

  3760.                     }

  3761.                 }

  3762.                 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  3763.                 s->dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  3764.             }

  3765.             if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[mb_pos]);

  3766. 

  3767.             if(get_bits_count(&s->gb) > v->bits) {

  3768.                 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));

  3769.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);

  3770.                 return;

  3771.             }

  3772.         }

  3773.         ff_draw_horiz_band(s, s->mb_y * 16, 16);

  3774.         s->first_slice_line = 0;

  3775.     }

  3776.     ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));

  3777. }

  3778. 

  3779. static void vc1_decode_p_blocks(VC1Context *v)

  3780. {

  3781.     MpegEncContext *s = &v->s;

  3782. 

  3783.     /* select codingmode used for VLC tables selection */

  3784.     switch(v->c_ac_table_index){

  3785.     case 0:

  3786.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  3787.         break;

  3788.     case 1:

  3789.         v->codingset = CS_HIGH_MOT_INTRA;

  3790.         break;

  3791.     case 2:

  3792.         v->codingset = CS_MID_RATE_INTRA;

  3793.         break;

  3794.     }

  3795. 

  3796.     switch(v->c_ac_table_index){

  3797.     case 0:

  3798.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  3799.         break;

  3800.     case 1:

  3801.         v->codingset2 = CS_HIGH_MOT_INTER;

  3802.         break;

  3803.     case 2:

  3804.         v->codingset2 = CS_MID_RATE_INTER;

  3805.         break;

  3806.     }

  3807. 

  3808.     s->first_slice_line = 1;

  3809.     memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);

  3810.     for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {

  3811.         for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {

  3812.             ff_init_block_index(s);

  3813.             ff_update_block_index(s);

  3814.             s->dsp.clear_blocks(s->block[0]);

  3815. 

  3816.             vc1_decode_p_mb(v);

  3817.             if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {

  3818.                 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));

  3819.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n", get_bits_count(&s->gb), v->bits,s->mb_x,s->mb_y);

  3820.                 return;

  3821.             }

  3822.         }

  3823.         memmove(v->cbp_base, v->cbp, sizeof(v->cbp_base[0])*s->mb_stride);

  3824.         ff_draw_horiz_band(s, s->mb_y * 16, 16);

  3825.         s->first_slice_line = 0;

  3826.     }

  3827.     ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));

  3828. }

  3829. 

  3830. static void vc1_decode_b_blocks(VC1Context *v)

  3831. {

  3832.     MpegEncContext *s = &v->s;

  3833. 

  3834.     /* select codingmode used for VLC tables selection */

  3835.     switch(v->c_ac_table_index){

  3836.     case 0:

  3837.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  3838.         break;

  3839.     case 1:

  3840.         v->codingset = CS_HIGH_MOT_INTRA;

  3841.         break;

  3842.     case 2:

  3843.         v->codingset = CS_MID_RATE_INTRA;

  3844.         break;

  3845.     }

  3846. 

  3847.     switch(v->c_ac_table_index){

  3848.     case 0:

  3849.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  3850.         break;

  3851.     case 1:

  3852.         v->codingset2 = CS_HIGH_MOT_INTER;

  3853.         break;

  3854.     case 2:

  3855.         v->codingset2 = CS_MID_RATE_INTER;

  3856.         break;

  3857.     }

  3858. 

  3859.     s->first_slice_line = 1;

  3860.     for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {

  3861.         for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {

  3862.             ff_init_block_index(s);

  3863.             ff_update_block_index(s);

  3864.             s->dsp.clear_blocks(s->block[0]);

  3865. 

  3866.             vc1_decode_b_mb(v);

  3867.             if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {

  3868.                 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END));

  3869.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n", get_bits_count(&s->gb), v->bits,s->mb_x,s->mb_y);

  3870.                 return;

  3871.             }

  3872.             if(v->s.loop_filter) vc1_loop_filter_iblk(s, s->current_picture.qscale_table[s->mb_x + s->mb_y *s->mb_stride]);

  3873.         }

  3874.         ff_draw_horiz_band(s, s->mb_y * 16, 16);

  3875.         s->first_slice_line = 0;

  3876.     }

  3877.     ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));

  3878. }

  3879. 

  3880. static void vc1_decode_skip_blocks(VC1Context *v)

  3881. {

  3882.     MpegEncContext *s = &v->s;

  3883. 

