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.
40701 Commits
32 Branches
759MB
Tree: a28aa76ac8
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'a28aa76ac8'
${ noResults }
FFmpeg/libavcodec/vc1dec.c

5784 lines
223KB
Raw Blame History 

  1. /*

  2.  * VC-1 and WMV3 decoder

  3.  * Copyright (c) 2011 Mashiat Sarker Shakkhar

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

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

  6.  *

  7.  * This file is part of FFmpeg.

  8.  *

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

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

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

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

  13.  *

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

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

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

  17.  * Lesser General Public License for more details.

  18.  *

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

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

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

  22.  */

  23. 

  24. /**

  25.  * @file

  26.  * VC-1 and WMV3 decoder

  27.  */

  28. 

  29. #include "internal.h"

  30. #include "dsputil.h"

  31. #include "avcodec.h"

  32. #include "mpegvideo.h"

  33. #include "h263.h"

  34. #include "vc1.h"

  35. #include "vc1data.h"

  36. #include "vc1acdata.h"

  37. #include "msmpeg4data.h"

  38. #include "unary.h"

  39. #include "simple_idct.h"

  40. #include "mathops.h"

  41. #include "vdpau_internal.h"

  42. #include "libavutil/avassert.h"

  43. 

  44. #undef NDEBUG

  45. #include <assert.h>

  46. 

  47. #define MB_INTRA_VLC_BITS 9

  48. #define DC_VLC_BITS 9

  49. 

  50. 

  51. // offset tables for interlaced picture MVDATA decoding

  52. static const int offset_table1[9] = {  0,  1,  2,  4,  8, 16, 32,  64, 128 };

  53. static const int offset_table2[9] = {  0,  1,  3,  7, 15, 31, 63, 127, 255 };

  54. 

  55. /***********************************************************************/

  56. /**

  57.  * @name VC-1 Bitplane decoding

  58.  * @see 8.7, p56

  59.  * @{

  60.  */

  61. 

  62. /**

  63.  * Imode types

  64.  * @{

  65.  */

  66. enum Imode {

  67.     IMODE_RAW,

  68.     IMODE_NORM2,

  69.     IMODE_DIFF2,

  70.     IMODE_NORM6,

  71.     IMODE_DIFF6,

  72.     IMODE_ROWSKIP,

  73.     IMODE_COLSKIP

  74. };

  75. /** @} */ //imode defines

  76. 

  77. 

  78. /** @} */ //Bitplane group

  79. 

  80. static void vc1_put_signed_blocks_clamped(VC1Context *v)

  81. {

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

  83.     int topleft_mb_pos, top_mb_pos;

  84.     int stride_y, fieldtx;

  85.     int v_dist;

  86. 

  87.     /* The put pixels loop is always one MB row behind the decoding loop,

  88.      * because we can only put pixels when overlap filtering is done, and

  89.      * for filtering of the bottom edge of a MB, we need the next MB row

  90.      * present as well.

  91.      * Within the row, the put pixels loop is also one MB col behind the

  92.      * decoding loop. The reason for this is again, because for filtering

  93.      * of the right MB edge, we need the next MB present. */

  94.     if (!s->first_slice_line) {

  95.         if (s->mb_x) {

  96.             topleft_mb_pos = (s->mb_y - 1) * s->mb_stride + s->mb_x - 1;

  97.             fieldtx        = v->fieldtx_plane[topleft_mb_pos];

  98.             stride_y       = s->linesize << fieldtx;

  99.             v_dist         = (16 - fieldtx) >> (fieldtx == 0);

  100.             s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][0],

  101.                                              s->dest[0] - 16 * s->linesize - 16,

  102.                                              stride_y);

  103.             s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][1],

  104.                                              s->dest[0] - 16 * s->linesize - 8,

  105.                                              stride_y);

  106.             s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][2],

  107.                                              s->dest[0] - v_dist * s->linesize - 16,

  108.                                              stride_y);

  109.             s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][3],

  110.                                              s->dest[0] - v_dist * s->linesize - 8,

  111.                                              stride_y);

  112.             s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][4],

  113.                                              s->dest[1] - 8 * s->uvlinesize - 8,

  114.                                              s->uvlinesize);

  115.             s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][5],

  116.                                              s->dest[2] - 8 * s->uvlinesize - 8,

  117.                                              s->uvlinesize);

  118.         }

  119.         if (s->mb_x == s->mb_width - 1) {

  120.             top_mb_pos = (s->mb_y - 1) * s->mb_stride + s->mb_x;

  121.             fieldtx    = v->fieldtx_plane[top_mb_pos];

  122.             stride_y   = s->linesize << fieldtx;

  123.             v_dist     = fieldtx ? 15 : 8;

  124.             s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][0],

  125.                                              s->dest[0] - 16 * s->linesize,

  126.                                              stride_y);

  127.             s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][1],

  128.                                              s->dest[0] - 16 * s->linesize + 8,

  129.                                              stride_y);

  130.             s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][2],

  131.                                              s->dest[0] - v_dist * s->linesize,

  132.                                              stride_y);

  133.             s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][3],

  134.                                              s->dest[0] - v_dist * s->linesize + 8,

  135.                                              stride_y);

  136.             s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][4],

  137.                                              s->dest[1] - 8 * s->uvlinesize,

  138.                                              s->uvlinesize);

  139.             s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][5],

  140.                                              s->dest[2] - 8 * s->uvlinesize,

  141.                                              s->uvlinesize);

  142.         }

  143.     }

  144. 

  145. #define inc_blk_idx(idx) do { \

  146.         idx++; \

  147.         if (idx >= v->n_allocated_blks) \

  148.             idx = 0; \

  149.     } while (0)

  150. 

  151.     inc_blk_idx(v->topleft_blk_idx);

  152.     inc_blk_idx(v->top_blk_idx);

  153.     inc_blk_idx(v->left_blk_idx);

  154.     inc_blk_idx(v->cur_blk_idx);

  155. }

  156. 

  157. static void vc1_loop_filter_iblk(VC1Context *v, int pq)

  158. {

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

  160.     int j;

  161.     if (!s->first_slice_line) {

  162.         v->vc1dsp.vc1_v_loop_filter16(s->dest[0], s->linesize, pq);

  163.         if (s->mb_x)

  164.             v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);

  165.         v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize + 8, s->linesize, pq);

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

  167.             v->vc1dsp.vc1_v_loop_filter8(s->dest[j + 1], s->uvlinesize, pq);

  168.             if (s->mb_x)

  169.                 v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize, s->uvlinesize, pq);

  170.         }

  171.     }

  172.     v->vc1dsp.vc1_v_loop_filter16(s->dest[0] + 8 * s->linesize, s->linesize, pq);

  173. 

  174.     if (s->mb_y == s->end_mb_y - 1) {

  175.         if (s->mb_x) {

  176.             v->vc1dsp.vc1_h_loop_filter16(s->dest[0], s->linesize, pq);

  177.             v->vc1dsp.vc1_h_loop_filter8(s->dest[1], s->uvlinesize, pq);

  178.             v->vc1dsp.vc1_h_loop_filter8(s->dest[2], s->uvlinesize, pq);

  179.         }

  180.         v->vc1dsp.vc1_h_loop_filter16(s->dest[0] + 8, s->linesize, pq);

  181.     }

  182. }

  183. 

  184. static void vc1_loop_filter_iblk_delayed(VC1Context *v, int pq)

  185. {

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

  187.     int j;

  188. 

  189.     /* The loopfilter runs 1 row and 1 column behind the overlap filter, which

  190.      * means it runs two rows/cols behind the decoding loop. */

  191.     if (!s->first_slice_line) {

  192.         if (s->mb_x) {

  193.             if (s->mb_y >= s->start_mb_y + 2) {

  194.                 v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 16 * s->linesize - 16, s->linesize, pq);

  195. 

  196.                 if (s->mb_x >= 2)

  197.                     v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize - 16, s->linesize, pq);

  198.                 v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize - 8, s->linesize, pq);

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

  200.                     v->vc1dsp.vc1_v_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize - 8, s->uvlinesize, pq);

  201.                     if (s->mb_x >= 2) {

  202.                         v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 16 * s->uvlinesize - 8, s->uvlinesize, pq);

  203.                     }

  204.                 }

  205.             }

  206.             v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 8 * s->linesize - 16, s->linesize, pq);

  207.         }

  208. 

  209.         if (s->mb_x == s->mb_width - 1) {

  210.             if (s->mb_y >= s->start_mb_y + 2) {

  211.                 v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);

  212. 

  213.                 if (s->mb_x)

  214.                     v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize, s->linesize, pq);

  215.                 v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize + 8, s->linesize, pq);

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

  217.                     v->vc1dsp.vc1_v_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize, s->uvlinesize, pq);

  218.                     if (s->mb_x >= 2) {

  219.                         v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 16 * s->uvlinesize, s->uvlinesize, pq);

  220.                     }

  221.                 }

  222.             }

  223.             v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 8 * s->linesize, s->linesize, pq);

  224.         }

  225. 

  226.         if (s->mb_y == s->end_mb_y) {

  227.             if (s->mb_x) {

  228.                 if (s->mb_x >= 2)

  229.                     v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize - 16, s->linesize, pq);

  230.                 v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize - 8, s->linesize, pq);

  231.                 if (s->mb_x >= 2) {

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

  233.                         v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize - 8, s->uvlinesize, pq);

  234.                     }

  235.                 }

  236.             }

  237. 

  238.             if (s->mb_x == s->mb_width - 1) {

  239.                 if (s->mb_x)

  240.                     v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);

  241.                 v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize + 8, s->linesize, pq);

  242.                 if (s->mb_x) {

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

  244.                         v->vc1dsp.vc1_h_loop_filter8(s->dest[j + 1] - 8 * s->uvlinesize, s->uvlinesize, pq);

  245.                     }

  246.                 }

  247.             }

  248.         }

  249.     }

  250. }

  251. 

  252. static void vc1_smooth_overlap_filter_iblk(VC1Context *v)

  253. {

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

  255.     int mb_pos;

  256. 

  257.     if (v->condover == CONDOVER_NONE)

  258.         return;

  259. 

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

  261. 

  262.     /* Within a MB, the horizontal overlap always runs before the vertical.

  263.      * To accomplish that, we run the H on left and internal borders of the

  264.      * currently decoded MB. Then, we wait for the next overlap iteration

  265.      * to do H overlap on the right edge of this MB, before moving over and

  266.      * running the V overlap. Therefore, the V overlap makes us trail by one

  267.      * MB col and the H overlap filter makes us trail by one MB row. This

  268.      * is reflected in the time at which we run the put_pixels loop. */

  269.     if (v->condover == CONDOVER_ALL || v->pq >= 9 || v->over_flags_plane[mb_pos]) {

  270.         if (s->mb_x && (v->condover == CONDOVER_ALL || v->pq >= 9 ||

  271.                         v->over_flags_plane[mb_pos - 1])) {

  272.             v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][1],

  273.                                       v->block[v->cur_blk_idx][0]);

  274.             v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][3],

  275.                                       v->block[v->cur_blk_idx][2]);

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

  277.                 v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][4],

  278.                                           v->block[v->cur_blk_idx][4]);

  279.                 v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][5],

  280.                                           v->block[v->cur_blk_idx][5]);

  281.             }

  282.         }

  283.         v->vc1dsp.vc1_h_s_overlap(v->block[v->cur_blk_idx][0],

  284.                                   v->block[v->cur_blk_idx][1]);

  285.         v->vc1dsp.vc1_h_s_overlap(v->block[v->cur_blk_idx][2],

  286.                                   v->block[v->cur_blk_idx][3]);

  287. 

  288.         if (s->mb_x == s->mb_width - 1) {

  289.             if (!s->first_slice_line && (v->condover == CONDOVER_ALL || v->pq >= 9 ||

  290.                                          v->over_flags_plane[mb_pos - s->mb_stride])) {

  291.                 v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][2],

  292.                                           v->block[v->cur_blk_idx][0]);

  293.                 v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][3],

  294.                                           v->block[v->cur_blk_idx][1]);

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

  296.                     v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][4],

  297.                                               v->block[v->cur_blk_idx][4]);

  298.                     v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][5],

  299.                                               v->block[v->cur_blk_idx][5]);

  300.                 }

  301.             }

  302.             v->vc1dsp.vc1_v_s_overlap(v->block[v->cur_blk_idx][0],

  303.                                       v->block[v->cur_blk_idx][2]);

  304.             v->vc1dsp.vc1_v_s_overlap(v->block[v->cur_blk_idx][1],

  305.                                       v->block[v->cur_blk_idx][3]);

  306.         }

  307.     }

  308.     if (s->mb_x && (v->condover == CONDOVER_ALL || v->over_flags_plane[mb_pos - 1])) {

  309.         if (!s->first_slice_line && (v->condover == CONDOVER_ALL || v->pq >= 9 ||

  310.                                      v->over_flags_plane[mb_pos - s->mb_stride - 1])) {

  311.             v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][2],

  312.                                       v->block[v->left_blk_idx][0]);

  313.             v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][3],

  314.                                       v->block[v->left_blk_idx][1]);

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

  316.                 v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][4],

  317.                                           v->block[v->left_blk_idx][4]);

  318.                 v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][5],

  319.                                           v->block[v->left_blk_idx][5]);

  320.             }

  321.         }

  322.         v->vc1dsp.vc1_v_s_overlap(v->block[v->left_blk_idx][0],

  323.                                   v->block[v->left_blk_idx][2]);

  324.         v->vc1dsp.vc1_v_s_overlap(v->block[v->left_blk_idx][1],

  325.                                   v->block[v->left_blk_idx][3]);

  326.     }

  327. }

  328. 

  329. /** Do motion compensation over 1 macroblock

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

  331.  */

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

  333. {

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

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

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

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

  338.     int off, off_uv;

  339.     int v_edge_pos = s->v_edge_pos >> v->field_mode;

  340. 

  341.     if ((!v->field_mode ||

  342.          (v->ref_field_type[dir] == 1 && v->cur_field_type == 1)) &&

  343.         !v->s.last_picture.f.data[0])

  344.         return;

  345. 

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

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

  348. 

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

  350.     if (s->pict_type == AV_PICTURE_TYPE_P) {

  351.         s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = mx;

  352.         s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = my;

  353.     }

  354. 

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

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

  357.     v->luma_mv[s->mb_x][0] = uvmx;

  358.     v->luma_mv[s->mb_x][1] = uvmy;

  359. 

  360.     if (v->field_mode &&

  361.         v->cur_field_type != v->ref_field_type[dir]) {

  362.         my   = my   - 2 + 4 * v->cur_field_type;

  363.         uvmy = uvmy - 2 + 4 * v->cur_field_type;

  364.     }

  365. 

  366.     // fastuvmc shall be ignored for interlaced frame picture

  367.     if (v->fastuvmc && (v->fcm != ILACE_FRAME)) {

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

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

  370.     }

  371.     if (v->field_mode) { // interlaced field picture

  372.         if (!dir) {

  373.             if ((v->cur_field_type != v->ref_field_type[dir]) && v->second_field) {

  374.                 srcY = s->current_picture.f.data[0];

  375.                 srcU = s->current_picture.f.data[1];

  376.                 srcV = s->current_picture.f.data[2];

  377.             } else {

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

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

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

  381.             }

  382.         } else {

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

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

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

  386.         }

  387.     } else {

  388.         if (!dir) {

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

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

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

  392.         } else {

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

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

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

  396.         }

  397.     }

  398. 

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

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

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

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

  403. 

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

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

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

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

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

  409.     } else {

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

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

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

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

  414.     }

  415. 

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

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

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

  419. 

  420.     if (v->field_mode && v->ref_field_type[dir]) {

  421.         srcY += s->current_picture_ptr->f.linesize[0];

  422.         srcU += s->current_picture_ptr->f.linesize[1];

  423.         srcV += s->current_picture_ptr->f.linesize[2];

  424.     }

  425. 

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

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

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

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

  430.     }

  431. 

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

  433.         || s->h_edge_pos < 22 || v_edge_pos < 22

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

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

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

  437. 

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

  439.         s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize,

  440.                                 17 + s->mspel * 2, 17 + s->mspel * 2,

  441.                                 src_x - s->mspel, src_y - s->mspel,

  442.                                 s->h_edge_pos, v_edge_pos);

  443.         srcY = s->edge_emu_buffer;

  444.         s->dsp.emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8 + 1, 8 + 1,

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

  446.         s->dsp.emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8 + 1, 8 + 1,

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

  448.         srcU = uvbuf;

  449.         srcV = uvbuf + 16;

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

  451.         if (v->rangeredfrm) {

  452.             int i, j;

  453.             uint8_t *src, *src2;

  454. 

  455.             src = srcY;

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

  457.                 for (i = 0; i < 17 + s->mspel * 2; i++)

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

  459.                 src += s->linesize;

  460.             }

  461.             src  = srcU;

  462.             src2 = srcV;

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

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

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

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

  467.                 }

  468.                 src  += s->uvlinesize;

  469.                 src2 += s->uvlinesize;

  470.             }

  471.         }

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

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

  474.             int i, j;

  475.             uint8_t *src, *src2;

  476. 

  477.             src = srcY;

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

  479.                 for (i = 0; i < 17 + s->mspel * 2; i++)

  480.                     src[i] = v->luty[src[i]];

  481.                 src += s->linesize;

  482.             }

  483.             src  = srcU;

  484.             src2 = srcV;

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

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

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

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

  489.                 }

  490.                 src  += s->uvlinesize;

  491.                 src2 += s->uvlinesize;

  492.             }

  493.         }

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

  495.     }

  496. 

  497.     if (v->field_mode && v->second_field) {

  498.         off    = s->current_picture_ptr->f.linesize[0];

  499.         off_uv = s->current_picture_ptr->f.linesize[1];

  500.     } else {

  501.         off    = 0;

  502.         off_uv = 0;

  503.     }

  504.     if (s->mspel) {

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

  506.         v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off    , srcY    , s->linesize, v->rnd);

  507.         v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8, srcY + 8, s->linesize, v->rnd);

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

  509.         v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize    , srcY    , s->linesize, v->rnd);

  510.         v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);

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

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

  513.         if (!v->rnd)

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

  515.         else

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

  517.     }

  518. 

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

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

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

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

  523.     if (!v->rnd) {

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

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

  526.     } else {

  527.         v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy);

  528.         v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy);

  529.     }

  530. }

  531. 

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

  533. {

  534.     if (a < b) {

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

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

  537.     } else {

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

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

  540.     }

  541. }

  542. 

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

  544.  */

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

  546. {

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

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

  549.     uint8_t *srcY;

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

  551.     int off;

  552.     int fieldmv = (v->fcm == ILACE_FRAME) ? v->blk_mv_type[s->block_index[n]] : 0;

  553.     int v_edge_pos = s->v_edge_pos >> v->field_mode;

  554. 

  555.     if ((!v->field_mode ||

  556.          (v->ref_field_type[dir] == 1 && v->cur_field_type == 1)) &&

  557.         !v->s.last_picture.f.data[0])

  558.         return;

  559. 

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

  561.     my = s->mv[dir][n][1];

  562. 

  563.     if (!dir) {

  564.         if (v->field_mode) {

  565.             if ((v->cur_field_type != v->ref_field_type[dir]) && v->second_field)

  566.                 srcY = s->current_picture.f.data[0];

  567.             else

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

  569.         } else

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

  571.     } else

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

  573. 

  574.     if (v->field_mode) {

  575.         if (v->cur_field_type != v->ref_field_type[dir])

  576.             my = my - 2 + 4 * v->cur_field_type;

  577.     }

  578. 

  579.     if (s->pict_type == AV_PICTURE_TYPE_P && n == 3 && v->field_mode) {

  580.         int same_count = 0, opp_count = 0, k;

  581.         int chosen_mv[2][4][2], f;

  582.         int tx, ty;

  583.         for (k = 0; k < 4; k++) {

  584.             f = v->mv_f[0][s->block_index[k] + v->blocks_off];

  585.             chosen_mv[f][f ? opp_count : same_count][0] = s->mv[0][k][0];

  586.             chosen_mv[f][f ? opp_count : same_count][1] = s->mv[0][k][1];

  587.             opp_count  += f;

  588.             same_count += 1 - f;

  589.         }

  590.         f = opp_count > same_count;

  591.         switch (f ? opp_count : same_count) {

  592.         case 4:

  593.             tx = median4(chosen_mv[f][0][0], chosen_mv[f][1][0],

  594.                          chosen_mv[f][2][0], chosen_mv[f][3][0]);

  595.             ty = median4(chosen_mv[f][0][1], chosen_mv[f][1][1],

  596.                          chosen_mv[f][2][1], chosen_mv[f][3][1]);

  597.             break;

  598.         case 3:

  599.             tx = mid_pred(chosen_mv[f][0][0], chosen_mv[f][1][0], chosen_mv[f][2][0]);

  600.             ty = mid_pred(chosen_mv[f][0][1], chosen_mv[f][1][1], chosen_mv[f][2][1]);

  601.             break;

  602.         case 2:

  603.             tx = (chosen_mv[f][0][0] + chosen_mv[f][1][0]) / 2;

  604.             ty = (chosen_mv[f][0][1] + chosen_mv[f][1][1]) / 2;

  605.             break;

  606.         }

  607.         s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = tx;

  608.         s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = ty;

  609.         for (k = 0; k < 4; k++)

  610.             v->mv_f[1][s->block_index[k] + v->blocks_off] = f;

  611.     }

  612. 

  613.     if (v->fcm == ILACE_FRAME) {  // not sure if needed for other types of picture

  614.         int qx, qy;

  615.         int width  = s->avctx->coded_width;

  616.         int height = s->avctx->coded_height >> 1;

  617.         qx = (s->mb_x * 16) + (mx >> 2);

  618.         qy = (s->mb_y *  8) + (my >> 3);

  619. 

  620.         if (qx < -17)

  621.             mx -= 4 * (qx + 17);

  622.         else if (qx > width)

  623.             mx -= 4 * (qx - width);

  624.         if (qy < -18)

  625.             my -= 8 * (qy + 18);

  626.         else if (qy > height + 1)

  627.             my -= 8 * (qy - height - 1);

  628.     }

  629. 

  630.     if ((v->fcm == ILACE_FRAME) && fieldmv)

  631.         off = ((n > 1) ? s->linesize : 0) + (n & 1) * 8;

  632.     else

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

  634.     if (v->field_mode && v->second_field)

  635.         off += s->current_picture_ptr->f.linesize[0];

  636. 

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

  638.     if (!fieldmv)

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

  640.     else

  641.         src_y = s->mb_y * 16 + ((n > 1) ? 1 : 0) + (my >> 2);

  642. 

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

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

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

  646.     } else {

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

  648.         if (v->fcm == ILACE_FRAME) {

  649.             if (src_y & 1)

  650.                 src_y = av_clip(src_y, -17, s->avctx->coded_height + 1);

  651.             else

  652.                 src_y = av_clip(src_y, -18, s->avctx->coded_height);

  653.         } else {

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

  655.         }

  656.     }

  657. 

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

  659.     if (v->field_mode && v->ref_field_type[dir])

  660.         srcY += s->current_picture_ptr->f.linesize[0];

  661. 

  662.     if (fieldmv && !(src_y & 1))

  663.         v_edge_pos--;

  664.     if (fieldmv && (src_y & 1) && src_y < 4)

  665.         src_y--;

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

  667.         || s->h_edge_pos < 13 || v_edge_pos < 23

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

  669.         || (unsigned)(src_y - (s->mspel << fieldmv)) > v_edge_pos - (my & 3) - ((8 + s->mspel * 2) << fieldmv)) {

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

  671.         /* check emulate edge stride and offset */

  672.         s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize,

  673.                                 9 + s->mspel * 2, (9 + s->mspel * 2) << fieldmv,

  674.                                 src_x - s->mspel, src_y - (s->mspel << fieldmv),

  675.                                 s->h_edge_pos, v_edge_pos);

  676.         srcY = s->edge_emu_buffer;

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

  678.         if (v->rangeredfrm) {

  679.             int i, j;

  680.             uint8_t *src;

  681. 

  682.             src = srcY;

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

  684.                 for (i = 0; i < 9 + s->mspel * 2; i++)

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

  686.                 src += s->linesize << fieldmv;

  687.             }

  688.         }

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

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

  691.             int i, j;

  692.             uint8_t *src;

  693. 

  694.             src = srcY;

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

  696.                 for (i = 0; i < 9 + s->mspel * 2; i++)

  697.                     src[i] = v->luty[src[i]];

  698.                 src += s->linesize << fieldmv;

  699.             }

  700.         }

  701.         srcY += s->mspel * (1 + (s->linesize << fieldmv));

  702.     }

  703. 

  704.     if (s->mspel) {

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

  706.         v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd);

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

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

  709.         if (!v->rnd)

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

  711.         else

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

  713.     }

  714. }

  715. 

  716. static av_always_inline int get_chroma_mv(int *mvx, int *mvy, int *a, int flag, int *tx, int *ty)

  717. {

  718.     int idx, i;

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

  720. 

  721.     idx =  ((a[3] != flag) << 3)

  722.          | ((a[2] != flag) << 2)

  723.          | ((a[1] != flag) << 1)

  724.          |  (a[0] != flag);

  725.     if (!idx) {

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

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

  728.         return 4;

  729.     } else if (count[idx] == 1) {

  730.         switch (idx) {

  731.         case 0x1:

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

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

  734.             return 3;

  735.         case 0x2:

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

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

  738.             return 3;

  739.         case 0x4:

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

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

  742.             return 3;

  743.         case 0x8:

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

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

  746.             return 3;

  747.         }

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

  749.         int t1 = 0, t2 = 0;

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

  751.             if (!a[i]) {

  752.                 t1 = i;

  753.                 break;

  754.             }

  755.         for (i = t1 + 1; i < 4; i++)

  756.             if (!a[i]) {

  757.                 t2 = i;

  758.                 break;

  759.             }

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

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

  762.         return 2;

  763.     } else {

  764.         return 0;

  765.     }

  766.     return -1;

  767. }

  768. 