  3884.     ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));

  3885.     s->first_slice_line = 1;

  3886.     for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {

  3887.         s->mb_x = 0;

  3888.         ff_init_block_index(s);

  3889.         ff_update_block_index(s);

  3890.         memcpy(s->dest[0], s->last_picture.data[0] + s->mb_y * 16 * s->linesize, s->linesize * 16);

  3891.         memcpy(s->dest[1], s->last_picture.data[1] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);

  3892.         memcpy(s->dest[2], s->last_picture.data[2] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8);

  3893.         ff_draw_horiz_band(s, s->mb_y * 16, 16);

  3894.         s->first_slice_line = 0;

  3895.     }

  3896.     s->pict_type = FF_P_TYPE;

  3897. }

  3898. 

  3899. static void vc1_decode_blocks(VC1Context *v)

  3900. {

  3901. 

  3902.     v->s.esc3_level_length = 0;

  3903.     if(v->x8_type){

  3904.         ff_intrax8_decode_picture(&v->x8, 2*v->pq+v->halfpq, v->pq*(!v->pquantizer) );

  3905.     }else{

  3906. 

  3907.         switch(v->s.pict_type) {

  3908.         case FF_I_TYPE:

  3909.             if(v->profile == PROFILE_ADVANCED)

  3910.                 vc1_decode_i_blocks_adv(v);

  3911.             else

  3912.                 vc1_decode_i_blocks(v);

  3913.             break;

  3914.         case FF_P_TYPE:

  3915.             if(v->p_frame_skipped)

  3916.                 vc1_decode_skip_blocks(v);

  3917.             else

  3918.                 vc1_decode_p_blocks(v);

  3919.             break;

  3920.         case FF_B_TYPE:

  3921.             if(v->bi_type){

  3922.                 if(v->profile == PROFILE_ADVANCED)

  3923.                     vc1_decode_i_blocks_adv(v);

  3924.                 else

  3925.                     vc1_decode_i_blocks(v);

  3926.             }else

  3927.                 vc1_decode_b_blocks(v);

  3928.             break;

  3929.         }

  3930.     }

  3931. }

  3932. 

  3933. /** Find VC-1 marker in buffer

  3934.  * @return position where next marker starts or end of buffer if no marker found

  3935.  */

  3936. static av_always_inline const uint8_t* find_next_marker(const uint8_t *src, const uint8_t *end)

  3937. {

  3938.     uint32_t mrk = 0xFFFFFFFF;

  3939. 

  3940.     if(end-src < 4) return end;

  3941.     while(src < end){

  3942.         mrk = (mrk << 8) | *src++;

  3943.         if(IS_MARKER(mrk))

  3944.             return src-4;

  3945.     }

  3946.     return end;

  3947. }

  3948. 

  3949. static av_always_inline int vc1_unescape_buffer(const uint8_t *src, int size, uint8_t *dst)

  3950. {

  3951.     int dsize = 0, i;

  3952. 

  3953.     if(size < 4){

  3954.         for(dsize = 0; dsize < size; dsize++) *dst++ = *src++;

  3955.         return size;

  3956.     }

  3957.     for(i = 0; i < size; i++, src++) {

  3958.         if(src[0] == 3 && i >= 2 && !src[-1] && !src[-2] && i < size-1 && src[1] < 4) {

  3959.             dst[dsize++] = src[1];

  3960.             src++;

  3961.             i++;

  3962.         } else

  3963.             dst[dsize++] = *src;

  3964.     }

  3965.     return dsize;

  3966. }

  3967. 

  3968. /** Initialize a VC1/WMV3 decoder

  3969.  * @todo TODO: Handle VC-1 IDUs (Transport level?)

  3970.  * @todo TODO: Decypher remaining bits in extra_data

  3971.  */

  3972. static av_cold int vc1_decode_init(AVCodecContext *avctx)

  3973. {

  3974.     VC1Context *v = avctx->priv_data;

  3975.     MpegEncContext *s = &v->s;

  3976.     GetBitContext gb;

  3977. 

  3978.     if (!avctx->extradata_size || !avctx->extradata) return -1;

  3979.     if (!(avctx->flags & CODEC_FLAG_GRAY))

  3980.         avctx->pix_fmt = PIX_FMT_YUV420P;

  3981.     else

  3982.         avctx->pix_fmt = PIX_FMT_GRAY8;

  3983.     v->s.avctx = avctx;

  3984.     avctx->flags |= CODEC_FLAG_EMU_EDGE;

  3985.     v->s.flags |= CODEC_FLAG_EMU_EDGE;

  3986. 

  3987.     if(avctx->idct_algo==FF_IDCT_AUTO){

  3988.         avctx->idct_algo=FF_IDCT_WMV2;

  3989.     }

  3990. 

  3991.     if(ff_h263_decode_init(avctx) < 0)

  3992.         return -1;

  3993.     if (vc1_init_common(v) < 0) return -1;

  3994. 

  3995.     avctx->coded_width = avctx->width;

  3996.     avctx->coded_height = avctx->height;

  3997.     if (avctx->codec_id == CODEC_ID_WMV3)

  3998.     {

  3999.         int count = 0;

  4000. 

  4001.         // looks like WMV3 has a sequence header stored in the extradata

  4002.         // advanced sequence header may be before the first frame

  4003.         // the last byte of the extradata is a version number, 1 for the

  4004.         // samples we can decode

  4005. 

  4006.         init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);

  4007. 

  4008.         if (decode_sequence_header(avctx, &gb) < 0)

  4009.           return -1;

  4010. 

  4011.         count = avctx->extradata_size*8 - get_bits_count(&gb);

  4012.         if (count>0)

  4013.         {

  4014.             av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",

  4015.                    count, get_bits(&gb, count));

  4016.         }

  4017.         else if (count < 0)

  4018.         {

  4019.             av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);

  4020.         }

  4021.     } else { // VC1/WVC1

  4022.         const uint8_t *start = avctx->extradata;

  4023.         uint8_t *end = avctx->extradata + avctx->extradata_size;

  4024.         const uint8_t *next;

  4025.         int size, buf2_size;

  4026.         uint8_t *buf2 = NULL;

  4027.         int seq_initialized = 0, ep_initialized = 0;

  4028. 

  4029.         if(avctx->extradata_size < 16) {

  4030.             av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);

  4031.             return -1;

  4032.         }

  4033. 

  4034.         buf2 = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);

  4035.         if(start[0]) start++; // in WVC1 extradata first byte is its size

  4036.         next = start;

  4037.         for(; next < end; start = next){

  4038.             next = find_next_marker(start + 4, end);

  4039.             size = next - start - 4;

  4040.             if(size <= 0) continue;

  4041.             buf2_size = vc1_unescape_buffer(start + 4, size, buf2);

  4042.             init_get_bits(&gb, buf2, buf2_size * 8);

  4043.             switch(AV_RB32(start)){

  4044.             case VC1_CODE_SEQHDR:

  4045.                 if(decode_sequence_header(avctx, &gb) < 0){

  4046.                     av_free(buf2);

  4047.                     return -1;

  4048.                 }

  4049.                 seq_initialized = 1;

  4050.                 break;

  4051.             case VC1_CODE_ENTRYPOINT:

  4052.                 if(decode_entry_point(avctx, &gb) < 0){

  4053.                     av_free(buf2);

  4054.                     return -1;

  4055.                 }

  4056.                 ep_initialized = 1;

  4057.                 break;

  4058.             }

  4059.         }

  4060.         av_free(buf2);

  4061.         if(!seq_initialized || !ep_initialized){

  4062.             av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");

  4063.             return -1;

  4064.         }

  4065.     }

  4066.     avctx->has_b_frames= !!(avctx->max_b_frames);

  4067.     s->low_delay = !avctx->has_b_frames;

  4068. 

  4069.     s->mb_width = (avctx->coded_width+15)>>4;

  4070.     s->mb_height = (avctx->coded_height+15)>>4;

  4071. 

  4072.     /* Allocate mb bitplanes */

  4073.     v->mv_type_mb_plane = av_malloc(s->mb_stride * s->mb_height);

  4074.     v->direct_mb_plane = av_malloc(s->mb_stride * s->mb_height);

  4075.     v->acpred_plane = av_malloc(s->mb_stride * s->mb_height);

  4076.     v->over_flags_plane = av_malloc(s->mb_stride * s->mb_height);

  4077. 

  4078.     v->cbp_base = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);

  4079.     v->cbp = v->cbp_base + s->mb_stride;

  4080. 

  4081.     /* allocate block type info in that way so it could be used with s->block_index[] */

  4082.     v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  4083.     v->mb_type[0] = v->mb_type_base + s->b8_stride + 1;

  4084.     v->mb_type[1] = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;

  4085.     v->mb_type[2] = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);

  4086. 

  4087.     /* Init coded blocks info */

  4088.     if (v->profile == PROFILE_ADVANCED)

  4089.     {

  4090. //        if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)

  4091. //            return -1;

  4092. //        if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)

  4093. //            return -1;

  4094.     }

  4095. 

  4096.     ff_intrax8_common_init(&v->x8,s);

  4097.     return 0;

  4098. }

  4099. 

  4100. 