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

  770.  */

  771. static void vc1_mc_4mv_chroma(VC1Context *v, int dir)

  772. {

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

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

  775.     uint8_t *srcU, *srcV;

  776.     int uvmx, uvmy, uvsrc_x, uvsrc_y;

  777.     int k, tx = 0, ty = 0;

  778.     int mvx[4], mvy[4], intra[4], mv_f[4];

  779.     int valid_count;

  780.     int chroma_ref_type = v->cur_field_type, off = 0;

  781.     int v_edge_pos = s->v_edge_pos >> v->field_mode;

  782. 

  783.     if (!v->field_mode && !v->s.last_picture.f.data[0])

  784.         return;

  785.     if (s->flags & CODEC_FLAG_GRAY)

  786.         return;

  787. 

  788.     for (k = 0; k < 4; k++) {

  789.         mvx[k] = s->mv[dir][k][0];

  790.         mvy[k] = s->mv[dir][k][1];

  791.         intra[k] = v->mb_type[0][s->block_index[k]];

  792.         if (v->field_mode)

  793.             mv_f[k] = v->mv_f[dir][s->block_index[k] + v->blocks_off];

  794.     }

  795. 

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

  797.     if (!v->field_mode || (v->field_mode && !v->numref)) {

  798.         valid_count = get_chroma_mv(mvx, mvy, intra, 0, &tx, &ty);

  799.         if (!valid_count) {

  800.             s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;

  801.             s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;

  802.             v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;

  803.             return; //no need to do MC for intra blocks

  804.         }

  805.     } else {

  806.         int dominant = 0;

  807.         if (mv_f[0] + mv_f[1] + mv_f[2] + mv_f[3] > 2)

  808.             dominant = 1;

  809.         valid_count = get_chroma_mv(mvx, mvy, mv_f, dominant, &tx, &ty);

  810.         if (dominant)

  811.             chroma_ref_type = !v->cur_field_type;

  812.     }

  813.     if (v->field_mode && chroma_ref_type == 1 && v->cur_field_type == 1 && !v->s.last_picture.f.data[0])

  814.         return;

  815.     s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = tx;

  816.     s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = ty;

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

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

  819. 

  820.     v->luma_mv[s->mb_x][0] = uvmx;

  821.     v->luma_mv[s->mb_x][1] = uvmy;

  822. 

  823.     if (v->fastuvmc) {

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

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

  826.     }

  827.     // Field conversion bias

  828.     if (v->cur_field_type != chroma_ref_type)

  829.         uvmy += 2 - 4 * chroma_ref_type;

  830. 

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

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

  833. 

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

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

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

  837.     } else {

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

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

  840.     }

  841. 

  842.     if (!dir) {

  843.         if (v->field_mode) {

  844.             if ((v->cur_field_type != chroma_ref_type) && v->cur_field_type) {

  845.                 srcU = s->current_picture.f.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;

  846.                 srcV = s->current_picture.f.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;

  847.             } else {

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

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

  850.             }

  851.         } else {

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

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

  854.         }

  855.     } else {

  856.         srcU = s->next_picture.f.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;

  857.         srcV = s->next_picture.f.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;

  858.     }

  859. 

  860.     if (v->field_mode) {

  861.         if (chroma_ref_type) {

  862.             srcU += s->current_picture_ptr->f.linesize[1];

  863.             srcV += s->current_picture_ptr->f.linesize[2];

  864.         }

  865.         off = v->second_field ? s->current_picture_ptr->f.linesize[1] : 0;

  866.     }

  867. 

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

  869.         || s->h_edge_pos < 18 || v_edge_pos < 18

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

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

  872.         s->dsp.emulated_edge_mc(s->edge_emu_buffer     , srcU, s->uvlinesize,

  873.                                 8 + 1, 8 + 1, uvsrc_x, uvsrc_y,

  874.                                 s->h_edge_pos >> 1, v_edge_pos >> 1);

  875.         s->dsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize,

  876.                                 8 + 1, 8 + 1, uvsrc_x, uvsrc_y,

  877.                                 s->h_edge_pos >> 1, v_edge_pos >> 1);

  878.         srcU = s->edge_emu_buffer;

  879.         srcV = s->edge_emu_buffer + 16;

  880. 

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

  882.         if (v->rangeredfrm) {

  883.             int i, j;

  884.             uint8_t *src, *src2;

  885. 

  886.             src  = srcU;

  887.             src2 = srcV;

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

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

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

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

  892.                 }

  893.                 src  += s->uvlinesize;

  894.                 src2 += s->uvlinesize;

  895.             }

  896.         }

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

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

  899.             int i, j;

  900.             uint8_t *src, *src2;

  901. 

  902.             src  = srcU;

  903.             src2 = srcV;

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

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

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

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

  908.                 }

  909.                 src  += s->uvlinesize;

  910.                 src2 += s->uvlinesize;

  911.             }

  912.         }

  913.     }

  914. 

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

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

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

  918.     if (!v->rnd) {

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

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

  921.     } else {

  922.         v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off, srcU, s->uvlinesize, 8, uvmx, uvmy);

  923.         v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off, srcV, s->uvlinesize, 8, uvmx, uvmy);

  924.     }

  925. }

  926. 

  927. /** Do motion compensation for 4-MV field chroma macroblock (both U and V)

  928.  */

  929. static void vc1_mc_4mv_chroma4(VC1Context *v)

  930. {

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

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

  933.     uint8_t *srcU, *srcV;

  934.     int uvsrc_x, uvsrc_y;

  935.     int uvmx_field[4], uvmy_field[4];

  936.     int i, off, tx, ty;

  937.     int fieldmv = v->blk_mv_type[s->block_index[0]];

  938.     static const int s_rndtblfield[16] = { 0, 0, 1, 2, 4, 4, 5, 6, 2, 2, 3, 8, 6, 6, 7, 12 };

  939.     int v_dist = fieldmv ? 1 : 4; // vertical offset for lower sub-blocks

  940.     int v_edge_pos = s->v_edge_pos >> 1;

  941. 

  942.     if (!v->s.last_picture.f.data[0])

  943.         return;

  944.     if (s->flags & CODEC_FLAG_GRAY)

  945.         return;

  946. 

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

  948.         tx = s->mv[0][i][0];

  949.         uvmx_field[i] = (tx + ((tx & 3) == 3)) >> 1;

  950.         ty = s->mv[0][i][1];

  951.         if (fieldmv)

  952.             uvmy_field[i] = (ty >> 4) * 8 + s_rndtblfield[ty & 0xF];

  953.         else

  954.             uvmy_field[i] = (ty + ((ty & 3) == 3)) >> 1;

  955.     }

  956. 

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

  958.         off = (i & 1) * 4 + ((i & 2) ? v_dist * s->uvlinesize : 0);

  959.         uvsrc_x = s->mb_x * 8 +  (i & 1) * 4           + (uvmx_field[i] >> 2);

  960.         uvsrc_y = s->mb_y * 8 + ((i & 2) ? v_dist : 0) + (uvmy_field[i] >> 2);

  961.         // FIXME: implement proper pull-back (see vc1cropmv.c, vc1CROPMV_ChromaPullBack())

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

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

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

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

  966.         uvmx_field[i] = (uvmx_field[i] & 3) << 1;

  967.         uvmy_field[i] = (uvmy_field[i] & 3) << 1;

  968. 

  969.         if (fieldmv && !(uvsrc_y & 1))

  970.             v_edge_pos--;

  971.         if (fieldmv && (uvsrc_y & 1) && uvsrc_y < 2)

  972.             uvsrc_y--;

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

  974.             || s->h_edge_pos < 10 || v_edge_pos < (5 << fieldmv)

  975.             || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 5

  976.             || (unsigned)uvsrc_y > v_edge_pos - (5 << fieldmv)) {

  977.             s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcU, s->uvlinesize,

  978.                                     5, (5 << fieldmv), uvsrc_x, uvsrc_y,

  979.                                     s->h_edge_pos >> 1, v_edge_pos);

  980.             s->dsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize,

  981.                                     5, (5 << fieldmv), uvsrc_x, uvsrc_y,

  982.                                     s->h_edge_pos >> 1, v_edge_pos);

  983.             srcU = s->edge_emu_buffer;

  984.             srcV = s->edge_emu_buffer + 16;

  985. 

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

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

  988.                 int i, j;

  989.                 uint8_t *src, *src2;

  990. 

  991.                 src  = srcU;

  992.                 src2 = srcV;

  993.                 for (j = 0; j < 5; j++) {

  994.                     for (i = 0; i < 5; i++) {

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

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

  997.                     }

  998.                     src  += s->uvlinesize << 1;

  999.                     src2 += s->uvlinesize << 1;

  1000.                 }

  1001.             }

  1002.         }

  1003.         if (!v->rnd) {

  1004.             dsp->put_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);

  1005.             dsp->put_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);

  1006.         } else {

  1007.             v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);

  1008.             v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);

  1009.         }

  1010.     }

  1011. }

  1012. 

  1013. /***********************************************************************/

  1014. /**

  1015.  * @name VC-1 Block-level functions

  1016.  * @see 7.1.4, p91 and 8.1.1.7, p(1)04

  1017.  * @{

  1018.  */

  1019. 

  1020. /**

  1021.  * @def GET_MQUANT

  1022.  * @brief Get macroblock-level quantizer scale

  1023.  */

  1024. #define GET_MQUANT()                                           \

  1025.     if (v->dquantfrm) {                                        \

  1026.         int edges = 0;                                         \

  1027.         if (v->dqprofile == DQPROFILE_ALL_MBS) {               \

  1028.             if (v->dqbilevel) {                                \

  1029.                 mquant = (get_bits1(gb)) ? v->altpq : v->pq;   \

  1030.             } else {                                           \

  1031.                 mqdiff = get_bits(gb, 3);                      \

  1032.                 if (mqdiff != 7)                               \

  1033.                     mquant = v->pq + mqdiff;                   \

  1034.                 else                                           \

  1035.                     mquant = get_bits(gb, 5);                  \

  1036.             }                                                  \

  1037.         }                                                      \

  1038.         if (v->dqprofile == DQPROFILE_SINGLE_EDGE)             \

  1039.             edges = 1 << v->dqsbedge;                          \

  1040.         else if (v->dqprofile == DQPROFILE_DOUBLE_EDGES)       \

  1041.             edges = (3 << v->dqsbedge) % 15;                   \

  1042.         else if (v->dqprofile == DQPROFILE_FOUR_EDGES)         \

  1043.             edges = 15;                                        \

  1044.         if ((edges&1) && !s->mb_x)                             \

  1045.             mquant = v->altpq;                                 \

  1046.         if ((edges&2) && s->first_slice_line)                  \

  1047.             mquant = v->altpq;                                 \

  1048.         if ((edges&4) && s->mb_x == (s->mb_width - 1))         \

  1049.             mquant = v->altpq;                                 \

  1050.         if ((edges&8) && s->mb_y == (s->mb_height - 1))        \

  1051.             mquant = v->altpq;                                 \

  1052.         if (!mquant || mquant > 31) {                          \

  1053.             av_log(v->s.avctx, AV_LOG_ERROR, "invalid mquant %d\n", mquant);   \

  1054.             mquant = 1;                                \

  1055.         }                                              \

  1056.     }

  1057. 

  1058. /**

  1059.  * @def GET_MVDATA(_dmv_x, _dmv_y)

  1060.  * @brief Get MV differentials

  1061.  * @see MVDATA decoding from 8.3.5.2, p(1)20

  1062.  * @param _dmv_x Horizontal differential for decoded MV

  1063.  * @param _dmv_y Vertical differential for decoded MV

  1064.  */

  1065. #define GET_MVDATA(_dmv_x, _dmv_y)                                      \

  1066.     index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table, \

  1067.                          VC1_MV_DIFF_VLC_BITS, 2);                      \

  1068.     if (index > 36) {                                                   \

  1069.         mb_has_coeffs = 1;                                              \

  1070.         index -= 37;                                                    \

  1071.     } else                                                              \

  1072.         mb_has_coeffs = 0;                                              \

  1073.     s->mb_intra = 0;                                                    \

  1074.     if (!index) {                                                       \

  1075.         _dmv_x = _dmv_y = 0;                                            \

  1076.     } else if (index == 35) {                                           \

  1077.         _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample);          \

  1078.         _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample);          \

  1079.     } else if (index == 36) {                                           \

  1080.         _dmv_x = 0;                                                     \

  1081.         _dmv_y = 0;                                                     \

  1082.         s->mb_intra = 1;                                                \

  1083.     } else {                                                            \

  1084.         index1 = index % 6;                                             \

  1085.         if (!s->quarter_sample && index1 == 5) val = 1;                 \

  1086.         else                                   val = 0;                 \

  1087.         if (size_table[index1] - val > 0)                               \

  1088.             val = get_bits(gb, size_table[index1] - val);               \

  1089.         else                                   val = 0;                 \

  1090.         sign = 0 - (val&1);                                             \

  1091.         _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \

  1092.                                                                         \

  1093.         index1 = index / 6;                                             \

  1094.         if (!s->quarter_sample && index1 == 5) val = 1;                 \

  1095.         else                                   val = 0;                 \

  1096.         if (size_table[index1] - val > 0)                               \

  1097.             val = get_bits(gb, size_table[index1] - val);               \

  1098.         else                                   val = 0;                 \

  1099.         sign = 0 - (val & 1);                                           \

  1100.         _dmv_y = (sign ^ ((val >> 1) + offset_table[index1])) - sign;   \

  1101.     }

  1102. 

  1103. static av_always_inline void get_mvdata_interlaced(VC1Context *v, int *dmv_x,

  1104.                                                    int *dmv_y, int *pred_flag)

  1105. {

  1106.     int index, index1;

  1107.     int extend_x = 0, extend_y = 0;

  1108.     GetBitContext *gb = &v->s.gb;

  1109.     int bits, esc;

  1110.     int val, sign;

  1111.     const int* offs_tab;

  1112. 

  1113.     if (v->numref) {

  1114.         bits = VC1_2REF_MVDATA_VLC_BITS;

  1115.         esc  = 125;

  1116.     } else {

  1117.         bits = VC1_1REF_MVDATA_VLC_BITS;

  1118.         esc  = 71;

  1119.     }

  1120.     switch (v->dmvrange) {

  1121.     case 1:

  1122.         extend_x = 1;

  1123.         break;

  1124.     case 2:

  1125.         extend_y = 1;

  1126.         break;

  1127.     case 3:

  1128.         extend_x = extend_y = 1;

  1129.         break;

  1130.     }

  1131.     index = get_vlc2(gb, v->imv_vlc->table, bits, 3);

  1132.     if (index == esc) {

  1133.         *dmv_x = get_bits(gb, v->k_x);

  1134.         *dmv_y = get_bits(gb, v->k_y);

  1135.         if (v->numref) {

  1136.             *pred_flag = *dmv_y & 1;

  1137.             *dmv_y     = (*dmv_y + *pred_flag) >> 1;

  1138.         }

  1139.     }

  1140.     else {

  1141.         av_assert0(index < esc);

  1142.         if (extend_x)

  1143.             offs_tab = offset_table2;

  1144.         else

  1145.             offs_tab = offset_table1;

  1146.         index1 = (index + 1) % 9;

  1147.         if (index1 != 0) {

  1148.             val    = get_bits(gb, index1 + extend_x);

  1149.             sign   = 0 -(val & 1);

  1150.             *dmv_x = (sign ^ ((val >> 1) + offs_tab[index1])) - sign;

  1151.         } else

  1152.             *dmv_x = 0;

  1153.         if (extend_y)

  1154.             offs_tab = offset_table2;

  1155.         else

  1156.             offs_tab = offset_table1;

  1157.         index1 = (index + 1) / 9;

  1158.         if (index1 > v->numref) {

  1159.             val    = get_bits(gb, (index1 + (extend_y << v->numref)) >> v->numref);

  1160.             sign   = 0 - (val & 1);

  1161.             *dmv_y = (sign ^ ((val >> 1) + offs_tab[index1 >> v->numref])) - sign;

  1162.         } else

  1163.             *dmv_y = 0;

  1164.         if (v->numref)

  1165.             *pred_flag = index1 & 1;

  1166.     }

  1167. }

  1168. 

  1169. static av_always_inline int scaleforsame_x(VC1Context *v, int n /* MV */, int dir)

  1170. {

  1171.     int scaledvalue, refdist;

  1172.     int scalesame1, scalesame2;

  1173.     int scalezone1_x, zone1offset_x;

  1174.     int table_index = dir ^ v->second_field;

  1175. 

  1176.     if (v->s.pict_type != AV_PICTURE_TYPE_B)

  1177.         refdist = v->refdist;

  1178.     else

  1179.         refdist = dir ? v->brfd : v->frfd;

  1180.     if (refdist > 3)

  1181.         refdist = 3;

  1182.     scalesame1    = ff_vc1_field_mvpred_scales[table_index][1][refdist];

  1183.     scalesame2    = ff_vc1_field_mvpred_scales[table_index][2][refdist];

  1184.     scalezone1_x  = ff_vc1_field_mvpred_scales[table_index][3][refdist];

  1185.     zone1offset_x = ff_vc1_field_mvpred_scales[table_index][5][refdist];

  1186. 

  1187.     if (FFABS(n) > 255)

  1188.         scaledvalue = n;

  1189.     else {

  1190.         if (FFABS(n) < scalezone1_x)

  1191.             scaledvalue = (n * scalesame1) >> 8;

  1192.         else {

  1193.             if (n < 0)

  1194.                 scaledvalue = ((n * scalesame2) >> 8) - zone1offset_x;

  1195.             else

  1196.                 scaledvalue = ((n * scalesame2) >> 8) + zone1offset_x;

  1197.         }

  1198.     }

  1199.     return av_clip(scaledvalue, -v->range_x, v->range_x - 1);

  1200. }

  1201. 

  1202. static av_always_inline int scaleforsame_y(VC1Context *v, int i, int n /* MV */, int dir)

  1203. {

  1204.     int scaledvalue, refdist;

  1205.     int scalesame1, scalesame2;

  1206.     int scalezone1_y, zone1offset_y;

  1207.     int table_index = dir ^ v->second_field;

  1208. 

  1209.     if (v->s.pict_type != AV_PICTURE_TYPE_B)

  1210.         refdist = v->refdist;

  1211.     else

  1212.         refdist = dir ? v->brfd : v->frfd;

  1213.     if (refdist > 3)

  1214.         refdist = 3;

  1215.     scalesame1    = ff_vc1_field_mvpred_scales[table_index][1][refdist];

  1216.     scalesame2    = ff_vc1_field_mvpred_scales[table_index][2][refdist];

  1217.     scalezone1_y  = ff_vc1_field_mvpred_scales[table_index][4][refdist];

  1218.     zone1offset_y = ff_vc1_field_mvpred_scales[table_index][6][refdist];

  1219. 

  1220.     if (FFABS(n) > 63)

  1221.         scaledvalue = n;

  1222.     else {

  1223.         if (FFABS(n) < scalezone1_y)

  1224.             scaledvalue = (n * scalesame1) >> 8;

  1225.         else {

  1226.             if (n < 0)

  1227.                 scaledvalue = ((n * scalesame2) >> 8) - zone1offset_y;

  1228.             else

  1229.                 scaledvalue = ((n * scalesame2) >> 8) + zone1offset_y;

  1230.         }

  1231.     }

  1232. 

  1233.     if (v->cur_field_type && !v->ref_field_type[dir])

  1234.         return av_clip(scaledvalue, -v->range_y / 2 + 1, v->range_y / 2);

  1235.     else

  1236.         return av_clip(scaledvalue, -v->range_y / 2, v->range_y / 2 - 1);

  1237. }

  1238. 

  1239. static av_always_inline int scaleforopp_x(VC1Context *v, int n /* MV */)

  1240. {

  1241.     int scalezone1_x, zone1offset_x;

  1242.     int scaleopp1, scaleopp2, brfd;

  1243.     int scaledvalue;

  1244. 

  1245.     brfd = FFMIN(v->brfd, 3);

  1246.     scalezone1_x  = ff_vc1_b_field_mvpred_scales[3][brfd];

  1247.     zone1offset_x = ff_vc1_b_field_mvpred_scales[5][brfd];

  1248.     scaleopp1     = ff_vc1_b_field_mvpred_scales[1][brfd];

  1249.     scaleopp2     = ff_vc1_b_field_mvpred_scales[2][brfd];

  1250. 

  1251.     if (FFABS(n) > 255)

  1252.         scaledvalue = n;

  1253.     else {

  1254.         if (FFABS(n) < scalezone1_x)

  1255.             scaledvalue = (n * scaleopp1) >> 8;

  1256.         else {

  1257.             if (n < 0)

  1258.                 scaledvalue = ((n * scaleopp2) >> 8) - zone1offset_x;

  1259.             else

  1260.                 scaledvalue = ((n * scaleopp2) >> 8) + zone1offset_x;

  1261.         }

  1262.     }

  1263.     return av_clip(scaledvalue, -v->range_x, v->range_x - 1);

  1264. }

  1265. 

  1266. static av_always_inline int scaleforopp_y(VC1Context *v, int n /* MV */, int dir)

  1267. {

  1268.     int scalezone1_y, zone1offset_y;

  1269.     int scaleopp1, scaleopp2, brfd;

  1270.     int scaledvalue;

  1271. 

  1272.     brfd = FFMIN(v->brfd, 3);

  1273.     scalezone1_y  = ff_vc1_b_field_mvpred_scales[4][brfd];

  1274.     zone1offset_y = ff_vc1_b_field_mvpred_scales[6][brfd];

  1275.     scaleopp1     = ff_vc1_b_field_mvpred_scales[1][brfd];

  1276.     scaleopp2     = ff_vc1_b_field_mvpred_scales[2][brfd];

  1277. 

  1278.     if (FFABS(n) > 63)

  1279.         scaledvalue = n;

  1280.     else {

  1281.         if (FFABS(n) < scalezone1_y)

  1282.             scaledvalue = (n * scaleopp1) >> 8;

  1283.         else {

  1284.             if (n < 0)

  1285.                 scaledvalue = ((n * scaleopp2) >> 8) - zone1offset_y;

  1286.             else

  1287.                 scaledvalue = ((n * scaleopp2) >> 8) + zone1offset_y;

  1288.         }

  1289.     }

  1290.     if (v->cur_field_type && !v->ref_field_type[dir]) {

  1291.         return av_clip(scaledvalue, -v->range_y / 2 + 1, v->range_y / 2);

  1292.     } else {

  1293.         return av_clip(scaledvalue, -v->range_y / 2, v->range_y / 2 - 1);

  1294.     }

  1295. }

  1296. 

  1297. static av_always_inline int scaleforsame(VC1Context *v, int i, int n /* MV */,

  1298.                                          int dim, int dir)

  1299. {

  1300.     int brfd, scalesame;

  1301.     int hpel = 1 - v->s.quarter_sample;

  1302. 

  1303.     n >>= hpel;

  1304.     if (v->s.pict_type != AV_PICTURE_TYPE_B || v->second_field || !dir) {

  1305.         if (dim)

  1306.             n = scaleforsame_y(v, i, n, dir) << hpel;

  1307.         else

  1308.             n = scaleforsame_x(v, n, dir) << hpel;

  1309.         return n;

  1310.     }

  1311.     brfd      = FFMIN(v->brfd, 3);

  1312.     scalesame = ff_vc1_b_field_mvpred_scales[0][brfd];

  1313. 

  1314.     n = (n * scalesame >> 8) << hpel;

  1315.     return n;

  1316. }

  1317. 

  1318. static av_always_inline int scaleforopp(VC1Context *v, int n /* MV */,

  1319.                                         int dim, int dir)

  1320. {

  1321.     int refdist, scaleopp;

  1322.     int hpel = 1 - v->s.quarter_sample;

  1323. 

  1324.     n >>= hpel;

  1325.     if (v->s.pict_type == AV_PICTURE_TYPE_B && !v->second_field && dir == 1) {

  1326.         if (dim)

  1327.             n = scaleforopp_y(v, n, dir) << hpel;

  1328.         else

  1329.             n = scaleforopp_x(v, n) << hpel;

  1330.         return n;

  1331.     }

  1332.     if (v->s.pict_type != AV_PICTURE_TYPE_B)

  1333.         refdist = FFMIN(v->refdist, 3);

  1334.     else

  1335.         refdist = dir ? v->brfd : v->frfd;

  1336.     scaleopp = ff_vc1_field_mvpred_scales[dir ^ v->second_field][0][refdist];

  1337. 

  1338.     n = (n * scaleopp >> 8) << hpel;

  1339.     return n;

  1340. }

  1341. 

  1342. /** Predict and set motion vector

  1343.  */

  1344. static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y,

  1345.                                int mv1, int r_x, int r_y, uint8_t* is_intra,

  1346.                                int pred_flag, int dir)

  1347. {

  1348.     MpegEncContext *s = &v->s;

  1349.     int xy, wrap, off = 0;

  1350.     int16_t *A, *B, *C;

  1351.     int px, py;

  1352.     int sum;

  1353.     int mixedmv_pic, num_samefield = 0, num_oppfield = 0;

  1354.     int opposit, a_f, b_f, c_f;

  1355.     int16_t field_predA[2];

  1356.     int16_t field_predB[2];

  1357.     int16_t field_predC[2];

  1358.     int a_valid, b_valid, c_valid;

  1359.     int hybridmv_thresh, y_bias = 0;

  1360. 