  4101. /** Decode a VC1/WMV3 frame

  4102.  * @todo TODO: Handle VC-1 IDUs (Transport level?)

  4103.  */

  4104. static int vc1_decode_frame(AVCodecContext *avctx,

  4105.                             void *data, int *data_size,

  4106.                             const uint8_t *buf, int buf_size)

  4107. {

  4108.     VC1Context *v = avctx->priv_data;

  4109.     MpegEncContext *s = &v->s;

  4110.     AVFrame *pict = data;

  4111.     uint8_t *buf2 = NULL;

  4112. 

  4113.     /* no supplementary picture */

  4114.     if (buf_size == 0) {

  4115.         /* special case for last picture */

  4116.         if (s->low_delay==0 && s->next_picture_ptr) {

  4117.             *pict= *(AVFrame*)s->next_picture_ptr;

  4118.             s->next_picture_ptr= NULL;

  4119. 

  4120.             *data_size = sizeof(AVFrame);

  4121.         }

  4122. 

  4123.         return 0;

  4124.     }

  4125. 

  4126.     /* We need to set current_picture_ptr before reading the header,

  4127.      * otherwise we cannot store anything in there. */

  4128.     if(s->current_picture_ptr==NULL || s->current_picture_ptr->data[0]){

  4129.         int i= ff_find_unused_picture(s, 0);

  4130.         s->current_picture_ptr= &s->picture[i];

  4131.     }

  4132. 

  4133.     //for advanced profile we may need to parse and unescape data

  4134.     if (avctx->codec_id == CODEC_ID_VC1) {

  4135.         int buf_size2 = 0;

  4136.         buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  4137. 

  4138.         if(IS_MARKER(AV_RB32(buf))){ /* frame starts with marker and needs to be parsed */

  4139.             const uint8_t *start, *end, *next;

  4140.             int size;

  4141. 

  4142.             next = buf;

  4143.             for(start = buf, end = buf + buf_size; next < end; start = next){

  4144.                 next = find_next_marker(start + 4, end);

  4145.                 size = next - start - 4;

  4146.                 if(size <= 0) continue;

  4147.                 switch(AV_RB32(start)){

  4148.                 case VC1_CODE_FRAME:

  4149.                     buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);

  4150.                     break;

  4151.                 case VC1_CODE_ENTRYPOINT: /* it should be before frame data */

  4152.                     buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);

  4153.                     init_get_bits(&s->gb, buf2, buf_size2*8);

  4154.                     decode_entry_point(avctx, &s->gb);

  4155.                     break;

  4156.                 case VC1_CODE_SLICE:

  4157.                     av_log(avctx, AV_LOG_ERROR, "Sliced decoding is not implemented (yet)\n");

  4158.                     av_free(buf2);

  4159.                     return -1;

  4160.                 }

  4161.             }

  4162.         }else if(v->interlace && ((buf[0] & 0xC0) == 0xC0)){ /* WVC1 interlaced stores both fields divided by marker */

  4163.             const uint8_t *divider;

  4164. 

  4165.             divider = find_next_marker(buf, buf + buf_size);

  4166.             if((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD){

  4167.                 av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");

  4168.                 av_free(buf2);

  4169.                 return -1;

  4170.             }

  4171. 

  4172.             buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);

  4173.             // TODO

  4174.             av_free(buf2);return -1;

  4175.         }else{

  4176.             buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);

  4177.         }

  4178.         init_get_bits(&s->gb, buf2, buf_size2*8);

  4179.     } else

  4180.         init_get_bits(&s->gb, buf, buf_size*8);

  4181.     // do parse frame header

  4182.     if(v->profile < PROFILE_ADVANCED) {

  4183.         if(vc1_parse_frame_header(v, &s->gb) == -1) {

  4184.             av_free(buf2);

  4185.             return -1;

  4186.         }

  4187.     } else {

  4188.         if(vc1_parse_frame_header_adv(v, &s->gb) == -1) {

  4189.             av_free(buf2);

  4190.             return -1;

  4191.         }

  4192.     }

  4193. 

  4194.     if(s->pict_type != FF_I_TYPE && !v->res_rtm_flag){

  4195.         av_free(buf2);

  4196.         return -1;

  4197.     }

  4198. 

  4199.     // for hurry_up==5

  4200.     s->current_picture.pict_type= s->pict_type;

  4201.     s->current_picture.key_frame= s->pict_type == FF_I_TYPE;

  4202. 

  4203.     /* skip B-frames if we don't have reference frames */

  4204.     if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)){

  4205.         av_free(buf2);

  4206.         return -1;//buf_size;

  4207.     }

  4208.     /* skip b frames if we are in a hurry */

  4209.     if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return -1;//buf_size;

  4210.     if(   (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)

  4211.        || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)

  4212.        ||  avctx->skip_frame >= AVDISCARD_ALL) {

  4213.         av_free(buf2);

  4214.         return buf_size;

  4215.     }

  4216.     /* skip everything if we are in a hurry>=5 */

  4217.     if(avctx->hurry_up>=5) {

  4218.         av_free(buf2);

  4219.         return -1;//buf_size;

  4220.     }

  4221. 

  4222.     if(s->next_p_frame_damaged){

  4223.         if(s->pict_type==FF_B_TYPE)

  4224.             return buf_size;

  4225.         else

  4226.             s->next_p_frame_damaged=0;

  4227.     }

  4228. 

  4229.     if(MPV_frame_start(s, avctx) < 0) {

  4230.         av_free(buf2);

  4231.         return -1;

  4232.     }

  4233. 

  4234.     s->me.qpel_put= s->dsp.put_qpel_pixels_tab;

  4235.     s->me.qpel_avg= s->dsp.avg_qpel_pixels_tab;

  4236. 

  4237.     ff_er_frame_start(s);

  4238. 

  4239.     v->bits = buf_size * 8;

  4240.     vc1_decode_blocks(v);

  4241. //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8);

  4242. //  if(get_bits_count(&s->gb) > buf_size * 8)

  4243. //      return -1;

  4244.     ff_er_frame_end(s);

  4245. 

  4246.     MPV_frame_end(s);

  4247. 

  4248. assert(s->current_picture.pict_type == s->current_picture_ptr->pict_type);

  4249. assert(s->current_picture.pict_type == s->pict_type);

  4250.     if (s->pict_type == FF_B_TYPE || s->low_delay) {

  4251.         *pict= *(AVFrame*)s->current_picture_ptr;

  4252.     } else if (s->last_picture_ptr != NULL) {

  4253.         *pict= *(AVFrame*)s->last_picture_ptr;

  4254.     }

  4255. 

  4256.     if(s->last_picture_ptr || s->low_delay){

  4257.         *data_size = sizeof(AVFrame);

  4258.         ff_print_debug_info(s, pict);

  4259.     }

  4260. 

  4261.     /* Return the Picture timestamp as the frame number */

  4262.     /* we subtract 1 because it is added on utils.c     */

  4263.     avctx->frame_number = s->picture_number - 1;

  4264. 

  4265.     av_free(buf2);

  4266.     return buf_size;

  4267. }

  4268. 

  4269. 

  4270. /** Close a VC1/WMV3 decoder

  4271.  * @warning Initial try at using MpegEncContext stuff

  4272.  */

  4273. static av_cold int vc1_decode_end(AVCodecContext *avctx)

  4274. {

  4275.     VC1Context *v = avctx->priv_data;

  4276. 

  4277.     av_freep(&v->hrd_rate);

  4278.     av_freep(&v->hrd_buffer);

  4279.     MPV_common_end(&v->s);

  4280.     av_freep(&v->mv_type_mb_plane);

  4281.     av_freep(&v->direct_mb_plane);

  4282.     av_freep(&v->acpred_plane);

  4283.     av_freep(&v->over_flags_plane);

  4284.     av_freep(&v->mb_type_base);

  4285.     av_freep(&v->cbp_base);

  4286.     ff_intrax8_common_end(&v->x8);

  4287.     return 0;

  4288. }

  4289. 

  4290. 

  4291. AVCodec vc1_decoder = {

  4292.     "vc1",

  4293.     CODEC_TYPE_VIDEO,

  4294.     CODEC_ID_VC1,

  4295.     sizeof(VC1Context),

  4296.     vc1_decode_init,

  4297.     NULL,

  4298.     vc1_decode_end,

  4299.     vc1_decode_frame,

  4300.     CODEC_CAP_DELAY,

  4301.     NULL,

  4302.     .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),

  4303. };

  4304. 

  4305. AVCodec wmv3_decoder = {

  4306.     "wmv3",

  4307.     CODEC_TYPE_VIDEO,

  4308.     CODEC_ID_WMV3,

  4309.     sizeof(VC1Context),

  4310.     vc1_decode_init,

  4311.     NULL,

  4312.     vc1_decode_end,

  4313.     vc1_decode_frame,

  4314.     CODEC_CAP_DELAY,

  4315.     NULL,

  4316.     .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),

  4317. };