  1361.     if (v->mv_mode == MV_PMODE_MIXED_MV ||

  1362.         ((v->mv_mode == MV_PMODE_INTENSITY_COMP) && (v->mv_mode2 == MV_PMODE_MIXED_MV)))

  1363.         mixedmv_pic = 1;

  1364.     else

  1365.         mixedmv_pic = 0;

  1366.     /* scale MV difference to be quad-pel */

  1367.     dmv_x <<= 1 - s->quarter_sample;

  1368.     dmv_y <<= 1 - s->quarter_sample;

  1369. 

  1370.     wrap = s->b8_stride;

  1371.     xy   = s->block_index[n];

  1372. 

  1373.     if (s->mb_intra) {

  1374.         s->mv[0][n][0] = s->current_picture.f.motion_val[0][xy + v->blocks_off][0] = 0;

  1375.         s->mv[0][n][1] = s->current_picture.f.motion_val[0][xy + v->blocks_off][1] = 0;

  1376.         s->current_picture.f.motion_val[1][xy + v->blocks_off][0] = 0;

  1377.         s->current_picture.f.motion_val[1][xy + v->blocks_off][1] = 0;

  1378.         if (mv1) { /* duplicate motion data for 1-MV block */

  1379.             s->current_picture.f.motion_val[0][xy + 1 + v->blocks_off][0]        = 0;

  1380.             s->current_picture.f.motion_val[0][xy + 1 + v->blocks_off][1]        = 0;

  1381.             s->current_picture.f.motion_val[0][xy + wrap + v->blocks_off][0]     = 0;

  1382.             s->current_picture.f.motion_val[0][xy + wrap + v->blocks_off][1]     = 0;

  1383.             s->current_picture.f.motion_val[0][xy + wrap + 1 + v->blocks_off][0] = 0;

  1384.             s->current_picture.f.motion_val[0][xy + wrap + 1 + v->blocks_off][1] = 0;

  1385.             v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;

  1386.             s->current_picture.f.motion_val[1][xy + 1 + v->blocks_off][0]        = 0;

  1387.             s->current_picture.f.motion_val[1][xy + 1 + v->blocks_off][1]        = 0;

  1388.             s->current_picture.f.motion_val[1][xy + wrap][0]                     = 0;

  1389.             s->current_picture.f.motion_val[1][xy + wrap + v->blocks_off][1]     = 0;

  1390.             s->current_picture.f.motion_val[1][xy + wrap + 1 + v->blocks_off][0] = 0;

  1391.             s->current_picture.f.motion_val[1][xy + wrap + 1 + v->blocks_off][1] = 0;

  1392.         }

  1393.         return;

  1394.     }

  1395. 

  1396.     C = s->current_picture.f.motion_val[dir][xy -    1 + v->blocks_off];

  1397.     A = s->current_picture.f.motion_val[dir][xy - wrap + v->blocks_off];

  1398.     if (mv1) {

  1399.         if (v->field_mode && mixedmv_pic)

  1400.             off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  1401.         else

  1402.             off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;

  1403.     } else {

  1404.         //in 4-MV mode different blocks have different B predictor position

  1405.         switch (n) {

  1406.         case 0:

  1407.             off = (s->mb_x > 0) ? -1 : 1;

  1408.             break;

  1409.         case 1:

  1410.             off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;

  1411.             break;

  1412.         case 2:

  1413.             off = 1;

  1414.             break;

  1415.         case 3:

  1416.             off = -1;

  1417.         }

  1418.     }

  1419.     B = s->current_picture.f.motion_val[dir][xy - wrap + off + v->blocks_off];

  1420. 

  1421.     a_valid = !s->first_slice_line || (n == 2 || n == 3);

  1422.     b_valid = a_valid && (s->mb_width > 1);

  1423.     c_valid = s->mb_x || (n == 1 || n == 3);

  1424.     if (v->field_mode) {

  1425.         a_valid = a_valid && !is_intra[xy - wrap];

  1426.         b_valid = b_valid && !is_intra[xy - wrap + off];

  1427.         c_valid = c_valid && !is_intra[xy - 1];

  1428.     }

  1429. 

  1430.     if (a_valid) {

  1431.         a_f = v->mv_f[dir][xy - wrap + v->blocks_off];

  1432.         num_oppfield  += a_f;

  1433.         num_samefield += 1 - a_f;

  1434.         field_predA[0] = A[0];

  1435.         field_predA[1] = A[1];

  1436.     } else {

  1437.         field_predA[0] = field_predA[1] = 0;

  1438.         a_f = 0;

  1439.     }

  1440.     if (b_valid) {

  1441.         b_f = v->mv_f[dir][xy - wrap + off + v->blocks_off];

  1442.         num_oppfield  += b_f;

  1443.         num_samefield += 1 - b_f;

  1444.         field_predB[0] = B[0];

  1445.         field_predB[1] = B[1];

  1446.     } else {

  1447.         field_predB[0] = field_predB[1] = 0;

  1448.         b_f = 0;

  1449.     }

  1450.     if (c_valid) {

  1451.         c_f = v->mv_f[dir][xy - 1 + v->blocks_off];

  1452.         num_oppfield  += c_f;

  1453.         num_samefield += 1 - c_f;

  1454.         field_predC[0] = C[0];

  1455.         field_predC[1] = C[1];

  1456.     } else {

  1457.         field_predC[0] = field_predC[1] = 0;

  1458.         c_f = 0;

  1459.     }

  1460. 

  1461.     if (v->field_mode) {

  1462.         if (num_samefield <= num_oppfield)

  1463.             opposit = 1 - pred_flag;

  1464.         else

  1465.             opposit = pred_flag;

  1466.     } else

  1467.         opposit = 0;

  1468.     if (opposit) {

  1469.         if (a_valid && !a_f) {

  1470.             field_predA[0] = scaleforopp(v, field_predA[0], 0, dir);

  1471.             field_predA[1] = scaleforopp(v, field_predA[1], 1, dir);

  1472.         }

  1473.         if (b_valid && !b_f) {

  1474.             field_predB[0] = scaleforopp(v, field_predB[0], 0, dir);

  1475.             field_predB[1] = scaleforopp(v, field_predB[1], 1, dir);

  1476.         }

  1477.         if (c_valid && !c_f) {

  1478.             field_predC[0] = scaleforopp(v, field_predC[0], 0, dir);

  1479.             field_predC[1] = scaleforopp(v, field_predC[1], 1, dir);

  1480.         }

  1481.         v->mv_f[dir][xy + v->blocks_off] = 1;

  1482.         v->ref_field_type[dir] = !v->cur_field_type;

  1483.     } else {

  1484.         if (a_valid && a_f) {

  1485.             field_predA[0] = scaleforsame(v, n, field_predA[0], 0, dir);

  1486.             field_predA[1] = scaleforsame(v, n, field_predA[1], 1, dir);

  1487.         }

  1488.         if (b_valid && b_f) {

  1489.             field_predB[0] = scaleforsame(v, n, field_predB[0], 0, dir);

  1490.             field_predB[1] = scaleforsame(v, n, field_predB[1], 1, dir);

  1491.         }

  1492.         if (c_valid && c_f) {

  1493.             field_predC[0] = scaleforsame(v, n, field_predC[0], 0, dir);

  1494.             field_predC[1] = scaleforsame(v, n, field_predC[1], 1, dir);

  1495.         }

  1496.         v->mv_f[dir][xy + v->blocks_off] = 0;

  1497.         v->ref_field_type[dir] = v->cur_field_type;

  1498.     }

  1499. 

  1500.     if (a_valid) {

  1501.         px = field_predA[0];

  1502.         py = field_predA[1];

  1503.     } else if (c_valid) {

  1504.         px = field_predC[0];

  1505.         py = field_predC[1];

  1506.     } else if (b_valid) {

  1507.         px = field_predB[0];

  1508.         py = field_predB[1];

  1509.     } else {

  1510.         px = 0;

  1511.         py = 0;

  1512.     }

  1513. 

  1514.     if (num_samefield + num_oppfield > 1) {

  1515.         px = mid_pred(field_predA[0], field_predB[0], field_predC[0]);

  1516.         py = mid_pred(field_predA[1], field_predB[1], field_predC[1]);

  1517.     }

  1518. 

  1519.     /* Pullback MV as specified in 8.3.5.3.4 */

  1520.     if (!v->field_mode) {

  1521.         int qx, qy, X, Y;

  1522.         qx = (s->mb_x << 6) + ((n == 1 || n == 3) ? 32 : 0);

  1523.         qy = (s->mb_y << 6) + ((n == 2 || n == 3) ? 32 : 0);

  1524.         X  = (s->mb_width  << 6) - 4;

  1525.         Y  = (s->mb_height << 6) - 4;

  1526.         if (mv1) {

  1527.             if (qx + px < -60) px = -60 - qx;

  1528.             if (qy + py < -60) py = -60 - qy;

  1529.         } else {

  1530.             if (qx + px < -28) px = -28 - qx;

  1531.             if (qy + py < -28) py = -28 - qy;

  1532.         }

  1533.         if (qx + px > X) px = X - qx;

  1534.         if (qy + py > Y) py = Y - qy;

  1535.     }

  1536. 

  1537.     if (!v->field_mode || s->pict_type != AV_PICTURE_TYPE_B) {

  1538.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 (also 10.3.5.4.3.5) */

  1539.         hybridmv_thresh = 32;

  1540.         if (a_valid && c_valid) {

  1541.             if (is_intra[xy - wrap])

  1542.                 sum = FFABS(px) + FFABS(py);

  1543.             else

  1544.                 sum = FFABS(px - field_predA[0]) + FFABS(py - field_predA[1]);

  1545.             if (sum > hybridmv_thresh) {

  1546.                 if (get_bits1(&s->gb)) {     // read HYBRIDPRED bit

  1547.                     px = field_predA[0];

  1548.                     py = field_predA[1];

  1549.                 } else {

  1550.                     px = field_predC[0];

  1551.                     py = field_predC[1];

  1552.                 }

  1553.             } else {

  1554.                 if (is_intra[xy - 1])

  1555.                     sum = FFABS(px) + FFABS(py);

  1556.                 else

  1557.                     sum = FFABS(px - field_predC[0]) + FFABS(py - field_predC[1]);

  1558.                 if (sum > hybridmv_thresh) {

  1559.                     if (get_bits1(&s->gb)) {

  1560.                         px = field_predA[0];

  1561.                         py = field_predA[1];

  1562.                     } else {

  1563.                         px = field_predC[0];

  1564.                         py = field_predC[1];

  1565.                     }

  1566.                 }

  1567.             }

  1568.         }

  1569.     }

  1570. 

  1571.     if (v->field_mode && !s->quarter_sample) {

  1572.         r_x <<= 1;

  1573.         r_y <<= 1;

  1574.     }

  1575.     if (v->field_mode && v->numref)

  1576.         r_y >>= 1;

  1577.     if (v->field_mode && v->cur_field_type && v->ref_field_type[dir] == 0)

  1578.         y_bias = 1;

  1579.     /* store MV using signed modulus of MV range defined in 4.11 */

  1580.     s->mv[dir][n][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;

  1581.     s->mv[dir][n][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1] = ((py + dmv_y + r_y - y_bias) & ((r_y << 1) - 1)) - r_y + y_bias;

  1582.     if (mv1) { /* duplicate motion data for 1-MV block */

  1583.         s->current_picture.f.motion_val[dir][xy +    1 +     v->blocks_off][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0];

  1584.         s->current_picture.f.motion_val[dir][xy +    1 +     v->blocks_off][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1];

  1585.         s->current_picture.f.motion_val[dir][xy + wrap +     v->blocks_off][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0];

  1586.         s->current_picture.f.motion_val[dir][xy + wrap +     v->blocks_off][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1];

  1587.         s->current_picture.f.motion_val[dir][xy + wrap + 1 + v->blocks_off][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0];

  1588.         s->current_picture.f.motion_val[dir][xy + wrap + 1 + v->blocks_off][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1];

  1589.         v->mv_f[dir][xy +    1 + v->blocks_off] = v->mv_f[dir][xy +            v->blocks_off];

  1590.         v->mv_f[dir][xy + wrap + v->blocks_off] = v->mv_f[dir][xy + wrap + 1 + v->blocks_off] = v->mv_f[dir][xy + v->blocks_off];

  1591.     }

  1592. }

  1593. 

  1594. /** Predict and set motion vector for interlaced frame picture MBs

  1595.  */

  1596. static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y,

  1597.                                      int mvn, int r_x, int r_y, uint8_t* is_intra)

  1598. {

  1599.     MpegEncContext *s = &v->s;

  1600.     int xy, wrap, off = 0;

  1601.     int A[2], B[2], C[2];

  1602.     int px, py;

  1603.     int a_valid = 0, b_valid = 0, c_valid = 0;

  1604.     int field_a, field_b, field_c; // 0: same, 1: opposit

  1605.     int total_valid, num_samefield, num_oppfield;

  1606.     int pos_c, pos_b, n_adj;

  1607. 

  1608.     wrap = s->b8_stride;

  1609.     xy = s->block_index[n];

  1610. 

  1611.     if (s->mb_intra) {

  1612.         s->mv[0][n][0] = s->current_picture.f.motion_val[0][xy][0] = 0;

  1613.         s->mv[0][n][1] = s->current_picture.f.motion_val[0][xy][1] = 0;

  1614.         s->current_picture.f.motion_val[1][xy][0] = 0;

  1615.         s->current_picture.f.motion_val[1][xy][1] = 0;

  1616.         if (mvn == 1) { /* duplicate motion data for 1-MV block */

  1617.             s->current_picture.f.motion_val[0][xy + 1][0]        = 0;

  1618.             s->current_picture.f.motion_val[0][xy + 1][1]        = 0;

  1619.             s->current_picture.f.motion_val[0][xy + wrap][0]     = 0;

  1620.             s->current_picture.f.motion_val[0][xy + wrap][1]     = 0;

  1621.             s->current_picture.f.motion_val[0][xy + wrap + 1][0] = 0;

  1622.             s->current_picture.f.motion_val[0][xy + wrap + 1][1] = 0;

  1623.             v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;

  1624.             s->current_picture.f.motion_val[1][xy + 1][0]        = 0;

  1625.             s->current_picture.f.motion_val[1][xy + 1][1]        = 0;

  1626.             s->current_picture.f.motion_val[1][xy + wrap][0]     = 0;

  1627.             s->current_picture.f.motion_val[1][xy + wrap][1]     = 0;

  1628.             s->current_picture.f.motion_val[1][xy + wrap + 1][0] = 0;

  1629.             s->current_picture.f.motion_val[1][xy + wrap + 1][1] = 0;

  1630.         }

  1631.         return;

  1632.     }

  1633. 

  1634.     off = ((n == 0) || (n == 1)) ? 1 : -1;

  1635.     /* predict A */

  1636.     if (s->mb_x || (n == 1) || (n == 3)) {

  1637.         if ((v->blk_mv_type[xy]) // current block (MB) has a field MV

  1638.             || (!v->blk_mv_type[xy] && !v->blk_mv_type[xy - 1])) { // or both have frame MV

  1639.             A[0] = s->current_picture.f.motion_val[0][xy - 1][0];

  1640.             A[1] = s->current_picture.f.motion_val[0][xy - 1][1];

  1641.             a_valid = 1;

  1642.         } else { // current block has frame mv and cand. has field MV (so average)

  1643.             A[0] = (s->current_picture.f.motion_val[0][xy - 1][0]

  1644.                     + s->current_picture.f.motion_val[0][xy - 1 + off * wrap][0] + 1) >> 1;

  1645.             A[1] = (s->current_picture.f.motion_val[0][xy - 1][1]

  1646.                     + s->current_picture.f.motion_val[0][xy - 1 + off * wrap][1] + 1) >> 1;

  1647.             a_valid = 1;

  1648.         }

  1649.         if (!(n & 1) && v->is_intra[s->mb_x - 1]) {

  1650.             a_valid = 0;

  1651.             A[0] = A[1] = 0;

  1652.         }

  1653.     } else

  1654.         A[0] = A[1] = 0;

  1655.     /* Predict B and C */

  1656.     B[0] = B[1] = C[0] = C[1] = 0;

  1657.     if (n == 0 || n == 1 || v->blk_mv_type[xy]) {

  1658.         if (!s->first_slice_line) {

  1659.             if (!v->is_intra[s->mb_x - s->mb_stride]) {

  1660.                 b_valid = 1;

  1661.                 n_adj   = n | 2;

  1662.                 pos_b   = s->block_index[n_adj] - 2 * wrap;

  1663.                 if (v->blk_mv_type[pos_b] && v->blk_mv_type[xy]) {

  1664.                     n_adj = (n & 2) | (n & 1);

  1665.                 }

  1666.                 B[0] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap][0];

  1667.                 B[1] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap][1];

  1668.                 if (v->blk_mv_type[pos_b] && !v->blk_mv_type[xy]) {

  1669.                     B[0] = (B[0] + s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap][0] + 1) >> 1;

  1670.                     B[1] = (B[1] + s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap][1] + 1) >> 1;

  1671.                 }

  1672.             }

  1673.             if (s->mb_width > 1) {

  1674.                 if (!v->is_intra[s->mb_x - s->mb_stride + 1]) {

  1675.                     c_valid = 1;

  1676.                     n_adj   = 2;

  1677.                     pos_c   = s->block_index[2] - 2 * wrap + 2;

  1678.                     if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) {

  1679.                         n_adj = n & 2;

  1680.                     }

  1681.                     C[0] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap + 2][0];

  1682.                     C[1] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap + 2][1];

  1683.                     if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) {

  1684.                         C[0] = (1 + C[0] + (s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap + 2][0])) >> 1;

  1685.                         C[1] = (1 + C[1] + (s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap + 2][1])) >> 1;

  1686.                     }

  1687.                     if (s->mb_x == s->mb_width - 1) {

  1688.                         if (!v->is_intra[s->mb_x - s->mb_stride - 1]) {

  1689.                             c_valid = 1;

  1690.                             n_adj   = 3;

  1691.                             pos_c   = s->block_index[3] - 2 * wrap - 2;

  1692.                             if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) {

  1693.                                 n_adj = n | 1;

  1694.                             }

  1695.                             C[0] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap - 2][0];

  1696.                             C[1] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap - 2][1];

  1697.                             if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) {

  1698.                                 C[0] = (1 + C[0] + s->current_picture.f.motion_val[0][s->block_index[1] - 2 * wrap - 2][0]) >> 1;

  1699.                                 C[1] = (1 + C[1] + s->current_picture.f.motion_val[0][s->block_index[1] - 2 * wrap - 2][1]) >> 1;

  1700.                             }

  1701.                         } else

  1702.                             c_valid = 0;

  1703.                     }

  1704.                 }

  1705.             }

  1706.         }

  1707.     } else {

  1708.         pos_b   = s->block_index[1];

  1709.         b_valid = 1;

  1710.         B[0]    = s->current_picture.f.motion_val[0][pos_b][0];

  1711.         B[1]    = s->current_picture.f.motion_val[0][pos_b][1];

  1712.         pos_c   = s->block_index[0];

  1713.         c_valid = 1;

  1714.         C[0]    = s->current_picture.f.motion_val[0][pos_c][0];

  1715.         C[1]    = s->current_picture.f.motion_val[0][pos_c][1];

  1716.     }

  1717. 

  1718.     total_valid = a_valid + b_valid + c_valid;

  1719.     // check if predictor A is out of bounds

  1720.     if (!s->mb_x && !(n == 1 || n == 3)) {

  1721.         A[0] = A[1] = 0;

  1722.     }

  1723.     // check if predictor B is out of bounds

  1724.     if ((s->first_slice_line && v->blk_mv_type[xy]) || (s->first_slice_line && !(n & 2))) {

  1725.         B[0] = B[1] = C[0] = C[1] = 0;

  1726.     }

  1727.     if (!v->blk_mv_type[xy]) {

  1728.         if (s->mb_width == 1) {

  1729.             px = B[0];

  1730.             py = B[1];

  1731.         } else {

  1732.             if (total_valid >= 2) {

  1733.                 px = mid_pred(A[0], B[0], C[0]);

  1734.                 py = mid_pred(A[1], B[1], C[1]);

  1735.             } else if (total_valid) {

  1736.                 if (a_valid) { px = A[0]; py = A[1]; }

  1737.                 if (b_valid) { px = B[0]; py = B[1]; }

  1738.                 if (c_valid) { px = C[0]; py = C[1]; }

  1739.             } else

  1740.                 px = py = 0;

  1741.         }

  1742.     } else {

  1743.         if (a_valid)

  1744.             field_a = (A[1] & 4) ? 1 : 0;

  1745.         else

  1746.             field_a = 0;

  1747.         if (b_valid)

  1748.             field_b = (B[1] & 4) ? 1 : 0;

  1749.         else

  1750.             field_b = 0;

  1751.         if (c_valid)

  1752.             field_c = (C[1] & 4) ? 1 : 0;

  1753.         else

  1754.             field_c = 0;

  1755. 

  1756.         num_oppfield  = field_a + field_b + field_c;

  1757.         num_samefield = total_valid - num_oppfield;

  1758.         if (total_valid == 3) {

  1759.             if ((num_samefield == 3) || (num_oppfield == 3)) {

  1760.                 px = mid_pred(A[0], B[0], C[0]);

  1761.                 py = mid_pred(A[1], B[1], C[1]);

  1762.             } else if (num_samefield >= num_oppfield) {

  1763.                 /* take one MV from same field set depending on priority

  1764.                 the check for B may not be necessary */

  1765.                 px = !field_a ? A[0] : B[0];

  1766.                 py = !field_a ? A[1] : B[1];

  1767.             } else {

  1768.                 px =  field_a ? A[0] : B[0];

  1769.                 py =  field_a ? A[1] : B[1];

  1770.             }

  1771.         } else if (total_valid == 2) {

  1772.             if (num_samefield >= num_oppfield) {

  1773.                 if (!field_a && a_valid) {

  1774.                     px = A[0];

  1775.                     py = A[1];

  1776.                 } else if (!field_b && b_valid) {

  1777.                     px = B[0];

  1778.                     py = B[1];

  1779.                 } else if (c_valid) {

  1780.                     px = C[0];

  1781.                     py = C[1];

  1782.                 } else px = py = 0;

  1783.             } else {

  1784.                 if (field_a && a_valid) {

  1785.                     px = A[0];

  1786.                     py = A[1];

  1787.                 } else if (field_b && b_valid) {

  1788.                     px = B[0];

  1789.                     py = B[1];

  1790.                 } else if (c_valid) {

  1791.                     px = C[0];

  1792.                     py = C[1];

  1793.                 } else px = py = 0;

  1794.             }

  1795.         } else if (total_valid == 1) {

  1796.             px = (a_valid) ? A[0] : ((b_valid) ? B[0] : C[0]);

  1797.             py = (a_valid) ? A[1] : ((b_valid) ? B[1] : C[1]);

  1798.         } else

  1799.             px = py = 0;

  1800.     }

  1801. 

  1802.     /* store MV using signed modulus of MV range defined in 4.11 */

  1803.     s->mv[0][n][0] = s->current_picture.f.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;

  1804.     s->mv[0][n][1] = s->current_picture.f.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;

  1805.     if (mvn == 1) { /* duplicate motion data for 1-MV block */

  1806.         s->current_picture.f.motion_val[0][xy +    1    ][0] = s->current_picture.f.motion_val[0][xy][0];

  1807.         s->current_picture.f.motion_val[0][xy +    1    ][1] = s->current_picture.f.motion_val[0][xy][1];

  1808.         s->current_picture.f.motion_val[0][xy + wrap    ][0] = s->current_picture.f.motion_val[0][xy][0];

  1809.         s->current_picture.f.motion_val[0][xy + wrap    ][1] = s->current_picture.f.motion_val[0][xy][1];

  1810.         s->current_picture.f.motion_val[0][xy + wrap + 1][0] = s->current_picture.f.motion_val[0][xy][0];

  1811.         s->current_picture.f.motion_val[0][xy + wrap + 1][1] = s->current_picture.f.motion_val[0][xy][1];

  1812.     } else if (mvn == 2) { /* duplicate motion data for 2-Field MV block */

  1813.         s->current_picture.f.motion_val[0][xy + 1][0] = s->current_picture.f.motion_val[0][xy][0];

  1814.         s->current_picture.f.motion_val[0][xy + 1][1] = s->current_picture.f.motion_val[0][xy][1];

  1815.         s->mv[0][n + 1][0] = s->mv[0][n][0];

  1816.         s->mv[0][n + 1][1] = s->mv[0][n][1];

  1817.     }

  1818. }

  1819. 

  1820. /** Motion compensation for direct or interpolated blocks in B-frames

  1821.  */

  1822. static void vc1_interp_mc(VC1Context *v)

  1823. {

  1824.     MpegEncContext *s = &v->s;

  1825.     DSPContext *dsp = &v->s.dsp;

  1826.     uint8_t *srcY, *srcU, *srcV;

  1827.     int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;

  1828.     int off, off_uv;

  1829.     int v_edge_pos = s->v_edge_pos >> v->field_mode;

  1830. 

  1831.     if (!v->field_mode && !v->s.next_picture.f.data[0])

  1832.         return;

  1833. 

  1834.     mx   = s->mv[1][0][0];

  1835.     my   = s->mv[1][0][1];

  1836.     uvmx = (mx + ((mx & 3) == 3)) >> 1;

  1837.     uvmy = (my + ((my & 3) == 3)) >> 1;

  1838.     if (v->field_mode) {

  1839.         if (v->cur_field_type != v->ref_field_type[1])

  1840.             my   = my   - 2 + 4 * v->cur_field_type;

  1841.             uvmy = uvmy - 2 + 4 * v->cur_field_type;

  1842.     }

  1843.     if (v->fastuvmc) {

  1844.         uvmx = uvmx + ((uvmx < 0) ? -(uvmx & 1) : (uvmx & 1));

  1845.         uvmy = uvmy + ((uvmy < 0) ? -(uvmy & 1) : (uvmy & 1));

  1846.     }

  1847.     srcY = s->next_picture.f.data[0];

  1848.     srcU = s->next_picture.f.data[1];

  1849.     srcV = s->next_picture.f.data[2];

  1850. 

  1851.     src_x   = s->mb_x * 16 + (mx   >> 2);

  1852.     src_y   = s->mb_y * 16 + (my   >> 2);

  1853.     uvsrc_x = s->mb_x *  8 + (uvmx >> 2);

  1854.     uvsrc_y = s->mb_y *  8 + (uvmy >> 2);

  1855. 

  1856.     if (v->profile != PROFILE_ADVANCED) {

  1857.         src_x   = av_clip(  src_x, -16, s->mb_width  * 16);

  1858.         src_y   = av_clip(  src_y, -16, s->mb_height * 16);

  1859.         uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);

  1860.         uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);

  1861.     } else {

  1862.         src_x   = av_clip(  src_x, -17, s->avctx->coded_width);

  1863.         src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);

  1864.         uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);

  1865.         uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);

  1866.     }

  1867. 

  1868.     srcY += src_y   * s->linesize   + src_x;

  1869.     srcU += uvsrc_y * s->uvlinesize + uvsrc_x;

  1870.     srcV += uvsrc_y * s->uvlinesize + uvsrc_x;

  1871. 

  1872.     if (v->field_mode && v->ref_field_type[1]) {

  1873.         srcY += s->current_picture_ptr->f.linesize[0];

  1874.         srcU += s->current_picture_ptr->f.linesize[1];

  1875.         srcV += s->current_picture_ptr->f.linesize[2];

  1876.     }

  1877. 

  1878.     /* for grayscale we should not try to read from unknown area */

  1879.     if (s->flags & CODEC_FLAG_GRAY) {

  1880.         srcU = s->edge_emu_buffer + 18 * s->linesize;

  1881.         srcV = s->edge_emu_buffer + 18 * s->linesize;

  1882.     }

  1883. 

  1884.     if (v->rangeredfrm || s->h_edge_pos < 22 || v_edge_pos < 22

  1885.         || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx & 3) - 16 - s->mspel * 3

  1886.         || (unsigned)(src_y - s->mspel) > v_edge_pos    - (my & 3) - 16 - s->mspel * 3) {

  1887.         uint8_t *uvbuf = s->edge_emu_buffer + 19 * s->linesize;

  1888. 

  1889.         srcY -= s->mspel * (1 + s->linesize);

  1890.         s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize,

  1891.                                 17 + s->mspel * 2, 17 + s->mspel * 2,

  1892.                                 src_x - s->mspel, src_y - s->mspel,

  1893.                                 s->h_edge_pos, v_edge_pos);

  1894.         srcY = s->edge_emu_buffer;

  1895.         s->dsp.emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8 + 1, 8 + 1,

  1896.                                 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);

  1897.         s->dsp.emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8 + 1, 8 + 1,

  1898.                                 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);

  1899.         srcU = uvbuf;

  1900.         srcV = uvbuf + 16;

  1901.         /* if we deal with range reduction we need to scale source blocks */

  1902.         if (v->rangeredfrm) {

  1903.             int i, j;

  1904.             uint8_t *src, *src2;

  1905. 

  1906.             src = srcY;

  1907.             for (j = 0; j < 17 + s->mspel * 2; j++) {

  1908.                 for (i = 0; i < 17 + s->mspel * 2; i++)

  1909.                     src[i] = ((src[i] - 128) >> 1) + 128;

  1910.                 src += s->linesize;

  1911.             }

  1912.             src = srcU;

  1913.             src2 = srcV;

  1914.             for (j = 0; j < 9; j++) {

  1915.                 for (i = 0; i < 9; i++) {

  1916.                     src[i]  = ((src[i]  - 128) >> 1) + 128;

  1917.                     src2[i] = ((src2[i] - 128) >> 1) + 128;

  1918.                 }

  1919.                 src  += s->uvlinesize;

  1920.                 src2 += s->uvlinesize;

  1921.             }

  1922.         }

  1923.         srcY += s->mspel * (1 + s->linesize);

  1924.     }

  1925. 

  1926.     if (v->field_mode && v->second_field) {

  1927.         off    = s->current_picture_ptr->f.linesize[0];

  1928.         off_uv = s->current_picture_ptr->f.linesize[1];

  1929.     } else {

  1930.         off    = 0;

  1931.         off_uv = 0;

  1932.     }

  1933. 

  1934.     if (s->mspel) {

  1935.         dxy = ((my & 3) << 2) | (mx & 3);

  1936.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off    , srcY    , s->linesize, v->rnd);

  1937.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8, srcY + 8, s->linesize, v->rnd);

  1938.         srcY += s->linesize * 8;

  1939.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize    , srcY    , s->linesize, v->rnd);

  1940.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);

  1941.     } else { // hpel mc

  1942.         dxy = (my & 2) | ((mx & 2) >> 1);

  1943. 

  1944.         if (!v->rnd)

  1945.             dsp->avg_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16);

  1946.         else

  1947.             dsp->avg_no_rnd_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16);

  1948.     }

  1949. 

  1950.     if (s->flags & CODEC_FLAG_GRAY) return;

  1951.     /* Chroma MC always uses qpel blilinear */

  1952.     uvmx = (uvmx & 3) << 1;

  1953.     uvmy = (uvmy & 3) << 1;

  1954.     if (!v->rnd) {

  1955.         dsp->avg_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy);

  1956.         dsp->avg_h264_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy);

  1957.     } else {

  1958.         v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy);

  1959.         v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy);

  1960.     }

  1961. }

  1962. 

  1963. static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)

  1964. {

  1965.     int n = bfrac;

  1966. 

  1967. #if B_FRACTION_DEN==256

  1968.     if (inv)

  1969.         n -= 256;

  1970.     if (!qs)

  1971.         return 2 * ((value * n + 255) >> 9);

  1972.     return (value * n + 128) >> 8;

  1973. #else

  1974.     if (inv)

  1975.         n -= B_FRACTION_DEN;

  1976.     if (!qs)

  1977.         return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));

  1978.     return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;

  1979. #endif

  1980. }

  1981. 

  1982. static av_always_inline int scale_mv_intfi(int value, int bfrac, int inv,

  1983.                                            int qs, int qs_last)

  1984. {

  1985.     int n = bfrac;

  1986. 

  1987.     if (inv)

  1988.         n -= 256;

  1989.     n <<= !qs_last;

  1990.     if (!qs)

  1991.         return (value * n + 255) >> 9;

  1992.     else

  1993.         return (value * n + 128) >> 8;

  1994. }

  1995. 

  1996. /** Reconstruct motion vector for B-frame and do motion compensation

  1997.  */

  1998. static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2],

  1999.                             int direct, int mode)

  2000. {

  2001.     if (v->use_ic) {

  2002.         v->mv_mode2 = v->mv_mode;

  2003.         v->mv_mode  = MV_PMODE_INTENSITY_COMP;

  2004.     }

  2005.     if (direct) {

  2006.         vc1_mc_1mv(v, 0);

  2007.         vc1_interp_mc(v);

  2008.         if (v->use_ic)

  2009.             v->mv_mode = v->mv_mode2;

  2010.         return;

  2011.     }

  2012.     if (mode == BMV_TYPE_INTERPOLATED) {

  2013.         vc1_mc_1mv(v, 0);

  2014.         vc1_interp_mc(v);

  2015.         if (v->use_ic)

  2016.             v->mv_mode = v->mv_mode2;

  2017.         return;

  2018.     }

  2019. 

  2020.     if (v->use_ic && (mode == BMV_TYPE_BACKWARD))

  2021.         v->mv_mode = v->mv_mode2;

  2022.     vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));

  2023.     if (v->use_ic)

  2024.         v->mv_mode = v->mv_mode2;

  2025. }

  2026. 

  2027. static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2],

  2028.                                  int direct, int mvtype)

  2029. {

  2030.     MpegEncContext *s = &v->s;

  2031.     int xy, wrap, off = 0;

  2032.     int16_t *A, *B, *C;

  2033.     int px, py;

  2034.     int sum;

  2035.     int r_x, r_y;

  2036.     const uint8_t *is_intra = v->mb_type[0];

  2037. 

  2038.     r_x = v->range_x;

  2039.     r_y = v->range_y;

  2040.     /* scale MV difference to be quad-pel */

  2041.     dmv_x[0] <<= 1 - s->quarter_sample;

  2042.     dmv_y[0] <<= 1 - s->quarter_sample;

  2043.     dmv_x[1] <<= 1 - s->quarter_sample;

  2044.     dmv_y[1] <<= 1 - s->quarter_sample;

  2045. 

  2046.     wrap = s->b8_stride;

  2047.     xy = s->block_index[0];

  2048. 

  2049.     if (s->mb_intra) {

  2050.         s->current_picture.f.motion_val[0][xy + v->blocks_off][0] =

  2051.         s->current_picture.f.motion_val[0][xy + v->blocks_off][1] =

  2052.         s->current_picture.f.motion_val[1][xy + v->blocks_off][0] =

  2053.         s->current_picture.f.motion_val[1][xy + v->blocks_off][1] = 0;

  2054.         return;

  2055.     }

  2056.     if (!v->field_mode) {

  2057.         s->mv[0][0][0] = scale_mv(s->next_picture.f.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);

  2058.         s->mv[0][0][1] = scale_mv(s->next_picture.f.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);

  2059.         s->mv[1][0][0] = scale_mv(s->next_picture.f.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);

  2060.         s->mv[1][0][1] = scale_mv(s->next_picture.f.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);

  2061. 

  2062.         /* Pullback predicted motion vectors as specified in 8.4.5.4 */

  2063.         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));

  2064.         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));

  2065.         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));

  2066.         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));

  2067.     }

  2068.     if (direct) {

  2069.         s->current_picture.f.motion_val[0][xy + v->blocks_off][0] = s->mv[0][0][0];

  2070.         s->current_picture.f.motion_val[0][xy + v->blocks_off][1] = s->mv[0][0][1];

  2071.         s->current_picture.f.motion_val[1][xy + v->blocks_off][0] = s->mv[1][0][0];

  2072.         s->current_picture.f.motion_val[1][xy + v->blocks_off][1] = s->mv[1][0][1];

  2073.         return;

  2074.     }

  2075. 

  2076.     if ((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {

  2077.         C   = s->current_picture.f.motion_val[0][xy - 2];

  2078.         A   = s->current_picture.f.motion_val[0][xy - wrap * 2];

  2079.         off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  2080.         B   = s->current_picture.f.motion_val[0][xy - wrap * 2 + off];

  2081. 

  2082.         if (!s->mb_x) C[0] = C[1] = 0;

  2083.         if (!s->first_slice_line) { // predictor A is not out of bounds

  2084.             if (s->mb_width == 1) {

  2085.                 px = A[0];

  2086.                 py = A[1];

  2087.             } else {

  2088.                 px = mid_pred(A[0], B[0], C[0]);

  2089.                 py = mid_pred(A[1], B[1], C[1]);

  2090.             }

  2091.         } else if (s->mb_x) { // predictor C is not out of bounds

  2092.             px = C[0];

  2093.             py = C[1];

  2094.         } else {

  2095.             px = py = 0;

  2096.         }

  2097.         /* Pullback MV as specified in 8.3.5.3.4 */

  2098.         {

  2099.             int qx, qy, X, Y;

  2100.             if (v->profile < PROFILE_ADVANCED) {

  2101.                 qx = (s->mb_x << 5);

  2102.                 qy = (s->mb_y << 5);

  2103.                 X  = (s->mb_width  << 5) - 4;

  2104.                 Y  = (s->mb_height << 5) - 4;

  2105.                 if (qx + px < -28) px = -28 - qx;

  2106.                 if (qy + py < -28) py = -28 - qy;

  2107.                 if (qx + px > X) px = X - qx;

  2108.                 if (qy + py > Y) py = Y - qy;

  2109.             } else {

  2110.                 qx = (s->mb_x << 6);

  2111.                 qy = (s->mb_y << 6);

  2112.                 X  = (s->mb_width  << 6) - 4;

  2113.                 Y  = (s->mb_height << 6) - 4;

  2114.                 if (qx + px < -60) px = -60 - qx;

  2115.                 if (qy + py < -60) py = -60 - qy;

  2116.                 if (qx + px > X) px = X - qx;

  2117.                 if (qy + py > Y) py = Y - qy;

  2118.             }

  2119.         }

  2120.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 */

  2121.         if (0 && !s->first_slice_line && s->mb_x) {

  2122.             if (is_intra[xy - wrap])

  2123.                 sum = FFABS(px) + FFABS(py);

  2124.             else

  2125.                 sum = FFABS(px - A[0]) + FFABS(py - A[1]);

  2126.             if (sum > 32) {

  2127.                 if (get_bits1(&s->gb)) {

  2128.                     px = A[0];

  2129.                     py = A[1];

  2130.                 } else {

  2131.                     px = C[0];

  2132.                     py = C[1];

  2133.                 }

  2134.             } else {

  2135.                 if (is_intra[xy - 2])

  2136.                     sum = FFABS(px) + FFABS(py);

  2137.                 else

  2138.                     sum = FFABS(px - C[0]) + FFABS(py - C[1]);

  2139.                 if (sum > 32) {

  2140.                     if (get_bits1(&s->gb)) {

  2141.                         px = A[0];

  2142.                         py = A[1];

  2143.                     } else {

  2144.                         px = C[0];

  2145.                         py = C[1];

  2146.                     }

  2147.                 }

  2148.             }

  2149.         }

  2150.         /* store MV using signed modulus of MV range defined in 4.11 */

  2151.         s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;

  2152.         s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;

  2153.     }

  2154.     if ((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {

  2155.         C   = s->current_picture.f.motion_val[1][xy - 2];

  2156.         A   = s->current_picture.f.motion_val[1][xy - wrap * 2];

  2157.         off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  2158.         B   = s->current_picture.f.motion_val[1][xy - wrap * 2 + off];

  2159. 

  2160.         if (!s->mb_x)

  2161.             C[0] = C[1] = 0;

  2162.         if (!s->first_slice_line) { // predictor A is not out of bounds

  2163.             if (s->mb_width == 1) {

  2164.                 px = A[0];

  2165.                 py = A[1];

  2166.             } else {

  2167.                 px = mid_pred(A[0], B[0], C[0]);

  2168.                 py = mid_pred(A[1], B[1], C[1]);

  2169.             }

  2170.         } else if (s->mb_x) { // predictor C is not out of bounds

  2171.             px = C[0];

  2172.             py = C[1];

  2173.         } else {

  2174.             px = py = 0;

  2175.         }

  2176.         /* Pullback MV as specified in 8.3.5.3.4 */

  2177.         {

  2178.             int qx, qy, X, Y;

  2179.             if (v->profile < PROFILE_ADVANCED) {

  2180.                 qx = (s->mb_x << 5);

  2181.                 qy = (s->mb_y << 5);

  2182.                 X  = (s->mb_width  << 5) - 4;

  2183.                 Y  = (s->mb_height << 5) - 4;

  2184.                 if (qx + px < -28) px = -28 - qx;

  2185.                 if (qy + py < -28) py = -28 - qy;

  2186.                 if (qx + px > X) px = X - qx;

  2187.                 if (qy + py > Y) py = Y - qy;

  2188.             } else {

  2189.                 qx = (s->mb_x << 6);

  2190.                 qy = (s->mb_y << 6);

  2191.                 X  = (s->mb_width  << 6) - 4;

  2192.                 Y  = (s->mb_height << 6) - 4;

  2193.                 if (qx + px < -60) px = -60 - qx;

  2194.                 if (qy + py < -60) py = -60 - qy;

  2195.                 if (qx + px > X) px = X - qx;

  2196.                 if (qy + py > Y) py = Y - qy;

  2197.             }

  2198.         }

  2199.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 */

  2200.         if (0 && !s->first_slice_line && s->mb_x) {

  2201.             if (is_intra[xy - wrap])

  2202.                 sum = FFABS(px) + FFABS(py);

  2203.             else

  2204.                 sum = FFABS(px - A[0]) + FFABS(py - A[1]);

  2205.             if (sum > 32) {

  2206.                 if (get_bits1(&s->gb)) {

  2207.                     px = A[0];

  2208.                     py = A[1];

  2209.                 } else {

  2210.                     px = C[0];

  2211.                     py = C[1];

  2212.                 }

  2213.             } else {

  2214.                 if (is_intra[xy - 2])

  2215.                     sum = FFABS(px) + FFABS(py);

  2216.                 else

  2217.                     sum = FFABS(px - C[0]) + FFABS(py - C[1]);

  2218.                 if (sum > 32) {

  2219.                     if (get_bits1(&s->gb)) {

  2220.                         px = A[0];

  2221.                         py = A[1];

  2222.                     } else {

  2223.                         px = C[0];

  2224.                         py = C[1];

  2225.                     }

  2226.                 }

  2227.             }

  2228.         }

  2229.         /* store MV using signed modulus of MV range defined in 4.11 */

  2230. 

  2231.         s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;

  2232.         s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;

  2233.     }

  2234.     s->current_picture.f.motion_val[0][xy][0] = s->mv[0][0][0];

  2235.     s->current_picture.f.motion_val[0][xy][1] = s->mv[0][0][1];

  2236.     s->current_picture.f.motion_val[1][xy][0] = s->mv[1][0][0];

  2237.     s->current_picture.f.motion_val[1][xy][1] = s->mv[1][0][1];

  2238. }

  2239. 

  2240. static inline void vc1_pred_b_mv_intfi(VC1Context *v, int n, int *dmv_x, int *dmv_y, int mv1, int *pred_flag)

  2241. {

  2242.     int dir = (v->bmvtype == BMV_TYPE_BACKWARD) ? 1 : 0;

  2243.     MpegEncContext *s = &v->s;

  2244.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2245. 

  2246.     if (v->bmvtype == BMV_TYPE_DIRECT) {

  2247.         int total_opp, k, f;

  2248.         if (s->next_picture.f.mb_type[mb_pos + v->mb_off] != MB_TYPE_INTRA) {

  2249.             s->mv[0][0][0] = scale_mv_intfi(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0],

  2250.                                             v->bfraction, 0, s->quarter_sample, v->qs_last);

  2251.             s->mv[0][0][1] = scale_mv_intfi(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1],

  2252.                                             v->bfraction, 0, s->quarter_sample, v->qs_last);

  2253.             s->mv[1][0][0] = scale_mv_intfi(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0],

  2254.                                             v->bfraction, 1, s->quarter_sample, v->qs_last);

  2255.             s->mv[1][0][1] = scale_mv_intfi(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1],

  2256.                                             v->bfraction, 1, s->quarter_sample, v->qs_last);

  2257. 

  2258.             total_opp = v->mv_f_next[0][s->block_index[0] + v->blocks_off]

  2259.                       + v->mv_f_next[0][s->block_index[1] + v->blocks_off]

  2260.                       + v->mv_f_next[0][s->block_index[2] + v->blocks_off]

  2261.                       + v->mv_f_next[0][s->block_index[3] + v->blocks_off];

  2262.             f = (total_opp > 2) ? 1 : 0;

  2263.         } else {

  2264.             s->mv[0][0][0] = s->mv[0][0][1] = 0;

  2265.             s->mv[1][0][0] = s->mv[1][0][1] = 0;

  2266.             f = 0;

  2267.         }

  2268.         v->ref_field_type[0] = v->ref_field_type[1] = v->cur_field_type ^ f;

  2269.         for (k = 0; k < 4; k++) {

  2270.             s->current_picture.f.motion_val[0][s->block_index[k] + v->blocks_off][0] = s->mv[0][0][0];

  2271.             s->current_picture.f.motion_val[0][s->block_index[k] + v->blocks_off][1] = s->mv[0][0][1];

  2272.             s->current_picture.f.motion_val[1][s->block_index[k] + v->blocks_off][0] = s->mv[1][0][0];

  2273.             s->current_picture.f.motion_val[1][s->block_index[k] + v->blocks_off][1] = s->mv[1][0][1];

  2274.             v->mv_f[0][s->block_index[k] + v->blocks_off] = f;

  2275.             v->mv_f[1][s->block_index[k] + v->blocks_off] = f;

  2276.         }

  2277.         return;

  2278.     }

  2279.     if (v->bmvtype == BMV_TYPE_INTERPOLATED) {

  2280.         vc1_pred_mv(v, 0, dmv_x[0], dmv_y[0],   1, v->range_x, v->range_y, v->mb_type[0], pred_flag[0], 0);

  2281.         vc1_pred_mv(v, 0, dmv_x[1], dmv_y[1],   1, v->range_x, v->range_y, v->mb_type[0], pred_flag[1], 1);

  2282.         return;

  2283.     }

  2284.     if (dir) { // backward

  2285.         vc1_pred_mv(v, n, dmv_x[1], dmv_y[1], mv1, v->range_x, v->range_y, v->mb_type[0], pred_flag[1], 1);

  2286.         if (n == 3 || mv1) {

  2287.             vc1_pred_mv(v, 0, dmv_x[0], dmv_y[0],   1, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  2288.         }

  2289.     } else { // forward

  2290.         vc1_pred_mv(v, n, dmv_x[0], dmv_y[0], mv1, v->range_x, v->range_y, v->mb_type[0], pred_flag[0], 0);

  2291.         if (n == 3 || mv1) {

  2292.             vc1_pred_mv(v, 0, dmv_x[1], dmv_y[1],   1, v->range_x, v->range_y, v->mb_type[0], 0, 1);

  2293.         }

  2294.     }

  2295. }

  2296. 

  2297. /** Get predicted DC value for I-frames only

  2298.  * prediction dir: left=0, top=1

  2299.  * @param s MpegEncContext

  2300.  * @param overlap flag indicating that overlap filtering is used

  2301.  * @param pq integer part of picture quantizer

  2302.  * @param[in] n block index in the current MB

  2303.  * @param dc_val_ptr Pointer to DC predictor

  2304.  * @param dir_ptr Prediction direction for use in AC prediction

  2305.  */

  2306. static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,

  2307.                                 int16_t **dc_val_ptr, int *dir_ptr)

  2308. {

  2309.     int a, b, c, wrap, pred, scale;

  2310.     int16_t *dc_val;

  2311.     static const uint16_t dcpred[32] = {

  2312.         -1, 1024,  512,  341,  256,  205,  171,  146,  128,

  2313.              114,  102,   93,   85,   79,   73,   68,   64,

  2314.               60,   57,   54,   51,   49,   47,   45,   43,

  2315.               41,   39,   38,   37,   35,   34,   33

  2316.     };

  2317. 

  2318.     /* find prediction - wmv3_dc_scale always used here in fact */

  2319.     if (n < 4) scale = s->y_dc_scale;

  2320.     else       scale = s->c_dc_scale;

  2321. 

  2322.     wrap   = s->block_wrap[n];

  2323.     dc_val = s->dc_val[0] + s->block_index[n];

  2324. 

  2325.     /* B A

  2326.      * C X

  2327.      */

  2328.     c = dc_val[ - 1];

  2329.     b = dc_val[ - 1 - wrap];

  2330.     a = dc_val[ - wrap];

  2331. 

  2332.     if (pq < 9 || !overlap) {

  2333.         /* Set outer values */

  2334.         if (s->first_slice_line && (n != 2 && n != 3))

  2335.             b = a = dcpred[scale];

  2336.         if (s->mb_x == 0 && (n != 1 && n != 3))

  2337.             b = c = dcpred[scale];

  2338.     } else {

  2339.         /* Set outer values */

  2340.         if (s->first_slice_line && (n != 2 && n != 3))

  2341.             b = a = 0;

  2342.         if (s->mb_x == 0 && (n != 1 && n != 3))

  2343.             b = c = 0;

  2344.     }

  2345. 

  2346.     if (abs(a - b) <= abs(b - c)) {

  2347.         pred     = c;

  2348.         *dir_ptr = 1; // left

  2349.     } else {

  2350.         pred     = a;

  2351.         *dir_ptr = 0; // top

  2352.     }

  2353. 

  2354.     /* update predictor */

  2355.     *dc_val_ptr = &dc_val[0];

  2356.     return pred;

  2357. }

  2358. 

  2359. 

  2360. /** Get predicted DC value

  2361.  * prediction dir: left=0, top=1

  2362.  * @param s MpegEncContext

  2363.  * @param overlap flag indicating that overlap filtering is used

  2364.  * @param pq integer part of picture quantizer

  2365.  * @param[in] n block index in the current MB

  2366.  * @param a_avail flag indicating top block availability

  2367.  * @param c_avail flag indicating left block availability

  2368.  * @param dc_val_ptr Pointer to DC predictor

  2369.  * @param dir_ptr Prediction direction for use in AC prediction

  2370.  */

  2371. static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,

  2372.                               int a_avail, int c_avail,

  2373.                               int16_t **dc_val_ptr, int *dir_ptr)

  2374. {

  2375.     int a, b, c, wrap, pred;

  2376.     int16_t *dc_val;

  2377.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2378.     int q1, q2 = 0;

  2379.     int dqscale_index;

  2380. 

  2381.     wrap = s->block_wrap[n];

  2382.     dc_val = s->dc_val[0] + s->block_index[n];

  2383. 

  2384.     /* B A

  2385.      * C X

  2386.      */

  2387.     c = dc_val[ - 1];

  2388.     b = dc_val[ - 1 - wrap];

  2389.     a = dc_val[ - wrap];

  2390.     /* scale predictors if needed */

  2391.     q1 = s->current_picture.f.qscale_table[mb_pos];

  2392.     dqscale_index = s->y_dc_scale_table[q1] - 1;

  2393.     if (dqscale_index < 0)

  2394.         return 0;

  2395.     if (c_avail && (n != 1 && n != 3)) {

  2396.         q2 = s->current_picture.f.qscale_table[mb_pos - 1];

  2397.         if (q2 && q2 != q1)

  2398.             c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;

  2399.     }

  2400.     if (a_avail && (n != 2 && n != 3)) {

  2401.         q2 = s->current_picture.f.qscale_table[mb_pos - s->mb_stride];

  2402.         if (q2 && q2 != q1)

  2403.             a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;

  2404.     }

  2405.     if (a_avail && c_avail && (n != 3)) {

  2406.         int off = mb_pos;

  2407.         if (n != 1)

  2408.             off--;

  2409.         if (n != 2)

  2410.             off -= s->mb_stride;

  2411.         q2 = s->current_picture.f.qscale_table[off];

  2412.         if (q2 && q2 != q1)

  2413.             b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;

  2414.     }

  2415. 

  2416.     if (a_avail && c_avail) {

  2417.         if (abs(a - b) <= abs(b - c)) {

  2418.             pred     = c;

  2419.             *dir_ptr = 1; // left

  2420.         } else {

  2421.             pred     = a;

  2422.             *dir_ptr = 0; // top

  2423.         }

  2424.     } else if (a_avail) {

  2425.         pred     = a;

  2426.         *dir_ptr = 0; // top

  2427.     } else if (c_avail) {

  2428.         pred     = c;

  2429.         *dir_ptr = 1; // left

  2430.     } else {

  2431.         pred     = 0;

  2432.         *dir_ptr = 1; // left

  2433.     }

  2434. 

  2435.     /* update predictor */

  2436.     *dc_val_ptr = &dc_val[0];

  2437.     return pred;

  2438. }

  2439. 

  2440. /** @} */ // Block group

  2441. 

  2442. /**

  2443.  * @name VC1 Macroblock-level functions in Simple/Main Profiles

  2444.  * @see 7.1.4, p91 and 8.1.1.7, p(1)04

  2445.  * @{

  2446.  */

  2447. 

  2448. static inline int vc1_coded_block_pred(MpegEncContext * s, int n,

  2449.                                        uint8_t **coded_block_ptr)

  2450. {

  2451.     int xy, wrap, pred, a, b, c;

  2452. 

  2453.     xy   = s->block_index[n];

  2454.     wrap = s->b8_stride;

  2455. 

  2456.     /* B C

  2457.      * A X

  2458.      */

  2459.     a = s->coded_block[xy - 1       ];

  2460.     b = s->coded_block[xy - 1 - wrap];

  2461.     c = s->coded_block[xy     - wrap];

  2462. 

  2463.     if (b == c) {

  2464.         pred = a;

  2465.     } else {

  2466.         pred = c;

  2467.     }

  2468. 

  2469.     /* store value */

  2470.     *coded_block_ptr = &s->coded_block[xy];

  2471. 

  2472.     return pred;

  2473. }

  2474. 

  2475. /**

  2476.  * Decode one AC coefficient

  2477.  * @param v The VC1 context

  2478.  * @param last Last coefficient

  2479.  * @param skip How much zero coefficients to skip

  2480.  * @param value Decoded AC coefficient value

  2481.  * @param codingset set of VLC to decode data

  2482.  * @see 8.1.3.4

  2483.  */

  2484. static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip,

  2485.                                 int *value, int codingset)

  2486. {

  2487.     GetBitContext *gb = &v->s.gb;

  2488.     int index, escape, run = 0, level = 0, lst = 0;

  2489. 

  2490.     index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);

  2491.     if (index != ff_vc1_ac_sizes[codingset] - 1) {

  2492.         run   = vc1_index_decode_table[codingset][index][0];

  2493.         level = vc1_index_decode_table[codingset][index][1];

  2494.         lst   = index >= vc1_last_decode_table[codingset] || get_bits_left(gb) < 0;

  2495.         if (get_bits1(gb))

  2496.             level = -level;

  2497.     } else {

  2498.         escape = decode210(gb);

  2499.         if (escape != 2) {

  2500.             index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);

  2501.             run   = vc1_index_decode_table[codingset][index][0];

  2502.             level = vc1_index_decode_table[codingset][index][1];

  2503.             lst   = index >= vc1_last_decode_table[codingset];

  2504.             if (escape == 0) {

  2505.                 if (lst)

  2506.                     level += vc1_last_delta_level_table[codingset][run];

  2507.                 else

  2508.                     level += vc1_delta_level_table[codingset][run];

  2509.             } else {

  2510.                 if (lst)

  2511.                     run += vc1_last_delta_run_table[codingset][level] + 1;

  2512.                 else

  2513.                     run += vc1_delta_run_table[codingset][level] + 1;

  2514.             }

  2515.             if (get_bits1(gb))

  2516.                 level = -level;

  2517.         } else {

  2518.             int sign;

  2519.             lst = get_bits1(gb);

  2520.             if (v->s.esc3_level_length == 0) {

  2521.                 if (v->pq < 8 || v->dquantfrm) { // table 59

  2522.                     v->s.esc3_level_length = get_bits(gb, 3);

  2523.                     if (!v->s.esc3_level_length)

  2524.                         v->s.esc3_level_length = get_bits(gb, 2) + 8;

  2525.                 } else { // table 60

  2526.                     v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;

  2527.                 }

  2528.                 v->s.esc3_run_length = 3 + get_bits(gb, 2);

  2529.             }

  2530.             run   = get_bits(gb, v->s.esc3_run_length);

  2531.             sign  = get_bits1(gb);

  2532.             level = get_bits(gb, v->s.esc3_level_length);

  2533.             if (sign)

  2534.                 level = -level;

  2535.         }

  2536.     }

  2537. 

  2538.     *last  = lst;

  2539.     *skip  = run;

  2540.     *value = level;

  2541. }

  2542. 

  2543. /** Decode intra block in intra frames - should be faster than decode_intra_block

  2544.  * @param v VC1Context

  2545.  * @param block block to decode

  2546.  * @param[in] n subblock index

  2547.  * @param coded are AC coeffs present or not

  2548.  * @param codingset set of VLC to decode data

  2549.  */

  2550. static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n,

  2551.                               int coded, int codingset)

  2552. {

  2553.     GetBitContext *gb = &v->s.gb;

  2554.     MpegEncContext *s = &v->s;

  2555.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2556.     int i;

  2557.     int16_t *dc_val;

  2558.     int16_t *ac_val, *ac_val2;

  2559.     int dcdiff;

  2560. 

  2561.     /* Get DC differential */

  2562.     if (n < 4) {

  2563.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2564.     } else {

  2565.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2566.     }

  2567.     if (dcdiff < 0) {

  2568.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2569.         return -1;

  2570.     }

  2571.     if (dcdiff) {

  2572.         if (dcdiff == 119 /* ESC index value */) {

  2573.             /* TODO: Optimize */

  2574.             if (v->pq == 1)      dcdiff = get_bits(gb, 10);

  2575.             else if (v->pq == 2) dcdiff = get_bits(gb, 9);

  2576.             else                 dcdiff = get_bits(gb, 8);

  2577.         } else {

  2578.             if (v->pq == 1)

  2579.                 dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;

  2580.             else if (v->pq == 2)

  2581.                 dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;

  2582.         }

  2583.         if (get_bits1(gb))

  2584.             dcdiff = -dcdiff;

  2585.     }

  2586. 

  2587.     /* Prediction */

  2588.     dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);

  2589.     *dc_val = dcdiff;

  2590. 

  2591.     /* Store the quantized DC coeff, used for prediction */

  2592.     if (n < 4) {

  2593.         block[0] = dcdiff * s->y_dc_scale;

  2594.     } else {

  2595.         block[0] = dcdiff * s->c_dc_scale;

  2596.     }

  2597.     /* Skip ? */

  2598.     if (!coded) {

  2599.         goto not_coded;

  2600.     }

  2601. 

  2602.     // AC Decoding

  2603.     i = 1;

  2604. 

  2605.     {

  2606.         int last = 0, skip, value;

  2607.         const uint8_t *zz_table;

  2608.         int scale;

  2609.         int k;

  2610. 

  2611.         scale = v->pq * 2 + v->halfpq;

  2612. 

  2613.         if (v->s.ac_pred) {

  2614.             if (!dc_pred_dir)

  2615.                 zz_table = v->zz_8x8[2];

  2616.             else

  2617.                 zz_table = v->zz_8x8[3];

  2618.         } else

  2619.             zz_table = v->zz_8x8[1];

  2620. 

  2621.         ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;

  2622.         ac_val2 = ac_val;

  2623.         if (dc_pred_dir) // left

  2624.             ac_val -= 16;

  2625.         else // top

  2626.             ac_val -= 16 * s->block_wrap[n];

  2627. 

  2628.         while (!last) {

  2629.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2630.             i += skip;

  2631.             if (i > 63)

  2632.                 break;

  2633.             block[zz_table[i++]] = value;

  2634.         }

  2635. 

  2636.         /* apply AC prediction if needed */

  2637.         if (s->ac_pred) {

  2638.             if (dc_pred_dir) { // left

  2639.                 for (k = 1; k < 8; k++)

  2640.                     block[k << v->left_blk_sh] += ac_val[k];

  2641.             } else { // top

  2642.                 for (k = 1; k < 8; k++)

  2643.                     block[k << v->top_blk_sh] += ac_val[k + 8];

  2644.             }

  2645.         }

  2646.         /* save AC coeffs for further prediction */

  2647.         for (k = 1; k < 8; k++) {

  2648.             ac_val2[k]     = block[k << v->left_blk_sh];

  2649.             ac_val2[k + 8] = block[k << v->top_blk_sh];

  2650.         }

  2651. 

  2652.         /* scale AC coeffs */

  2653.         for (k = 1; k < 64; k++)

  2654.             if (block[k]) {

  2655.                 block[k] *= scale;

  2656.                 if (!v->pquantizer)

  2657.                     block[k] += (block[k] < 0) ? -v->pq : v->pq;

  2658.             }

  2659. 

  2660.         if (s->ac_pred) i = 63;

  2661.     }

  2662. 

  2663. not_coded:

  2664.     if (!coded) {

  2665.         int k, scale;

  2666.         ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;

  2667.         ac_val2 = ac_val;

  2668. 

  2669.         i = 0;

  2670.         scale = v->pq * 2 + v->halfpq;

  2671.         memset(ac_val2, 0, 16 * 2);

  2672.         if (dc_pred_dir) { // left

  2673.             ac_val -= 16;

  2674.             if (s->ac_pred)

  2675.                 memcpy(ac_val2, ac_val, 8 * 2);

  2676.         } else { // top

  2677.             ac_val -= 16 * s->block_wrap[n];

  2678.             if (s->ac_pred)

  2679.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  2680.         }

  2681. 

  2682.         /* apply AC prediction if needed */

  2683.         if (s->ac_pred) {

  2684.             if (dc_pred_dir) { //left

  2685.                 for (k = 1; k < 8; k++) {

  2686.                     block[k << v->left_blk_sh] = ac_val[k] * scale;

  2687.                     if (!v->pquantizer && block[k << v->left_blk_sh])

  2688.                         block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -v->pq : v->pq;

  2689.                 }

  2690.             } else { // top

  2691.                 for (k = 1; k < 8; k++) {

  2692.                     block[k << v->top_blk_sh] = ac_val[k + 8] * scale;

  2693.                     if (!v->pquantizer && block[k << v->top_blk_sh])

  2694.                         block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -v->pq : v->pq;

  2695.                 }

  2696.             }

  2697.             i = 63;

  2698.         }

  2699.     }

  2700.     s->block_last_index[n] = i;

  2701. 

  2702.     return 0;

  2703. }

  2704. 

  2705. /** Decode intra block in intra frames - should be faster than decode_intra_block

  2706.  * @param v VC1Context

  2707.  * @param block block to decode

  2708.  * @param[in] n subblock number

  2709.  * @param coded are AC coeffs present or not

  2710.  * @param codingset set of VLC to decode data

  2711.  * @param mquant quantizer value for this macroblock

  2712.  */

  2713. static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n,

  2714.                                   int coded, int codingset, int mquant)

  2715. {

  2716.     GetBitContext *gb = &v->s.gb;

  2717.     MpegEncContext *s = &v->s;

  2718.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2719.     int i;

  2720.     int16_t *dc_val;

  2721.     int16_t *ac_val, *ac_val2;

  2722.     int dcdiff;

  2723.     int a_avail = v->a_avail, c_avail = v->c_avail;

  2724.     int use_pred = s->ac_pred;

  2725.     int scale;

  2726.     int q1, q2 = 0;

  2727.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2728. 

  2729.     /* Get DC differential */

  2730.     if (n < 4) {

  2731.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2732.     } else {

  2733.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2734.     }

  2735.     if (dcdiff < 0) {

  2736.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2737.         return -1;

  2738.     }

  2739.     if (dcdiff) {

  2740.         if (dcdiff == 119 /* ESC index value */) {

  2741.             /* TODO: Optimize */

  2742.             if (mquant == 1)      dcdiff = get_bits(gb, 10);

  2743.             else if (mquant == 2) dcdiff = get_bits(gb, 9);

  2744.             else                  dcdiff = get_bits(gb, 8);

  2745.         } else {

  2746.             if (mquant == 1)

  2747.                 dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;

  2748.             else if (mquant == 2)

  2749.                 dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;

  2750.         }

  2751.         if (get_bits1(gb))

  2752.             dcdiff = -dcdiff;

  2753.     }

  2754. 

  2755.     /* Prediction */

  2756.     dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);

  2757.     *dc_val = dcdiff;

  2758. 

  2759.     /* Store the quantized DC coeff, used for prediction */

  2760.     if (n < 4) {

  2761.         block[0] = dcdiff * s->y_dc_scale;

  2762.     } else {

  2763.         block[0] = dcdiff * s->c_dc_scale;

  2764.     }

  2765. 

  2766.     //AC Decoding

  2767.     i = 1;

  2768. 

  2769.     /* check if AC is needed at all */

  2770.     if (!a_avail && !c_avail)

  2771.         use_pred = 0;

  2772.     ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;

  2773.     ac_val2 = ac_val;

  2774. 

  2775.     scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);

  2776. 

  2777.     if (dc_pred_dir) // left

  2778.         ac_val -= 16;

  2779.     else // top

  2780.         ac_val -= 16 * s->block_wrap[n];

  2781. 

  2782.     q1 = s->current_picture.f.qscale_table[mb_pos];

  2783.     if ( dc_pred_dir && c_avail && mb_pos)

  2784.         q2 = s->current_picture.f.qscale_table[mb_pos - 1];

  2785.     if (!dc_pred_dir && a_avail && mb_pos >= s->mb_stride)

  2786.         q2 = s->current_picture.f.qscale_table[mb_pos - s->mb_stride];

  2787.     if ( dc_pred_dir && n == 1)

  2788.         q2 = q1;

  2789.     if (!dc_pred_dir && n == 2)

  2790.         q2 = q1;

  2791.     if (n == 3)

  2792.         q2 = q1;

  2793. 

  2794.     if (coded) {

  2795.         int last = 0, skip, value;

  2796.         const uint8_t *zz_table;

  2797.         int k;

  2798. 

  2799.         if (v->s.ac_pred) {

  2800.             if (!use_pred && v->fcm == ILACE_FRAME) {

  2801.                 zz_table = v->zzi_8x8;

  2802.             } else {

  2803.                 if (!dc_pred_dir) // top

  2804.                     zz_table = v->zz_8x8[2];

  2805.                 else // left

  2806.                     zz_table = v->zz_8x8[3];

  2807.             }

  2808.         } else {

  2809.             if (v->fcm != ILACE_FRAME)

  2810.                 zz_table = v->zz_8x8[1];

  2811.             else

  2812.                 zz_table = v->zzi_8x8;

  2813.         }

  2814. 

  2815.         while (!last) {

  2816.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2817.             i += skip;

  2818.             if (i > 63)

  2819.                 break;

  2820.             block[zz_table[i++]] = value;

  2821.         }

  2822. 

  2823.         /* apply AC prediction if needed */

  2824.         if (use_pred) {

  2825.             /* scale predictors if needed*/

  2826.             if (q2 && q1 != q2) {

  2827.                 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2828.                 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2829. 

  2830.                 if (q1 < 1)

  2831.                     return AVERROR_INVALIDDATA;

  2832.                 if (dc_pred_dir) { // left

  2833.                     for (k = 1; k < 8; k++)

  2834.                         block[k << v->left_blk_sh] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2835.                 } else { // top

  2836.                     for (k = 1; k < 8; k++)

  2837.                         block[k << v->top_blk_sh] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2838.                 }

  2839.             } else {

  2840.                 if (dc_pred_dir) { //left

  2841.                     for (k = 1; k < 8; k++)

  2842.                         block[k << v->left_blk_sh] += ac_val[k];

  2843.                 } else { //top

  2844.                     for (k = 1; k < 8; k++)

  2845.                         block[k << v->top_blk_sh] += ac_val[k + 8];

  2846.                 }

  2847.             }

  2848.         }

  2849.         /* save AC coeffs for further prediction */

  2850.         for (k = 1; k < 8; k++) {

  2851.             ac_val2[k    ] = block[k << v->left_blk_sh];

  2852.             ac_val2[k + 8] = block[k << v->top_blk_sh];

  2853.         }

  2854. 

  2855.         /* scale AC coeffs */

  2856.         for (k = 1; k < 64; k++)

  2857.             if (block[k]) {

  2858.                 block[k] *= scale;

  2859.                 if (!v->pquantizer)

  2860.                     block[k] += (block[k] < 0) ? -mquant : mquant;

  2861.             }

  2862. 

  2863.         if (use_pred) i = 63;

  2864.     } else { // no AC coeffs

  2865.         int k;

  2866. 

  2867.         memset(ac_val2, 0, 16 * 2);

  2868.         if (dc_pred_dir) { // left

  2869.             if (use_pred) {

  2870.                 memcpy(ac_val2, ac_val, 8 * 2);

  2871.                 if (q2 && q1 != q2) {

  2872.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2873.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2874.                     if (q1 < 1)

  2875.                         return AVERROR_INVALIDDATA;

  2876.                     for (k = 1; k < 8; k++)

  2877.                         ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2878.                 }

  2879.             }

  2880.         } else { // top

  2881.             if (use_pred) {

  2882.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  2883.                 if (q2 && q1 != q2) {

  2884.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2885.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2886.                     if (q1 < 1)

  2887.                         return AVERROR_INVALIDDATA;

  2888.                     for (k = 1; k < 8; k++)

  2889.                         ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2890.                 }

  2891.             }

  2892.         }

  2893. 

  2894.         /* apply AC prediction if needed */

  2895.         if (use_pred) {

  2896.             if (dc_pred_dir) { // left

  2897.                 for (k = 1; k < 8; k++) {

  2898.                     block[k << v->left_blk_sh] = ac_val2[k] * scale;

  2899.                     if (!v->pquantizer && block[k << v->left_blk_sh])

  2900.                         block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -mquant : mquant;

  2901.                 }

  2902.             } else { // top

  2903.                 for (k = 1; k < 8; k++) {

  2904.                     block[k << v->top_blk_sh] = ac_val2[k + 8] * scale;

  2905.                     if (!v->pquantizer && block[k << v->top_blk_sh])

  2906.                         block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -mquant : mquant;

  2907.                 }

  2908.             }

  2909.             i = 63;

  2910.         }

  2911.     }

  2912.     s->block_last_index[n] = i;

  2913. 

  2914.     return 0;

  2915. }

  2916. 

  2917. /** Decode intra block in inter frames - more generic version than vc1_decode_i_block

  2918.  * @param v VC1Context

  2919.  * @param block block to decode

  2920.  * @param[in] n subblock index

  2921.  * @param coded are AC coeffs present or not

  2922.  * @param mquant block quantizer

  2923.  * @param codingset set of VLC to decode data

  2924.  */

  2925. static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n,

  2926.                                   int coded, int mquant, int codingset)

  2927. {

  2928.     GetBitContext *gb = &v->s.gb;

  2929.     MpegEncContext *s = &v->s;

  2930.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2931.     int i;

  2932.     int16_t *dc_val;

  2933.     int16_t *ac_val, *ac_val2;

  2934.     int dcdiff;

  2935.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2936.     int a_avail = v->a_avail, c_avail = v->c_avail;

  2937.     int use_pred = s->ac_pred;

  2938.     int scale;

  2939.     int q1, q2 = 0;

  2940. 

  2941.     s->dsp.clear_block(block);

  2942. 

  2943.     /* XXX: Guard against dumb values of mquant */

  2944.     mquant = (mquant < 1) ? 0 : ((mquant > 31) ? 31 : mquant);

  2945. 

  2946.     /* Set DC scale - y and c use the same */

  2947.     s->y_dc_scale = s->y_dc_scale_table[mquant];

  2948.     s->c_dc_scale = s->c_dc_scale_table[mquant];

  2949. 

  2950.     /* Get DC differential */

  2951.     if (n < 4) {

  2952.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2953.     } else {

  2954.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2955.     }

  2956.     if (dcdiff < 0) {

  2957.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2958.         return -1;

  2959.     }

  2960.     if (dcdiff) {

  2961.         if (dcdiff == 119 /* ESC index value */) {

  2962.             /* TODO: Optimize */

  2963.             if (mquant == 1)      dcdiff = get_bits(gb, 10);

  2964.             else if (mquant == 2) dcdiff = get_bits(gb, 9);

  2965.             else                  dcdiff = get_bits(gb, 8);

  2966.         } else {

  2967.             if (mquant == 1)

  2968.                 dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;

  2969.             else if (mquant == 2)

  2970.                 dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;

  2971.         }

  2972.         if (get_bits1(gb))

  2973.             dcdiff = -dcdiff;

  2974.     }

  2975. 

  2976.     /* Prediction */

  2977.     dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);

  2978.     *dc_val = dcdiff;

  2979. 

  2980.     /* Store the quantized DC coeff, used for prediction */

  2981. 

  2982.     if (n < 4) {

  2983.         block[0] = dcdiff * s->y_dc_scale;

  2984.     } else {

  2985.         block[0] = dcdiff * s->c_dc_scale;

  2986.     }

  2987. 

  2988.     //AC Decoding

  2989.     i = 1;

  2990. 

  2991.     /* check if AC is needed at all and adjust direction if needed */

  2992.     if (!a_avail) dc_pred_dir = 1;

  2993.     if (!c_avail) dc_pred_dir = 0;

  2994.     if (!a_avail && !c_avail) use_pred = 0;

  2995.     ac_val = s->ac_val[0][0] + s->block_index[n] * 16;

  2996.     ac_val2 = ac_val;

  2997. 

  2998.     scale = mquant * 2 + v->halfpq;

  2999. 

  3000.     if (dc_pred_dir) //left

  3001.         ac_val -= 16;

  3002.     else //top

  3003.         ac_val -= 16 * s->block_wrap[n];

  3004. 

  3005.     q1 = s->current_picture.f.qscale_table[mb_pos];

  3006.     if (dc_pred_dir && c_avail && mb_pos)

  3007.         q2 = s->current_picture.f.qscale_table[mb_pos - 1];

  3008.     if (!dc_pred_dir && a_avail && mb_pos >= s->mb_stride)

  3009.         q2 = s->current_picture.f.qscale_table[mb_pos - s->mb_stride];

  3010.     if ( dc_pred_dir && n == 1)

  3011.         q2 = q1;

  3012.     if (!dc_pred_dir && n == 2)

  3013.         q2 = q1;

  3014.     if (n == 3) q2 = q1;

  3015. 

  3016.     if (coded) {

  3017.         int last = 0, skip, value;

  3018.         int k;

  3019. 

  3020.         while (!last) {

  3021.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  3022.             i += skip;

  3023.             if (i > 63)

  3024.                 break;

  3025.             if (v->fcm == PROGRESSIVE)

  3026.                 block[v->zz_8x8[0][i++]] = value;

  3027.             else {

  3028.                 if (use_pred && (v->fcm == ILACE_FRAME)) {

  3029.                     if (!dc_pred_dir) // top

  3030.                         block[v->zz_8x8[2][i++]] = value;

  3031.                     else // left

  3032.                         block[v->zz_8x8[3][i++]] = value;

  3033.                 } else {

  3034.                     block[v->zzi_8x8[i++]] = value;

  3035.                 }

  3036.             }

  3037.         }

  3038. 

  3039.         /* apply AC prediction if needed */

  3040.         if (use_pred) {

  3041.             /* scale predictors if needed*/

  3042.             if (q2 && q1 != q2) {

  3043.                 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3044.                 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3045. 

  3046.                 if (q1 < 1)

  3047.                     return AVERROR_INVALIDDATA;

  3048.                 if (dc_pred_dir) { // left

  3049.                     for (k = 1; k < 8; k++)

  3050.                         block[k << v->left_blk_sh] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3051.                 } else { //top

  3052.                     for (k = 1; k < 8; k++)

  3053.                         block[k << v->top_blk_sh] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3054.                 }

  3055.             } else {

  3056.                 if (dc_pred_dir) { // left

  3057.                     for (k = 1; k < 8; k++)

  3058.                         block[k << v->left_blk_sh] += ac_val[k];

  3059.                 } else { // top

  3060.                     for (k = 1; k < 8; k++)

  3061.                         block[k << v->top_blk_sh] += ac_val[k + 8];

  3062.                 }

  3063.             }

  3064.         }

  3065.         /* save AC coeffs for further prediction */

  3066.         for (k = 1; k < 8; k++) {

  3067.             ac_val2[k    ] = block[k << v->left_blk_sh];

  3068.             ac_val2[k + 8] = block[k << v->top_blk_sh];

  3069.         }

  3070. 

  3071.         /* scale AC coeffs */

  3072.         for (k = 1; k < 64; k++)

  3073.             if (block[k]) {

  3074.                 block[k] *= scale;

  3075.                 if (!v->pquantizer)

  3076.                     block[k] += (block[k] < 0) ? -mquant : mquant;

  3077.             }

  3078. 

  3079.         if (use_pred) i = 63;

  3080.     } else { // no AC coeffs

  3081.         int k;

  3082. 

  3083.         memset(ac_val2, 0, 16 * 2);

  3084.         if (dc_pred_dir) { // left

  3085.             if (use_pred) {

  3086.                 memcpy(ac_val2, ac_val, 8 * 2);

  3087.                 if (q2 && q1 != q2) {

  3088.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3089.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3090.                     if (q1 < 1)

  3091.                         return AVERROR_INVALIDDATA;

  3092.                     for (k = 1; k < 8; k++)

  3093.                         ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3094.                 }

  3095.             }

  3096.         } else { // top

  3097.             if (use_pred) {

  3098.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  3099.                 if (q2 && q1 != q2) {

  3100.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3101.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3102.                     if (q1 < 1)

  3103.                         return AVERROR_INVALIDDATA;

  3104.                     for (k = 1; k < 8; k++)

  3105.                         ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3106.                 }

  3107.             }

  3108.         }

  3109. 

  3110.         /* apply AC prediction if needed */

  3111.         if (use_pred) {

  3112.             if (dc_pred_dir) { // left

  3113.                 for (k = 1; k < 8; k++) {

  3114.                     block[k << v->left_blk_sh] = ac_val2[k] * scale;

  3115.                     if (!v->pquantizer && block[k << v->left_blk_sh])

  3116.                         block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -mquant : mquant;

  3117.                 }

  3118.             } else { // top

  3119.                 for (k = 1; k < 8; k++) {

  3120.                     block[k << v->top_blk_sh] = ac_val2[k + 8] * scale;

  3121.                     if (!v->pquantizer && block[k << v->top_blk_sh])

  3122.                         block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -mquant : mquant;

  3123.                 }

  3124.             }

  3125.             i = 63;

  3126.         }

  3127.     }

  3128.     s->block_last_index[n] = i;

  3129. 

  3130.     return 0;

  3131. }

  3132. 

  3133. /** Decode P block

  3134.  */

  3135. static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n,

  3136.                               int mquant, int ttmb, int first_block,

  3137.                               uint8_t *dst, int linesize, int skip_block,

  3138.                               int *ttmb_out)

  3139. {

  3140.     MpegEncContext *s = &v->s;

  3141.     GetBitContext *gb = &s->gb;

  3142.     int i, j;

  3143.     int subblkpat = 0;

  3144.     int scale, off, idx, last, skip, value;

  3145.     int ttblk = ttmb & 7;

  3146.     int pat = 0;

  3147. 

  3148.     s->dsp.clear_block(block);

  3149. 

  3150.     if (ttmb == -1) {

  3151.         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)];

  3152.     }

  3153.     if (ttblk == TT_4X4) {

  3154.         subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);

  3155.     }

  3156.     if ((ttblk != TT_8X8 && ttblk != TT_4X4)

  3157.         && ((v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))

  3158.             || (!v->res_rtm_flag && !first_block))) {

  3159.         subblkpat = decode012(gb);

  3160.         if (subblkpat)

  3161.             subblkpat ^= 3; // swap decoded pattern bits

  3162.         if (ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM)

  3163.             ttblk = TT_8X4;

  3164.         if (ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT)

  3165.             ttblk = TT_4X8;

  3166.     }

  3167.     scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);

  3168. 

  3169.     // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT

  3170.     if (ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {

  3171.         subblkpat = 2 - (ttblk == TT_8X4_TOP);

  3172.         ttblk     = TT_8X4;

  3173.     }

  3174.     if (ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {

  3175.         subblkpat = 2 - (ttblk == TT_4X8_LEFT);

  3176.         ttblk     = TT_4X8;

  3177.     }

  3178.     switch (ttblk) {

  3179.     case TT_8X8:

  3180.         pat  = 0xF;

  3181.         i    = 0;

  3182.         last = 0;

  3183.         while (!last) {

  3184.             vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3185.             i += skip;

  3186.             if (i > 63)

  3187.                 break;

  3188.             if (!v->fcm)

  3189.                 idx = v->zz_8x8[0][i++];

  3190.             else

  3191.                 idx = v->zzi_8x8[i++];

  3192.             block[idx] = value * scale;

  3193.             if (!v->pquantizer)

  3194.                 block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3195.         }

  3196.         if (!skip_block) {

  3197.             if (i == 1)

  3198.                 v->vc1dsp.vc1_inv_trans_8x8_dc(dst, linesize, block);

  3199.             else {

  3200.                 v->vc1dsp.vc1_inv_trans_8x8(block);

  3201.                 s->dsp.add_pixels_clamped(block, dst, linesize);

  3202.             }

  3203.         }

  3204.         break;

  3205.     case TT_4X4:

  3206.         pat = ~subblkpat & 0xF;

  3207.         for (j = 0; j < 4; j++) {

  3208.             last = subblkpat & (1 << (3 - j));

  3209.             i    = 0;

  3210.             off  = (j & 1) * 4 + (j & 2) * 16;

  3211.             while (!last) {

  3212.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3213.                 i += skip;

  3214.                 if (i > 15)

  3215.                     break;

  3216.                 if (!v->fcm)

  3217.                     idx = ff_vc1_simple_progressive_4x4_zz[i++];

  3218.                 else

  3219.                     idx = ff_vc1_adv_interlaced_4x4_zz[i++];

  3220.                 block[idx + off] = value * scale;

  3221.                 if (!v->pquantizer)

  3222.                     block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;

  3223.             }

  3224.             if (!(subblkpat & (1 << (3 - j))) && !skip_block) {

  3225.                 if (i == 1)

  3226.                     v->vc1dsp.vc1_inv_trans_4x4_dc(dst + (j & 1) * 4 + (j & 2) * 2 * linesize, linesize, block + off);

  3227.                 else

  3228.                     v->vc1dsp.vc1_inv_trans_4x4(dst + (j & 1) * 4 + (j & 2) *  2 * linesize, linesize, block + off);

  3229.             }

  3230.         }

  3231.         break;

  3232.     case TT_8X4:

  3233.         pat = ~((subblkpat & 2) * 6 + (subblkpat & 1) * 3) & 0xF;

  3234.         for (j = 0; j < 2; j++) {

  3235.             last = subblkpat & (1 << (1 - j));

  3236.             i    = 0;

  3237.             off  = j * 32;

  3238.             while (!last) {

  3239.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3240.                 i += skip;

  3241.                 if (i > 31)

  3242.                     break;

  3243.                 if (!v->fcm)

  3244.                     idx = v->zz_8x4[i++] + off;

  3245.                 else

  3246.                     idx = ff_vc1_adv_interlaced_8x4_zz[i++] + off;

  3247.                 block[idx] = value * scale;

  3248.                 if (!v->pquantizer)

  3249.                     block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3250.             }

  3251.             if (!(subblkpat & (1 << (1 - j))) && !skip_block) {

  3252.                 if (i == 1)

  3253.                     v->vc1dsp.vc1_inv_trans_8x4_dc(dst + j * 4 * linesize, linesize, block + off);

  3254.                 else

  3255.                     v->vc1dsp.vc1_inv_trans_8x4(dst + j * 4 * linesize, linesize, block + off);

  3256.             }

  3257.         }

  3258.         break;

  3259.     case TT_4X8:

  3260.         pat = ~(subblkpat * 5) & 0xF;

  3261.         for (j = 0; j < 2; j++) {

  3262.             last = subblkpat & (1 << (1 - j));

  3263.             i    = 0;

  3264.             off  = j * 4;

  3265.             while (!last) {

  3266.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3267.                 i += skip;

  3268.                 if (i > 31)

  3269.                     break;

  3270.                 if (!v->fcm)

  3271.                     idx = v->zz_4x8[i++] + off;

  3272.                 else

  3273.                     idx = ff_vc1_adv_interlaced_4x8_zz[i++] + off;

  3274.                 block[idx] = value * scale;

  3275.                 if (!v->pquantizer)

  3276.                     block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3277.             }

  3278.             if (!(subblkpat & (1 << (1 - j))) && !skip_block) {

  3279.                 if (i == 1)

  3280.                     v->vc1dsp.vc1_inv_trans_4x8_dc(dst + j * 4, linesize, block + off);

  3281.                 else

  3282.                     v->vc1dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);

  3283.             }

  3284.         }

  3285.         break;

  3286.     }

  3287.     if (ttmb_out)

  3288.         *ttmb_out |= ttblk << (n * 4);

  3289.     return pat;

  3290. }

  3291. 

  3292. /** @} */ // Macroblock group

  3293. 

  3294. static const int size_table  [6] = { 0, 2, 3, 4,  5,  8 };

  3295. static const int offset_table[6] = { 0, 1, 3, 7, 15, 31 };

  3296. 

  3297. static av_always_inline void vc1_apply_p_v_loop_filter(VC1Context *v, int block_num)

  3298. {

  3299.     MpegEncContext *s  = &v->s;

  3300.     int mb_cbp         = v->cbp[s->mb_x - s->mb_stride],

  3301.         block_cbp      = mb_cbp      >> (block_num * 4), bottom_cbp,

  3302.         mb_is_intra    = v->is_intra[s->mb_x - s->mb_stride],

  3303.         block_is_intra = mb_is_intra >> (block_num * 4), bottom_is_intra;

  3304.     int idx, linesize  = block_num > 3 ? s->uvlinesize : s->linesize, ttblk;

  3305.     uint8_t *dst;

  3306. 

  3307.     if (block_num > 3) {

  3308.         dst      = s->dest[block_num - 3];

  3309.     } else {

  3310.         dst      = s->dest[0] + (block_num & 1) * 8 + ((block_num & 2) * 4 - 8) * linesize;

  3311.     }

  3312.     if (s->mb_y != s->end_mb_y || block_num < 2) {

  3313.         int16_t (*mv)[2];

  3314.         int mv_stride;

  3315. 

  3316.         if (block_num > 3) {

  3317.             bottom_cbp      = v->cbp[s->mb_x]      >> (block_num * 4);

  3318.             bottom_is_intra = v->is_intra[s->mb_x] >> (block_num * 4);

  3319.             mv              = &v->luma_mv[s->mb_x - s->mb_stride];

  3320.             mv_stride       = s->mb_stride;

  3321.         } else {

  3322.             bottom_cbp      = (block_num < 2) ? (mb_cbp               >> ((block_num + 2) * 4))

  3323.                                               : (v->cbp[s->mb_x]      >> ((block_num - 2) * 4));

  3324.             bottom_is_intra = (block_num < 2) ? (mb_is_intra          >> ((block_num + 2) * 4))

  3325.                                               : (v->is_intra[s->mb_x] >> ((block_num - 2) * 4));

  3326.             mv_stride       = s->b8_stride;

  3327.             mv              = &s->current_picture.f.motion_val[0][s->block_index[block_num] - 2 * mv_stride];

  3328.         }

  3329. 

  3330.         if (bottom_is_intra & 1 || block_is_intra & 1 ||

  3331.             mv[0][0] != mv[mv_stride][0] || mv[0][1] != mv[mv_stride][1]) {

  3332.             v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);

  3333.         } else {

  3334.             idx = ((bottom_cbp >> 2) | block_cbp) & 3;

  3335.             if (idx == 3) {

  3336.                 v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);

  3337.             } else if (idx) {

  3338.                 if (idx == 1)

  3339.                     v->vc1dsp.vc1_v_loop_filter4(dst + 4, linesize, v->pq);

  3340.                 else

  3341.                     v->vc1dsp.vc1_v_loop_filter4(dst,     linesize, v->pq);

  3342.             }

  3343.         }

  3344.     }

  3345. 

  3346.     dst -= 4 * linesize;

  3347.     ttblk = (v->ttblk[s->mb_x - s->mb_stride] >> (block_num * 4)) & 0xF;

  3348.     if (ttblk == TT_4X4 || ttblk == TT_8X4) {

  3349.         idx = (block_cbp | (block_cbp >> 2)) & 3;

  3350.         if (idx == 3) {

  3351.             v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);

  3352.         } else if (idx) {

  3353.             if (idx == 1)

  3354.                 v->vc1dsp.vc1_v_loop_filter4(dst + 4, linesize, v->pq);

  3355.             else

  3356.                 v->vc1dsp.vc1_v_loop_filter4(dst,     linesize, v->pq);

  3357.         }

  3358.     }

  3359. }

  3360. 

  3361. static av_always_inline void vc1_apply_p_h_loop_filter(VC1Context *v, int block_num)

  3362. {

  3363.     MpegEncContext *s  = &v->s;

  3364.     int mb_cbp         = v->cbp[s->mb_x - 1 - s->mb_stride],

  3365.         block_cbp      = mb_cbp      >> (block_num * 4), right_cbp,

  3366.         mb_is_intra    = v->is_intra[s->mb_x - 1 - s->mb_stride],

  3367.         block_is_intra = mb_is_intra >> (block_num * 4), right_is_intra;

  3368.     int idx, linesize  = block_num > 3 ? s->uvlinesize : s->linesize, ttblk;

  3369.     uint8_t *dst;

  3370. 

  3371.     if (block_num > 3) {

  3372.         dst = s->dest[block_num - 3] - 8 * linesize;

  3373.     } else {

  3374.         dst = s->dest[0] + (block_num & 1) * 8 + ((block_num & 2) * 4 - 16) * linesize - 8;

  3375.     }

  3376. 

  3377.     if (s->mb_x != s->mb_width || !(block_num & 5)) {

  3378.         int16_t (*mv)[2];

  3379. 

  3380.         if (block_num > 3) {

  3381.             right_cbp      = v->cbp[s->mb_x - s->mb_stride] >> (block_num * 4);

  3382.             right_is_intra = v->is_intra[s->mb_x - s->mb_stride] >> (block_num * 4);

  3383.             mv             = &v->luma_mv[s->mb_x - s->mb_stride - 1];

  3384.         } else {

  3385.             right_cbp      = (block_num & 1) ? (v->cbp[s->mb_x - s->mb_stride]      >> ((block_num - 1) * 4))

  3386.                                              : (mb_cbp                              >> ((block_num + 1) * 4));

  3387.             right_is_intra = (block_num & 1) ? (v->is_intra[s->mb_x - s->mb_stride] >> ((block_num - 1) * 4))

  3388.                                              : (mb_is_intra                         >> ((block_num + 1) * 4));

  3389.             mv             = &s->current_picture.f.motion_val[0][s->block_index[block_num] - s->b8_stride * 2 - 2];

  3390.         }

  3391.         if (block_is_intra & 1 || right_is_intra & 1 || mv[0][0] != mv[1][0] || mv[0][1] != mv[1][1]) {

  3392.             v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);

  3393.         } else {

  3394.             idx = ((right_cbp >> 1) | block_cbp) & 5; // FIXME check

  3395.             if (idx == 5) {

  3396.                 v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);

  3397.             } else if (idx) {

  3398.                 if (idx == 1)

  3399.                     v->vc1dsp.vc1_h_loop_filter4(dst + 4 * linesize, linesize, v->pq);

  3400.                 else

  3401.                     v->vc1dsp.vc1_h_loop_filter4(dst,                linesize, v->pq);

  3402.             }

  3403.         }

  3404.     }

  3405. 

  3406.     dst -= 4;

  3407.     ttblk = (v->ttblk[s->mb_x - s->mb_stride - 1] >> (block_num * 4)) & 0xf;

  3408.     if (ttblk == TT_4X4 || ttblk == TT_4X8) {

  3409.         idx = (block_cbp | (block_cbp >> 1)) & 5;

  3410.         if (idx == 5) {

  3411.             v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);

  3412.         } else if (idx) {

  3413.             if (idx == 1)

  3414.                 v->vc1dsp.vc1_h_loop_filter4(dst + linesize * 4, linesize, v->pq);

  3415.             else

  3416.                 v->vc1dsp.vc1_h_loop_filter4(dst,                linesize, v->pq);

  3417.         }

  3418.     }

  3419. }

  3420. 

  3421. static void vc1_apply_p_loop_filter(VC1Context *v)

  3422. {

  3423.     MpegEncContext *s = &v->s;

  3424.     int i;

  3425. 

  3426.     for (i = 0; i < 6; i++) {

  3427.         vc1_apply_p_v_loop_filter(v, i);

  3428.     }

  3429. 

  3430.     /* V always precedes H, therefore we run H one MB before V;

  3431.      * at the end of a row, we catch up to complete the row */

  3432.     if (s->mb_x) {

  3433.         for (i = 0; i < 6; i++) {

  3434.             vc1_apply_p_h_loop_filter(v, i);

  3435.         }

  3436.         if (s->mb_x == s->mb_width - 1) {

  3437.             s->mb_x++;

  3438.             ff_update_block_index(s);

  3439.             for (i = 0; i < 6; i++) {

  3440.                 vc1_apply_p_h_loop_filter(v, i);

  3441.             }

  3442.         }

  3443.     }

  3444. }

  3445. 

  3446. /** Decode one P-frame MB

  3447.  */

  3448. static int vc1_decode_p_mb(VC1Context *v)

  3449. {

  3450.     MpegEncContext *s = &v->s;

  3451.     GetBitContext *gb = &s->gb;

  3452.     int i, j;

  3453.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3454.     int cbp; /* cbp decoding stuff */

  3455.     int mqdiff, mquant; /* MB quantization */

  3456.     int ttmb = v->ttfrm; /* MB Transform type */

  3457. 

  3458.     int mb_has_coeffs = 1; /* last_flag */

  3459.     int dmv_x, dmv_y; /* Differential MV components */

  3460.     int index, index1; /* LUT indexes */

  3461.     int val, sign; /* temp values */

  3462.     int first_block = 1;

  3463.     int dst_idx, off;

  3464.     int skipped, fourmv;

  3465.     int block_cbp = 0, pat, block_tt = 0, block_intra = 0;

  3466. 

  3467.     mquant = v->pq; /* lossy initialization */

  3468. 

  3469.     if (v->mv_type_is_raw)

  3470.         fourmv = get_bits1(gb);

  3471.     else

  3472.         fourmv = v->mv_type_mb_plane[mb_pos];

  3473.     if (v->skip_is_raw)

  3474.         skipped = get_bits1(gb);

  3475.     else

  3476.         skipped = v->s.mbskip_table[mb_pos];

  3477. 

  3478.     if (!fourmv) { /* 1MV mode */

  3479.         if (!skipped) {

  3480.             GET_MVDATA(dmv_x, dmv_y);

  3481. 

  3482.             if (s->mb_intra) {

  3483.                 s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0;

  3484.                 s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0;

  3485.             }

  3486.             s->current_picture.f.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;

  3487.             vc1_pred_mv(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3488. 

  3489.             /* FIXME Set DC val for inter block ? */

  3490.             if (s->mb_intra && !mb_has_coeffs) {

  3491.                 GET_MQUANT();

  3492.                 s->ac_pred = get_bits1(gb);

  3493.                 cbp        = 0;

  3494.             } else if (mb_has_coeffs) {

  3495.                 if (s->mb_intra)

  3496.                     s->ac_pred = get_bits1(gb);

  3497.                 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3498.                 GET_MQUANT();

  3499.             } else {

  3500.                 mquant = v->pq;

  3501.                 cbp    = 0;

  3502.             }

  3503.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  3504. 

  3505.             if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)

  3506.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,

  3507.                                 VC1_TTMB_VLC_BITS, 2);

  3508.             if (!s->mb_intra) vc1_mc_1mv(v, 0);

  3509.             dst_idx = 0;

  3510.             for (i = 0; i < 6; i++) {

  3511.                 s->dc_val[0][s->block_index[i]] = 0;

  3512.                 dst_idx += i >> 2;

  3513.                 val = ((cbp >> (5 - i)) & 1);

  3514.                 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3515.                 v->mb_type[0][s->block_index[i]] = s->mb_intra;

  3516.                 if (s->mb_intra) {

  3517.                     /* check if prediction blocks A and C are available */

  3518.                     v->a_avail = v->c_avail = 0;

  3519.                     if (i == 2 || i == 3 || !s->first_slice_line)

  3520.                         v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3521.                     if (i == 1 || i == 3 || s->mb_x)

  3522.                         v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3523. 

  3524.                     vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  3525.                                            (i & 4) ? v->codingset2 : v->codingset);

  3526.                     if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  3527.                         continue;

  3528.                     v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3529.                     if (v->rangeredfrm)

  3530.                         for (j = 0; j < 64; j++)

  3531.                             s->block[i][j] <<= 1;

  3532.                     s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3533.                     if (v->pq >= 9 && v->overlap) {

  3534.                         if (v->c_avail)

  3535.                             v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3536.                         if (v->a_avail)

  3537.                             v->vc1dsp.vc1_v_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3538.                     }

  3539.                     block_cbp   |= 0xF << (i << 2);

  3540.                     block_intra |= 1 << i;

  3541.                 } else if (val) {

  3542.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block,

  3543.                                              s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize,

  3544.                                              (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);

  3545.                     block_cbp |= pat << (i << 2);

  3546.                     if (!v->ttmbf && ttmb < 8)

  3547.                         ttmb = -1;

  3548.                     first_block = 0;

  3549.                 }

  3550.             }

  3551.         } else { // skipped

  3552.             s->mb_intra = 0;

  3553.             for (i = 0; i < 6; i++) {

  3554.                 v->mb_type[0][s->block_index[i]] = 0;

  3555.                 s->dc_val[0][s->block_index[i]]  = 0;

  3556.             }

  3557.             s->current_picture.f.mb_type[mb_pos]      = MB_TYPE_SKIP;

  3558.             s->current_picture.f.qscale_table[mb_pos] = 0;

  3559.             vc1_pred_mv(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3560.             vc1_mc_1mv(v, 0);

  3561.         }

  3562.     } else { // 4MV mode

  3563.         if (!skipped /* unskipped MB */) {

  3564.             int intra_count = 0, coded_inter = 0;

  3565.             int is_intra[6], is_coded[6];

  3566.             /* Get CBPCY */

  3567.             cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3568.             for (i = 0; i < 6; i++) {

  3569.                 val = ((cbp >> (5 - i)) & 1);

  3570.                 s->dc_val[0][s->block_index[i]] = 0;

  3571.                 s->mb_intra                     = 0;

  3572.                 if (i < 4) {

  3573.                     dmv_x = dmv_y = 0;

  3574.                     s->mb_intra   = 0;

  3575.                     mb_has_coeffs = 0;

  3576.                     if (val) {

  3577.                         GET_MVDATA(dmv_x, dmv_y);

  3578.                     }

  3579.                     vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3580.                     if (!s->mb_intra)

  3581.                         vc1_mc_4mv_luma(v, i, 0);

  3582.                     intra_count += s->mb_intra;

  3583.                     is_intra[i]  = s->mb_intra;

  3584.                     is_coded[i]  = mb_has_coeffs;

  3585.                 }

  3586.                 if (i & 4) {

  3587.                     is_intra[i] = (intra_count >= 3);

  3588.                     is_coded[i] = val;

  3589.                 }

  3590.                 if (i == 4)

  3591.                     vc1_mc_4mv_chroma(v, 0);

  3592.                 v->mb_type[0][s->block_index[i]] = is_intra[i];

  3593.                 if (!coded_inter)

  3594.                     coded_inter = !is_intra[i] & is_coded[i];

  3595.             }

  3596.             // if there are no coded blocks then don't do anything more

  3597.             dst_idx = 0;

  3598.             if (!intra_count && !coded_inter)

  3599.                 goto end;

  3600.             GET_MQUANT();

  3601.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  3602.             /* test if block is intra and has pred */

  3603.             {

  3604.                 int intrapred = 0;

  3605.                 for (i = 0; i < 6; i++)

  3606.                     if (is_intra[i]) {

  3607.                         if (((!s->first_slice_line || (i == 2 || i == 3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])

  3608.                             || ((s->mb_x || (i == 1 || i == 3)) && v->mb_type[0][s->block_index[i] - 1])) {

  3609.                             intrapred = 1;

  3610.                             break;

  3611.                         }

  3612.                     }

  3613.                 if (intrapred)

  3614.                     s->ac_pred = get_bits1(gb);

  3615.                 else

  3616.                     s->ac_pred = 0;

  3617.             }

  3618.             if (!v->ttmbf && coded_inter)

  3619.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  3620.             for (i = 0; i < 6; i++) {

  3621.                 dst_idx    += i >> 2;

  3622.                 off         = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3623.                 s->mb_intra = is_intra[i];

  3624.                 if (is_intra[i]) {

  3625.                     /* check if prediction blocks A and C are available */

  3626.                     v->a_avail = v->c_avail = 0;

  3627.                     if (i == 2 || i == 3 || !s->first_slice_line)

  3628.                         v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3629.                     if (i == 1 || i == 3 || s->mb_x)

  3630.                         v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3631. 

  3632.                     vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant,

  3633.                                            (i & 4) ? v->codingset2 : v->codingset);

  3634.                     if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  3635.                         continue;

  3636.                     v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3637.                     if (v->rangeredfrm)

  3638.                         for (j = 0; j < 64; j++)

  3639.                             s->block[i][j] <<= 1;

  3640.                     s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off,

  3641.                                                      (i & 4) ? s->uvlinesize : s->linesize);

  3642.                     if (v->pq >= 9 && v->overlap) {

  3643.                         if (v->c_avail)

  3644.                             v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3645.                         if (v->a_avail)

  3646.                             v->vc1dsp.vc1_v_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3647.                     }

  3648.                     block_cbp   |= 0xF << (i << 2);

  3649.                     block_intra |= 1 << i;

  3650.                 } else if (is_coded[i]) {

  3651.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  3652.                                              first_block, s->dest[dst_idx] + off,

  3653.                                              (i & 4) ? s->uvlinesize : s->linesize,

  3654.                                              (i & 4) && (s->flags & CODEC_FLAG_GRAY),

  3655.                                              &block_tt);

  3656.                     block_cbp |= pat << (i << 2);

  3657.                     if (!v->ttmbf && ttmb < 8)

  3658.                         ttmb = -1;

  3659.                     first_block = 0;

  3660.                 }

  3661.             }

  3662.         } else { // skipped MB

  3663.             s->mb_intra                               = 0;

  3664.             s->current_picture.f.qscale_table[mb_pos] = 0;

  3665.             for (i = 0; i < 6; i++) {

  3666.                 v->mb_type[0][s->block_index[i]] = 0;

  3667.                 s->dc_val[0][s->block_index[i]]  = 0;

  3668.             }

  3669.             for (i = 0; i < 4; i++) {

  3670.                 vc1_pred_mv(v, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3671.                 vc1_mc_4mv_luma(v, i, 0);

  3672.             }

  3673.             vc1_mc_4mv_chroma(v, 0);

  3674.             s->current_picture.f.qscale_table[mb_pos] = 0;

  3675.         }

  3676.     }

  3677. end:

  3678.     v->cbp[s->mb_x]      = block_cbp;

  3679.     v->ttblk[s->mb_x]    = block_tt;

  3680.     v->is_intra[s->mb_x] = block_intra;

  3681. 

  3682.     return 0;

  3683. }

  3684. 

  3685. /* Decode one macroblock in an interlaced frame p picture */

  3686. 

  3687. static int vc1_decode_p_mb_intfr(VC1Context *v)

  3688. {

  3689.     MpegEncContext *s = &v->s;

  3690.     GetBitContext *gb = &s->gb;

  3691.     int i;

  3692.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3693.     int cbp = 0; /* cbp decoding stuff */

  3694.     int mqdiff, mquant; /* MB quantization */

  3695.     int ttmb = v->ttfrm; /* MB Transform type */

  3696. 

  3697.     int mb_has_coeffs = 1; /* last_flag */

  3698.     int dmv_x, dmv_y; /* Differential MV components */

  3699.     int val; /* temp value */

  3700.     int first_block = 1;

  3701.     int dst_idx, off;

  3702.     int skipped, fourmv = 0, twomv = 0;

  3703.     int block_cbp = 0, pat, block_tt = 0;

  3704.     int idx_mbmode = 0, mvbp;

  3705.     int stride_y, fieldtx;

  3706. 

  3707.     mquant = v->pq; /* Lossy initialization */

  3708. 

  3709.     if (v->skip_is_raw)

  3710.         skipped = get_bits1(gb);

  3711.     else

  3712.         skipped = v->s.mbskip_table[mb_pos];

  3713.     if (!skipped) {

  3714.         if (v->fourmvswitch)

  3715.             idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_4MV_MBMODE_VLC_BITS, 2); // try getting this done

  3716.         else

  3717.             idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_NON4MV_MBMODE_VLC_BITS, 2); // in a single line

  3718.         switch (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0]) {

  3719.         /* store the motion vector type in a flag (useful later) */

  3720.         case MV_PMODE_INTFR_4MV:

  3721.             fourmv = 1;

  3722.             v->blk_mv_type[s->block_index[0]] = 0;

  3723.             v->blk_mv_type[s->block_index[1]] = 0;

  3724.             v->blk_mv_type[s->block_index[2]] = 0;

  3725.             v->blk_mv_type[s->block_index[3]] = 0;

  3726.             break;

  3727.         case MV_PMODE_INTFR_4MV_FIELD:

  3728.             fourmv = 1;

  3729.             v->blk_mv_type[s->block_index[0]] = 1;

  3730.             v->blk_mv_type[s->block_index[1]] = 1;

  3731.             v->blk_mv_type[s->block_index[2]] = 1;

  3732.             v->blk_mv_type[s->block_index[3]] = 1;

  3733.             break;

  3734.         case MV_PMODE_INTFR_2MV_FIELD:

  3735.             twomv = 1;

  3736.             v->blk_mv_type[s->block_index[0]] = 1;

  3737.             v->blk_mv_type[s->block_index[1]] = 1;

  3738.             v->blk_mv_type[s->block_index[2]] = 1;

  3739.             v->blk_mv_type[s->block_index[3]] = 1;

  3740.             break;

  3741.         case MV_PMODE_INTFR_1MV:

  3742.             v->blk_mv_type[s->block_index[0]] = 0;

  3743.             v->blk_mv_type[s->block_index[1]] = 0;

  3744.             v->blk_mv_type[s->block_index[2]] = 0;

  3745.             v->blk_mv_type[s->block_index[3]] = 0;

  3746.             break;

  3747.         }

  3748.         if (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_INTRA) { // intra MB

  3749.             s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0;

  3750.             s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0;

  3751.             s->current_picture.f.mb_type[mb_pos]                     = MB_TYPE_INTRA;

  3752.             s->mb_intra = v->is_intra[s->mb_x] = 1;

  3753.             for (i = 0; i < 6; i++)

  3754.                 v->mb_type[0][s->block_index[i]] = 1;

  3755.             fieldtx = v->fieldtx_plane[mb_pos] = get_bits1(gb);

  3756.             mb_has_coeffs = get_bits1(gb);

  3757.             if (mb_has_coeffs)

  3758.                 cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3759.             v->s.ac_pred = v->acpred_plane[mb_pos] = get_bits1(gb);

  3760.             GET_MQUANT();

  3761.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  3762.             /* Set DC scale - y and c use the same (not sure if necessary here) */

  3763.             s->y_dc_scale = s->y_dc_scale_table[mquant];

  3764.             s->c_dc_scale = s->c_dc_scale_table[mquant];

  3765.             dst_idx = 0;

  3766.             for (i = 0; i < 6; i++) {

  3767.                 s->dc_val[0][s->block_index[i]] = 0;

  3768.                 dst_idx += i >> 2;

  3769.                 val = ((cbp >> (5 - i)) & 1);

  3770.                 v->mb_type[0][s->block_index[i]] = s->mb_intra;

  3771.                 v->a_avail = v->c_avail = 0;

  3772.                 if (i == 2 || i == 3 || !s->first_slice_line)

  3773.                     v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3774.                 if (i == 1 || i == 3 || s->mb_x)

  3775.                     v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3776. 

  3777.                 vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  3778.                                        (i & 4) ? v->codingset2 : v->codingset);

  3779.                 if ((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;

  3780.                 v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3781.                 if (i < 4) {

  3782.                     stride_y = s->linesize << fieldtx;

  3783.                     off = (fieldtx) ? ((i & 1) * 8) + ((i & 2) >> 1) * s->linesize : (i & 1) * 8 + 4 * (i & 2) * s->linesize;

  3784.                 } else {

  3785.                     stride_y = s->uvlinesize;

  3786.                     off = 0;

  3787.                 }

  3788.                 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, stride_y);

  3789.                 //TODO: loop filter

  3790.             }

  3791. 

  3792.         } else { // inter MB

  3793.             mb_has_coeffs = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][3];

  3794.             if (mb_has_coeffs)

  3795.                 cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3796.             if (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_2MV_FIELD) {

  3797.                 v->twomvbp = get_vlc2(gb, v->twomvbp_vlc->table, VC1_2MV_BLOCK_PATTERN_VLC_BITS, 1);

  3798.             } else {

  3799.                 if ((ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_4MV)

  3800.                     || (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_4MV_FIELD)) {

  3801.                     v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);

  3802.                 }

  3803.             }

  3804.             s->mb_intra = v->is_intra[s->mb_x] = 0;

  3805.             for (i = 0; i < 6; i++)

  3806.                 v->mb_type[0][s->block_index[i]] = 0;

  3807.             fieldtx = v->fieldtx_plane[mb_pos] = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][1];

  3808.             /* for all motion vector read MVDATA and motion compensate each block */

  3809.             dst_idx = 0;

  3810.             if (fourmv) {

  3811.                 mvbp = v->fourmvbp;

  3812.                 for (i = 0; i < 6; i++) {

  3813.                     if (i < 4) {

  3814.                         dmv_x = dmv_y = 0;

  3815.                         val   = ((mvbp >> (3 - i)) & 1);

  3816.                         if (val) {

  3817.                             get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3818.                         }

  3819.                         vc1_pred_mv_intfr(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);

  3820.                         vc1_mc_4mv_luma(v, i, 0);

  3821.                     } else if (i == 4) {

  3822.                         vc1_mc_4mv_chroma4(v);

  3823.                     }

  3824.                 }

  3825.             } else if (twomv) {

  3826.                 mvbp  = v->twomvbp;

  3827.                 dmv_x = dmv_y = 0;

  3828.                 if (mvbp & 2) {

  3829.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3830.                 }

  3831.                 vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0]);

  3832.                 vc1_mc_4mv_luma(v, 0, 0);

  3833.                 vc1_mc_4mv_luma(v, 1, 0);

  3834.                 dmv_x = dmv_y = 0;

  3835.                 if (mvbp & 1) {

  3836.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3837.                 }

  3838.                 vc1_pred_mv_intfr(v, 2, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0]);

  3839.                 vc1_mc_4mv_luma(v, 2, 0);

  3840.                 vc1_mc_4mv_luma(v, 3, 0);

  3841.                 vc1_mc_4mv_chroma4(v);

  3842.             } else {

  3843.                 mvbp = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][2];

  3844.                 dmv_x = dmv_y = 0;

  3845.                 if (mvbp) {

  3846.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3847.                 }

  3848.                 vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);

  3849.                 vc1_mc_1mv(v, 0);

  3850.             }

  3851.             if (cbp)

  3852.                 GET_MQUANT();  // p. 227

  3853.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  3854.             if (!v->ttmbf && cbp)

  3855.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  3856.             for (i = 0; i < 6; i++) {

  3857.                 s->dc_val[0][s->block_index[i]] = 0;

  3858.                 dst_idx += i >> 2;

  3859.                 val = ((cbp >> (5 - i)) & 1);

  3860.                 if (!fieldtx)

  3861.                     off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3862.                 else

  3863.                     off = (i & 4) ? 0 : ((i & 1) * 8 + ((i > 1) * s->linesize));

  3864.                 if (val) {

  3865.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  3866.                                              first_block, s->dest[dst_idx] + off,

  3867.                                              (i & 4) ? s->uvlinesize : (s->linesize << fieldtx),

  3868.                                              (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);

  3869.                     block_cbp |= pat << (i << 2);

  3870.                     if (!v->ttmbf && ttmb < 8)

  3871.                         ttmb = -1;

  3872.                     first_block = 0;

  3873.                 }

  3874.             }

  3875.         }

  3876.     } else { // skipped

  3877.         s->mb_intra = v->is_intra[s->mb_x] = 0;

  3878.         for (i = 0; i < 6; i++) {

  3879.             v->mb_type[0][s->block_index[i]] = 0;

  3880.             s->dc_val[0][s->block_index[i]] = 0;

  3881.         }

  3882.         s->current_picture.f.mb_type[mb_pos]      = MB_TYPE_SKIP;

  3883.         s->current_picture.f.qscale_table[mb_pos] = 0;

  3884.         v->blk_mv_type[s->block_index[0]] = 0;

  3885.         v->blk_mv_type[s->block_index[1]] = 0;

  3886.         v->blk_mv_type[s->block_index[2]] = 0;

  3887.         v->blk_mv_type[s->block_index[3]] = 0;

  3888.         vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);

  3889.         vc1_mc_1mv(v, 0);

  3890.     }

  3891.     if (s->mb_x == s->mb_width - 1)

  3892.         memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0])*s->mb_stride);

  3893.     return 0;

  3894. }

  3895. 

  3896. static int vc1_decode_p_mb_intfi(VC1Context *v)

  3897. {

  3898.     MpegEncContext *s = &v->s;

  3899.     GetBitContext *gb = &s->gb;

  3900.     int i;

  3901.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3902.     int cbp = 0; /* cbp decoding stuff */

  3903.     int mqdiff, mquant; /* MB quantization */

  3904.     int ttmb = v->ttfrm; /* MB Transform type */

  3905. 

  3906.     int mb_has_coeffs = 1; /* last_flag */

  3907.     int dmv_x, dmv_y; /* Differential MV components */

  3908.     int val; /* temp values */

  3909.     int first_block = 1;

  3910.     int dst_idx, off;

  3911.     int pred_flag = 0;

  3912.     int block_cbp = 0, pat, block_tt = 0;

  3913.     int idx_mbmode = 0;

  3914. 

  3915.     mquant = v->pq; /* Lossy initialization */

  3916. 

  3917.     idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_IF_MBMODE_VLC_BITS, 2);

  3918.     if (idx_mbmode <= 1) { // intra MB

  3919.         s->mb_intra = v->is_intra[s->mb_x] = 1;

  3920.         s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;

  3921.         s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;

  3922.         s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_INTRA;

  3923.         GET_MQUANT();

  3924.         s->current_picture.f.qscale_table[mb_pos] = mquant;

  3925.         /* Set DC scale - y and c use the same (not sure if necessary here) */

  3926.         s->y_dc_scale = s->y_dc_scale_table[mquant];

  3927.         s->c_dc_scale = s->c_dc_scale_table[mquant];

  3928.         v->s.ac_pred  = v->acpred_plane[mb_pos] = get_bits1(gb);

  3929.         mb_has_coeffs = idx_mbmode & 1;

  3930.         if (mb_has_coeffs)

  3931.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_ICBPCY_VLC_BITS, 2);

  3932.         dst_idx = 0;

  3933.         for (i = 0; i < 6; i++) {

  3934.             s->dc_val[0][s->block_index[i]]  = 0;

  3935.             v->mb_type[0][s->block_index[i]] = 1;

  3936.             dst_idx += i >> 2;

  3937.             val = ((cbp >> (5 - i)) & 1);

  3938.             v->a_avail = v->c_avail = 0;

  3939.             if (i == 2 || i == 3 || !s->first_slice_line)

  3940.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3941.             if (i == 1 || i == 3 || s->mb_x)

  3942.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3943. 

  3944.             vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  3945.                                    (i & 4) ? v->codingset2 : v->codingset);

  3946.             if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  3947.                 continue;

  3948.             v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3949.             off  = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3950.             off += v->second_field ? ((i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0]) : 0;

  3951.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize);

  3952.             // TODO: loop filter

  3953.         }

  3954.     } else {

  3955.         s->mb_intra = v->is_intra[s->mb_x] = 0;

  3956.         s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16;

  3957.         for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 0;

  3958.         if (idx_mbmode <= 5) { // 1-MV

  3959.             dmv_x = dmv_y = 0;

  3960.             if (idx_mbmode & 1) {

  3961.                 get_mvdata_interlaced(v, &dmv_x, &dmv_y, &pred_flag);

  3962.             }

  3963.             vc1_pred_mv(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], pred_flag, 0);

  3964.             vc1_mc_1mv(v, 0);

  3965.             mb_has_coeffs = !(idx_mbmode & 2);

  3966.         } else { // 4-MV

  3967.             v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);

  3968.             for (i = 0; i < 6; i++) {

  3969.                 if (i < 4) {

  3970.                     dmv_x = dmv_y = pred_flag = 0;

  3971.                     val   = ((v->fourmvbp >> (3 - i)) & 1);

  3972.                     if (val) {

  3973.                         get_mvdata_interlaced(v, &dmv_x, &dmv_y, &pred_flag);

  3974.                     }

  3975.                     vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], pred_flag, 0);

  3976.                     vc1_mc_4mv_luma(v, i, 0);

  3977.                 } else if (i == 4)

  3978.                     vc1_mc_4mv_chroma(v, 0);

  3979.             }

  3980.             mb_has_coeffs = idx_mbmode & 1;

  3981.         }

  3982.         if (mb_has_coeffs)

  3983.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3984.         if (cbp) {

  3985.             GET_MQUANT();

  3986.         }

  3987.         s->current_picture.f.qscale_table[mb_pos] = mquant;

  3988.         if (!v->ttmbf && cbp) {

  3989.             ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  3990.         }

  3991.         dst_idx = 0;

  3992.         for (i = 0; i < 6; i++) {

  3993.             s->dc_val[0][s->block_index[i]] = 0;

  3994.             dst_idx += i >> 2;

  3995.             val = ((cbp >> (5 - i)) & 1);

  3996.             off = (i & 4) ? 0 : (i & 1) * 8 + (i & 2) * 4 * s->linesize;

  3997.             if (v->second_field)

  3998.                 off += (i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0];

  3999.             if (val) {

  4000.                 pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4001.                                          first_block, s->dest[dst_idx] + off,

  4002.                                          (i & 4) ? s->uvlinesize : s->linesize,

  4003.                                          (i & 4) && (s->flags & CODEC_FLAG_GRAY),

  4004.                                          &block_tt);

  4005.                 block_cbp |= pat << (i << 2);

  4006.                 if (!v->ttmbf && ttmb < 8) ttmb = -1;

  4007.                 first_block = 0;

  4008.             }

  4009.         }

  4010.     }

  4011.     if (s->mb_x == s->mb_width - 1)

  4012.         memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride);

  4013.     return 0;

  4014. }

  4015. 

  4016. /** Decode one B-frame MB (in Main profile)

  4017.  */

  4018. static void vc1_decode_b_mb(VC1Context *v)

  4019. {

  4020.     MpegEncContext *s = &v->s;

  4021.     GetBitContext *gb = &s->gb;

  4022.     int i, j;

  4023.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4024.     int cbp = 0; /* cbp decoding stuff */

  4025.     int mqdiff, mquant; /* MB quantization */

  4026.     int ttmb = v->ttfrm; /* MB Transform type */

  4027.     int mb_has_coeffs = 0; /* last_flag */

  4028.     int index, index1; /* LUT indexes */

  4029.     int val, sign; /* temp values */

  4030.     int first_block = 1;

  4031.     int dst_idx, off;

  4032.     int skipped, direct;

  4033.     int dmv_x[2], dmv_y[2];

  4034.     int bmvtype = BMV_TYPE_BACKWARD;

  4035. 

  4036.     mquant      = v->pq; /* lossy initialization */

  4037.     s->mb_intra = 0;

  4038. 

  4039.     if (v->dmb_is_raw)

  4040.         direct = get_bits1(gb);

  4041.     else

  4042.         direct = v->direct_mb_plane[mb_pos];

  4043.     if (v->skip_is_raw)

  4044.         skipped = get_bits1(gb);

  4045.     else

  4046.         skipped = v->s.mbskip_table[mb_pos];

  4047. 

  4048.     dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;

  4049.     for (i = 0; i < 6; i++) {

  4050.         v->mb_type[0][s->block_index[i]] = 0;

  4051.         s->dc_val[0][s->block_index[i]]  = 0;

  4052.     }

  4053.     s->current_picture.f.qscale_table[mb_pos] = 0;

  4054. 

  4055.     if (!direct) {

  4056.         if (!skipped) {

  4057.             GET_MVDATA(dmv_x[0], dmv_y[0]);

  4058.             dmv_x[1] = dmv_x[0];

  4059.             dmv_y[1] = dmv_y[0];

  4060.         }

  4061.         if (skipped || !s->mb_intra) {

  4062.             bmvtype = decode012(gb);

  4063.             switch (bmvtype) {

  4064.             case 0:

  4065.                 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;

  4066.                 break;

  4067.             case 1:

  4068.                 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;

  4069.                 break;

  4070.             case 2:

  4071.                 bmvtype  = BMV_TYPE_INTERPOLATED;

  4072.                 dmv_x[0] = dmv_y[0] = 0;

  4073.             }

  4074.         }

  4075.     }

  4076.     for (i = 0; i < 6; i++)

  4077.         v->mb_type[0][s->block_index[i]] = s->mb_intra;

  4078. 

  4079.     if (skipped) {

  4080.         if (direct)

  4081.             bmvtype = BMV_TYPE_INTERPOLATED;

  4082.         vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4083.         vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4084.         return;

  4085.     }

  4086.     if (direct) {

  4087.         cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4088.         GET_MQUANT();

  4089.         s->mb_intra = 0;

  4090.         s->current_picture.f.qscale_table[mb_pos] = mquant;

  4091.         if (!v->ttmbf)

  4092.             ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4093.         dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;

  4094.         vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4095.         vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4096.     } else {

  4097.         if (!mb_has_coeffs && !s->mb_intra) {

  4098.             /* no coded blocks - effectively skipped */

  4099.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4100.             vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4101.             return;

  4102.         }

  4103.         if (s->mb_intra && !mb_has_coeffs) {

  4104.             GET_MQUANT();

  4105.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  4106.             s->ac_pred = get_bits1(gb);

  4107.             cbp = 0;

  4108.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4109.         } else {

  4110.             if (bmvtype == BMV_TYPE_INTERPOLATED) {

  4111.                 GET_MVDATA(dmv_x[0], dmv_y[0]);

  4112.                 if (!mb_has_coeffs) {

  4113.                     /* interpolated skipped block */

  4114.                     vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4115.                     vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4116.                     return;

  4117.                 }

  4118.             }

  4119.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4120.             if (!s->mb_intra) {

  4121.                 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4122.             }

  4123.             if (s->mb_intra)

  4124.                 s->ac_pred = get_bits1(gb);

  4125.             cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4126.             GET_MQUANT();

  4127.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  4128.             if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)

  4129.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4130.         }

  4131.     }

  4132.     dst_idx = 0;

  4133.     for (i = 0; i < 6; i++) {

  4134.         s->dc_val[0][s->block_index[i]] = 0;

  4135.         dst_idx += i >> 2;

  4136.         val = ((cbp >> (5 - i)) & 1);

  4137.         off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  4138.         v->mb_type[0][s->block_index[i]] = s->mb_intra;

  4139.         if (s->mb_intra) {

  4140.             /* check if prediction blocks A and C are available */

  4141.             v->a_avail = v->c_avail = 0;

  4142.             if (i == 2 || i == 3 || !s->first_slice_line)

  4143.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  4144.             if (i == 1 || i == 3 || s->mb_x)

  4145.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  4146. 

  4147.             vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  4148.                                    (i & 4) ? v->codingset2 : v->codingset);

  4149.             if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  4150.                 continue;

  4151.             v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  4152.             if (v->rangeredfrm)

  4153.                 for (j = 0; j < 64; j++)

  4154.                     s->block[i][j] <<= 1;

  4155.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  4156.         } else if (val) {

  4157.             vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4158.                                first_block, s->dest[dst_idx] + off,

  4159.                                (i & 4) ? s->uvlinesize : s->linesize,

  4160.                                (i & 4) && (s->flags & CODEC_FLAG_GRAY), NULL);

  4161.             if (!v->ttmbf && ttmb < 8)

  4162.                 ttmb = -1;

  4163.             first_block = 0;

  4164.         }

  4165.     }

  4166. }

  4167. 

  4168. /** Decode one B-frame MB (in interlaced field B picture)

  4169.  */

  4170. static void vc1_decode_b_mb_intfi(VC1Context *v)

  4171. {

  4172.     MpegEncContext *s = &v->s;

  4173.     GetBitContext *gb = &s->gb;

  4174.     int i, j;

  4175.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4176.     int cbp = 0; /* cbp decoding stuff */

  4177.     int mqdiff, mquant; /* MB quantization */

  4178.     int ttmb = v->ttfrm; /* MB Transform type */

  4179.     int mb_has_coeffs = 0; /* last_flag */

  4180.     int val; /* temp value */

  4181.     int first_block = 1;

  4182.     int dst_idx, off;

  4183.     int fwd;

  4184.     int dmv_x[2], dmv_y[2], pred_flag[2];

  4185.     int bmvtype = BMV_TYPE_BACKWARD;

  4186.     int idx_mbmode, interpmvp;

  4187. 

  4188.     mquant      = v->pq; /* Lossy initialization */

  4189.     s->mb_intra = 0;

  4190. 

  4191.     idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_IF_MBMODE_VLC_BITS, 2);

  4192.     if (idx_mbmode <= 1) { // intra MB

  4193.         s->mb_intra = v->is_intra[s->mb_x] = 1;

  4194.         s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0;

  4195.         s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0;

  4196.         s->current_picture.f.mb_type[mb_pos + v->mb_off]         = MB_TYPE_INTRA;

  4197.         GET_MQUANT();

  4198.         s->current_picture.f.qscale_table[mb_pos] = mquant;

  4199.         /* Set DC scale - y and c use the same (not sure if necessary here) */

  4200.         s->y_dc_scale = s->y_dc_scale_table[mquant];

  4201.         s->c_dc_scale = s->c_dc_scale_table[mquant];

  4202.         v->s.ac_pred  = v->acpred_plane[mb_pos] = get_bits1(gb);

  4203.         mb_has_coeffs = idx_mbmode & 1;

  4204.         if (mb_has_coeffs)

  4205.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_ICBPCY_VLC_BITS, 2);

  4206.         dst_idx = 0;

  4207.         for (i = 0; i < 6; i++) {

  4208.             s->dc_val[0][s->block_index[i]] = 0;

  4209.             dst_idx += i >> 2;

  4210.             val = ((cbp >> (5 - i)) & 1);

  4211.             v->mb_type[0][s->block_index[i]] = s->mb_intra;

  4212.             v->a_avail                       = v->c_avail = 0;

  4213.             if (i == 2 || i == 3 || !s->first_slice_line)

  4214.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  4215.             if (i == 1 || i == 3 || s->mb_x)

  4216.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  4217. 

  4218.             vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  4219.                                    (i & 4) ? v->codingset2 : v->codingset);

  4220.             if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  4221.                 continue;

  4222.             v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  4223.             if (v->rangeredfrm)

  4224.                 for (j = 0; j < 64; j++)

  4225.                     s->block[i][j] <<= 1;

  4226.             off  = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  4227.             off += v->second_field ? ((i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0]) : 0;

  4228.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize);

  4229.             // TODO: yet to perform loop filter

  4230.         }

  4231.     } else {

  4232.         s->mb_intra = v->is_intra[s->mb_x] = 0;

  4233.         s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16;

  4234.         for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 0;

  4235.         if (v->fmb_is_raw)

  4236.             fwd = v->forward_mb_plane[mb_pos] = get_bits1(gb);

  4237.         else

  4238.             fwd = v->forward_mb_plane[mb_pos];

  4239.         if (idx_mbmode <= 5) { // 1-MV

  4240.             dmv_x[0]     = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;

  4241.             pred_flag[0] = pred_flag[1] = 0;

  4242.             if (fwd)

  4243.                 bmvtype = BMV_TYPE_FORWARD;

  4244.             else {

  4245.                 bmvtype = decode012(gb);

  4246.                 switch (bmvtype) {

  4247.                 case 0:

  4248.                     bmvtype = BMV_TYPE_BACKWARD;

  4249.                     break;

  4250.                 case 1:

  4251.                     bmvtype = BMV_TYPE_DIRECT;

  4252.                     break;

  4253.                 case 2:

  4254.                     bmvtype   = BMV_TYPE_INTERPOLATED;

  4255.                     interpmvp = get_bits1(gb);

  4256.                 }

  4257.             }

  4258.             v->bmvtype = bmvtype;

  4259.             if (bmvtype != BMV_TYPE_DIRECT && idx_mbmode & 1) {

  4260.                 get_mvdata_interlaced(v, &dmv_x[bmvtype == BMV_TYPE_BACKWARD], &dmv_y[bmvtype == BMV_TYPE_BACKWARD], &pred_flag[bmvtype == BMV_TYPE_BACKWARD]);

  4261.             }

  4262.             if (bmvtype == BMV_TYPE_INTERPOLATED && interpmvp) {

  4263.                 get_mvdata_interlaced(v, &dmv_x[1], &dmv_y[1], &pred_flag[1]);

  4264.             }

  4265.             if (bmvtype == BMV_TYPE_DIRECT) {

  4266.                 dmv_x[0] = dmv_y[0] = pred_flag[0] = 0;

  4267.                 dmv_x[1] = dmv_y[1] = pred_flag[0] = 0;

  4268.             }

  4269.             vc1_pred_b_mv_intfi(v, 0, dmv_x, dmv_y, 1, pred_flag);

  4270.             vc1_b_mc(v, dmv_x, dmv_y, (bmvtype == BMV_TYPE_DIRECT), bmvtype);

  4271.             mb_has_coeffs = !(idx_mbmode & 2);

  4272.         } else { // 4-MV

  4273.             if (fwd)

  4274.                 bmvtype = BMV_TYPE_FORWARD;

  4275.             v->bmvtype  = bmvtype;

  4276.             v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);

  4277.             for (i = 0; i < 6; i++) {

  4278.                 if (i < 4) {

  4279.                     dmv_x[0] = dmv_y[0] = pred_flag[0] = 0;

  4280.                     dmv_x[1] = dmv_y[1] = pred_flag[1] = 0;

  4281.                     val = ((v->fourmvbp >> (3 - i)) & 1);

  4282.                     if (val) {

  4283.                         get_mvdata_interlaced(v, &dmv_x[bmvtype == BMV_TYPE_BACKWARD],

  4284.                                                  &dmv_y[bmvtype == BMV_TYPE_BACKWARD],

  4285.                                              &pred_flag[bmvtype == BMV_TYPE_BACKWARD]);

  4286.                     }

  4287.                     vc1_pred_b_mv_intfi(v, i, dmv_x, dmv_y, 0, pred_flag);

  4288.                     vc1_mc_4mv_luma(v, i, bmvtype == BMV_TYPE_BACKWARD);

  4289.                 } else if (i == 4)

  4290.                     vc1_mc_4mv_chroma(v, bmvtype == BMV_TYPE_BACKWARD);

  4291.             }

  4292.             mb_has_coeffs = idx_mbmode & 1;

  4293.         }

  4294.         if (mb_has_coeffs)

  4295.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4296.         if (cbp) {

  4297.             GET_MQUANT();

  4298.         }

  4299.         s->current_picture.f.qscale_table[mb_pos] = mquant;

  4300.         if (!v->ttmbf && cbp) {

  4301.             ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4302.         }

  4303.         dst_idx = 0;

  4304.         for (i = 0; i < 6; i++) {

  4305.             s->dc_val[0][s->block_index[i]] = 0;

  4306.             dst_idx += i >> 2;

  4307.             val = ((cbp >> (5 - i)) & 1);

  4308.             off = (i & 4) ? 0 : (i & 1) * 8 + (i & 2) * 4 * s->linesize;

  4309.             if (v->second_field)

  4310.                 off += (i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0];

  4311.             if (val) {

  4312.                 vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4313.                                    first_block, s->dest[dst_idx] + off,

  4314.                                    (i & 4) ? s->uvlinesize : s->linesize,

  4315.                                    (i & 4) && (s->flags & CODEC_FLAG_GRAY), NULL);

  4316.                 if (!v->ttmbf && ttmb < 8)

  4317.                     ttmb = -1;

  4318.                 first_block = 0;

  4319.             }

  4320.         }

  4321.     }

  4322. }

  4323. 

  4324. /** Decode blocks of I-frame

  4325.  */

  4326. static void vc1_decode_i_blocks(VC1Context *v)

  4327. {

  4328.     int k, j;

  4329.     MpegEncContext *s = &v->s;

  4330.     int cbp, val;

  4331.     uint8_t *coded_val;

  4332.     int mb_pos;

  4333. 

  4334.     /* select codingmode used for VLC tables selection */

  4335.     switch (v->y_ac_table_index) {

  4336.     case 0:

  4337.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4338.         break;

  4339.     case 1:

  4340.         v->codingset = CS_HIGH_MOT_INTRA;

  4341.         break;

  4342.     case 2:

  4343.         v->codingset = CS_MID_RATE_INTRA;

  4344.         break;

  4345.     }

  4346. 

  4347.     switch (v->c_ac_table_index) {

  4348.     case 0:

  4349.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4350.         break;

  4351.     case 1:

  4352.         v->codingset2 = CS_HIGH_MOT_INTER;

  4353.         break;

  4354.     case 2:

  4355.         v->codingset2 = CS_MID_RATE_INTER;

  4356.         break;

  4357.     }

  4358. 

  4359.     /* Set DC scale - y and c use the same */

  4360.     s->y_dc_scale = s->y_dc_scale_table[v->pq];

  4361.     s->c_dc_scale = s->c_dc_scale_table[v->pq];

  4362. 

  4363.     //do frame decode

  4364.     s->mb_x = s->mb_y = 0;

  4365.     s->mb_intra         = 1;

  4366.     s->first_slice_line = 1;

  4367.     for (s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {

  4368.         s->mb_x = 0;

  4369.         ff_init_block_index(s);

  4370.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  4371.             uint8_t *dst[6];

  4372.             ff_update_block_index(s);

  4373.             dst[0] = s->dest[0];

  4374.             dst[1] = dst[0] + 8;

  4375.             dst[2] = s->dest[0] + s->linesize * 8;

  4376.             dst[3] = dst[2] + 8;

  4377.             dst[4] = s->dest[1];

  4378.             dst[5] = s->dest[2];

  4379.             s->dsp.clear_blocks(s->block[0]);

  4380.             mb_pos = s->mb_x + s->mb_y * s->mb_width;

  4381.             s->current_picture.f.mb_type[mb_pos]                     = MB_TYPE_INTRA;

  4382.             s->current_picture.f.qscale_table[mb_pos]                = v->pq;

  4383.             s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0;

  4384.             s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0;

  4385. 

  4386.             // do actual MB decoding and displaying

  4387.             cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);

  4388.             v->s.ac_pred = get_bits1(&v->s.gb);

  4389. 

  4390.             for (k = 0; k < 6; k++) {

  4391.                 val = ((cbp >> (5 - k)) & 1);

  4392. 

  4393.                 if (k < 4) {

  4394.                     int pred   = vc1_coded_block_pred(&v->s, k, &coded_val);

  4395.                     val        = val ^ pred;

  4396.                     *coded_val = val;

  4397.                 }

  4398.                 cbp |= val << (5 - k);

  4399. 

  4400.                 vc1_decode_i_block(v, s->block[k], k, val, (k < 4) ? v->codingset : v->codingset2);

  4401. 

  4402.                 if (k > 3 && (s->flags & CODEC_FLAG_GRAY))

  4403.                     continue;

  4404.                 v->vc1dsp.vc1_inv_trans_8x8(s->block[k]);

  4405.                 if (v->pq >= 9 && v->overlap) {

  4406.                     if (v->rangeredfrm)

  4407.                         for (j = 0; j < 64; j++)

  4408.                             s->block[k][j] <<= 1;

  4409.                     s->dsp.put_signed_pixels_clamped(s->block[k], dst[k], k & 4 ? s->uvlinesize : s->linesize);

  4410.                 } else {

  4411.                     if (v->rangeredfrm)

  4412.                         for (j = 0; j < 64; j++)

  4413.                             s->block[k][j] = (s->block[k][j] - 64) << 1;

  4414.                     s->dsp.put_pixels_clamped(s->block[k], dst[k], k & 4 ? s->uvlinesize : s->linesize);

  4415.                 }

  4416.             }

  4417. 

  4418.             if (v->pq >= 9 && v->overlap) {

  4419.                 if (s->mb_x) {

  4420.                     v->vc1dsp.vc1_h_overlap(s->dest[0], s->linesize);

  4421.                     v->vc1dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  4422.                     if (!(s->flags & CODEC_FLAG_GRAY)) {

  4423.                         v->vc1dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);

  4424.                         v->vc1dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);

  4425.                     }

  4426.                 }

  4427.                 v->vc1dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);

  4428.                 v->vc1dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  4429.                 if (!s->first_slice_line) {

  4430.                     v->vc1dsp.vc1_v_overlap(s->dest[0], s->linesize);

  4431.                     v->vc1dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);

  4432.                     if (!(s->flags & CODEC_FLAG_GRAY)) {

  4433.                         v->vc1dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);

  4434.                         v->vc1dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);

  4435.                     }

  4436.                 }

  4437.                 v->vc1dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  4438.                 v->vc1dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  4439.             }

  4440.             if (v->s.loop_filter) vc1_loop_filter_iblk(v, v->pq);

  4441. 

  4442.             if (get_bits_count(&s->gb) > v->bits) {

  4443.                 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, ER_MB_ERROR);

  4444.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n",

  4445.                        get_bits_count(&s->gb), v->bits);

  4446.                 return;

  4447.             }

  4448.         }

  4449.         if (!v->s.loop_filter)

  4450.             ff_draw_horiz_band(s, s->mb_y * 16, 16);

  4451.         else if (s->mb_y)

  4452.             ff_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);

  4453. 

  4454.         s->first_slice_line = 0;

  4455.     }

  4456.     if (v->s.loop_filter)

  4457.         ff_draw_horiz_band(s, (s->mb_height - 1) * 16, 16);

  4458.     ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, ER_MB_END);

  4459. }

  4460. 

  4461. /** Decode blocks of I-frame for advanced profile

  4462.  */

  4463. static void vc1_decode_i_blocks_adv(VC1Context *v)

  4464. {

  4465.     int k;

  4466.     MpegEncContext *s = &v->s;

  4467.     int cbp, val;

  4468.     uint8_t *coded_val;

  4469.     int mb_pos;

  4470.     int mquant = v->pq;

  4471.     int mqdiff;

  4472.     GetBitContext *gb = &s->gb;

  4473. 

  4474.     /* select codingmode used for VLC tables selection */

  4475.     switch (v->y_ac_table_index) {

  4476.     case 0:

  4477.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4478.         break;

  4479.     case 1:

  4480.         v->codingset = CS_HIGH_MOT_INTRA;

  4481.         break;

  4482.     case 2:

  4483.         v->codingset = CS_MID_RATE_INTRA;

  4484.         break;

  4485.     }

  4486. 

  4487.     switch (v->c_ac_table_index) {

  4488.     case 0:

  4489.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4490.         break;

  4491.     case 1:

  4492.         v->codingset2 = CS_HIGH_MOT_INTER;

  4493.         break;

  4494.     case 2:

  4495.         v->codingset2 = CS_MID_RATE_INTER;

  4496.         break;

  4497.     }

  4498. 

  4499.     // do frame decode

  4500.     s->mb_x             = s->mb_y = 0;

  4501.     s->mb_intra         = 1;

  4502.     s->first_slice_line = 1;

  4503.     s->mb_y             = s->start_mb_y;

  4504.     if (s->start_mb_y) {

  4505.         s->mb_x = 0;

  4506.         ff_init_block_index(s);

  4507.         memset(&s->coded_block[s->block_index[0] - s->b8_stride], 0,

  4508.                (1 + s->b8_stride) * sizeof(*s->coded_block));

  4509.     }

  4510.     for (; s->mb_y < s->end_mb_y; s->mb_y++) {

  4511.         s->mb_x = 0;

  4512.         ff_init_block_index(s);

  4513.         for (;s->mb_x < s->mb_width; s->mb_x++) {

  4514.             DCTELEM (*block)[64] = v->block[v->cur_blk_idx];

  4515.             ff_update_block_index(s);

  4516.             s->dsp.clear_blocks(block[0]);

  4517.             mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4518.             s->current_picture.f.mb_type[mb_pos + v->mb_off]                         = MB_TYPE_INTRA;

  4519.             s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;

  4520.             s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;

  4521. 

  4522.             // do actual MB decoding and displaying

  4523.             if (v->fieldtx_is_raw)

  4524.                 v->fieldtx_plane[mb_pos] = get_bits1(&v->s.gb);

  4525.             cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);

  4526.             if ( v->acpred_is_raw)

  4527.                 v->s.ac_pred = get_bits1(&v->s.gb);

  4528.             else

  4529.                 v->s.ac_pred = v->acpred_plane[mb_pos];

  4530. 

  4531.             if (v->condover == CONDOVER_SELECT && v->overflg_is_raw)

  4532.                 v->over_flags_plane[mb_pos] = get_bits1(&v->s.gb);

  4533. 

  4534.             GET_MQUANT();

  4535. 

  4536.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  4537.             /* Set DC scale - y and c use the same */

  4538.             s->y_dc_scale = s->y_dc_scale_table[mquant];

  4539.             s->c_dc_scale = s->c_dc_scale_table[mquant];

  4540. 

  4541.             for (k = 0; k < 6; k++) {

  4542.                 val = ((cbp >> (5 - k)) & 1);

  4543. 

  4544.                 if (k < 4) {

  4545.                     int pred   = vc1_coded_block_pred(&v->s, k, &coded_val);

  4546.                     val        = val ^ pred;

  4547.                     *coded_val = val;

  4548.                 }

  4549.                 cbp |= val << (5 - k);

  4550. 

  4551.                 v->a_avail = !s->first_slice_line || (k == 2 || k == 3);

  4552.                 v->c_avail = !!s->mb_x || (k == 1 || k == 3);

  4553. 

  4554.                 vc1_decode_i_block_adv(v, block[k], k, val,

  4555.                                        (k < 4) ? v->codingset : v->codingset2, mquant);

  4556. 

  4557.                 if (k > 3 && (s->flags & CODEC_FLAG_GRAY))

  4558.                     continue;

  4559.                 v->vc1dsp.vc1_inv_trans_8x8(block[k]);

  4560.             }

  4561. 

  4562.             vc1_smooth_overlap_filter_iblk(v);

  4563.             vc1_put_signed_blocks_clamped(v);

  4564.             if (v->s.loop_filter) vc1_loop_filter_iblk_delayed(v, v->pq);

  4565. 

  4566.             if (get_bits_count(&s->gb) > v->bits) {

  4567.                 // TODO: may need modification to handle slice coding

  4568.                 ff_er_add_slice(s, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);

  4569.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n",

  4570.                        get_bits_count(&s->gb), v->bits);

  4571.                 return;

  4572.             }

  4573.         }

  4574.         if (!v->s.loop_filter)

  4575.             ff_draw_horiz_band(s, s->mb_y * 16, 16);

  4576.         else if (s->mb_y)

  4577.             ff_draw_horiz_band(s, (s->mb_y-1) * 16, 16);

  4578.         s->first_slice_line = 0;

  4579.     }

  4580. 

  4581.     /* raw bottom MB row */

  4582.     s->mb_x = 0;

  4583.     ff_init_block_index(s);

  4584.     for (;s->mb_x < s->mb_width; s->mb_x++) {

  4585.         ff_update_block_index(s);

  4586.         vc1_put_signed_blocks_clamped(v);

  4587.         if (v->s.loop_filter)

  4588.             vc1_loop_filter_iblk_delayed(v, v->pq);

  4589.     }

  4590.     if (v->s.loop_filter)

  4591.         ff_draw_horiz_band(s, (s->end_mb_y-1)*16, 16);

  4592.     ff_er_add_slice(s, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,

  4593.                     (s->end_mb_y << v->field_mode) - 1, ER_MB_END);

  4594. }

  4595. 

  4596. static void vc1_decode_p_blocks(VC1Context *v)

  4597. {

  4598.     MpegEncContext *s = &v->s;

  4599.     int apply_loop_filter;

  4600. 

  4601.     /* select codingmode used for VLC tables selection */

  4602.     switch (v->c_ac_table_index) {

  4603.     case 0:

  4604.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4605.         break;

  4606.     case 1:

  4607.         v->codingset = CS_HIGH_MOT_INTRA;

  4608.         break;

  4609.     case 2:

  4610.         v->codingset = CS_MID_RATE_INTRA;

  4611.         break;

  4612.     }

  4613. 

  4614.     switch (v->c_ac_table_index) {

  4615.     case 0:

  4616.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4617.         break;

  4618.     case 1:

  4619.         v->codingset2 = CS_HIGH_MOT_INTER;

  4620.         break;

  4621.     case 2:

  4622.         v->codingset2 = CS_MID_RATE_INTER;

  4623.         break;

  4624.     }

  4625. 

  4626.     apply_loop_filter   = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY);

  4627.     s->first_slice_line = 1;

  4628.     memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);

  4629.     for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {

  4630.         s->mb_x = 0;

  4631.         ff_init_block_index(s);

  4632.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  4633.             ff_update_block_index(s);

  4634. 

  4635.             if (v->fcm == ILACE_FIELD)

  4636.                 vc1_decode_p_mb_intfi(v);

  4637.             else if (v->fcm == ILACE_FRAME)

  4638.                 vc1_decode_p_mb_intfr(v);

  4639.             else vc1_decode_p_mb(v);

  4640.             if (s->mb_y != s->start_mb_y && apply_loop_filter && v->fcm == PROGRESSIVE)

  4641.                 vc1_apply_p_loop_filter(v);

  4642.             if (get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {

  4643.                 // TODO: may need modification to handle slice coding

  4644.                 ff_er_add_slice(s, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);

  4645.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n",

  4646.                        get_bits_count(&s->gb), v->bits, s->mb_x, s->mb_y);

  4647.                 return;

  4648.             }

  4649.         }

  4650.         memmove(v->cbp_base,      v->cbp,      sizeof(v->cbp_base[0])      * s->mb_stride);

  4651.         memmove(v->ttblk_base,    v->ttblk,    sizeof(v->ttblk_base[0])    * s->mb_stride);

  4652.         memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride);

  4653.         memmove(v->luma_mv_base,  v->luma_mv,  sizeof(v->luma_mv_base[0])  * s->mb_stride);

  4654.         if (s->mb_y != s->start_mb_y) ff_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);

  4655.         s->first_slice_line = 0;

  4656.     }

  4657.     if (apply_loop_filter) {

  4658.         s->mb_x = 0;

  4659.         ff_init_block_index(s);

  4660.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  4661.             ff_update_block_index(s);

  4662.             vc1_apply_p_loop_filter(v);

  4663.         }

  4664.     }

  4665.     if (s->end_mb_y >= s->start_mb_y)

  4666.         ff_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);

  4667.     ff_er_add_slice(s, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,

  4668.                     (s->end_mb_y << v->field_mode) - 1, ER_MB_END);

  4669. }

  4670. 

  4671. static void vc1_decode_b_blocks(VC1Context *v)

  4672. {

  4673.     MpegEncContext *s = &v->s;

  4674. 

  4675.     /* select codingmode used for VLC tables selection */

  4676.     switch (v->c_ac_table_index) {

  4677.     case 0:

  4678.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4679.         break;

  4680.     case 1:

  4681.         v->codingset = CS_HIGH_MOT_INTRA;

  4682.         break;

  4683.     case 2:

  4684.         v->codingset = CS_MID_RATE_INTRA;

  4685.         break;

  4686.     }

  4687. 

  4688.     switch (v->c_ac_table_index) {

  4689.     case 0:

  4690.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4691.         break;

  4692.     case 1:

  4693.         v->codingset2 = CS_HIGH_MOT_INTER;

  4694.         break;

  4695.     case 2:

  4696.         v->codingset2 = CS_MID_RATE_INTER;

  4697.         break;

  4698.     }

  4699. 

  4700.     s->first_slice_line = 1;

  4701.     for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {

  4702.         s->mb_x = 0;

  4703.         ff_init_block_index(s);

  4704.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  4705.             ff_update_block_index(s);

  4706. 

  4707.             if (v->fcm == ILACE_FIELD)

  4708.                 vc1_decode_b_mb_intfi(v);

  4709.             else

  4710.                 vc1_decode_b_mb(v);

  4711.             if (get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {

  4712.                 // TODO: may need modification to handle slice coding

  4713.                 ff_er_add_slice(s, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);

  4714.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n",

  4715.                        get_bits_count(&s->gb), v->bits, s->mb_x, s->mb_y);

  4716.                 return;

  4717.             }

  4718.             if (v->s.loop_filter) vc1_loop_filter_iblk(v, v->pq);

  4719.         }

  4720.         if (!v->s.loop_filter)

  4721.             ff_draw_horiz_band(s, s->mb_y * 16, 16);

  4722.         else if (s->mb_y)

  4723.             ff_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);

  4724.         s->first_slice_line = 0;

  4725.     }

  4726.     if (v->s.loop_filter)

  4727.         ff_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);

  4728.     ff_er_add_slice(s, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,

  4729.                     (s->end_mb_y << v->field_mode) - 1, ER_MB_END);

  4730. }

  4731. 

  4732. static void vc1_decode_skip_blocks(VC1Context *v)

  4733. {

  4734.     MpegEncContext *s = &v->s;

  4735. 

  4736.     ff_er_add_slice(s, 0, s->start_mb_y, s->mb_width - 1, s->end_mb_y - 1, ER_MB_END);

  4737.     s->first_slice_line = 1;

  4738.     for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {

  4739.         s->mb_x = 0;

  4740.         ff_init_block_index(s);

  4741.         ff_update_block_index(s);

  4742.         memcpy(s->dest[0], s->last_picture.f.data[0] + s->mb_y * 16 * s->linesize,   s->linesize   * 16);

  4743.         memcpy(s->dest[1], s->last_picture.f.data[1] + s->mb_y *  8 * s->uvlinesize, s->uvlinesize *  8);

  4744.         memcpy(s->dest[2], s->last_picture.f.data[2] + s->mb_y *  8 * s->uvlinesize, s->uvlinesize *  8);

  4745.         ff_draw_horiz_band(s, s->mb_y * 16, 16);

  4746.         s->first_slice_line = 0;

  4747.     }

  4748.     s->pict_type = AV_PICTURE_TYPE_P;

  4749. }

  4750. 

  4751. static void vc1_decode_blocks(VC1Context *v)

  4752. {

  4753. 

  4754.     v->s.esc3_level_length = 0;

  4755.     if (v->x8_type) {

  4756.         ff_intrax8_decode_picture(&v->x8, 2*v->pq + v->halfpq, v->pq * !v->pquantizer);

  4757.     } else {

  4758.         v->cur_blk_idx     =  0;

  4759.         v->left_blk_idx    = -1;

  4760.         v->topleft_blk_idx =  1;

  4761.         v->top_blk_idx     =  2;

  4762.         switch (v->s.pict_type) {

  4763.         case AV_PICTURE_TYPE_I:

  4764.             if (v->profile == PROFILE_ADVANCED)

  4765.                 vc1_decode_i_blocks_adv(v);

  4766.             else

  4767.                 vc1_decode_i_blocks(v);

  4768.             break;

  4769.         case AV_PICTURE_TYPE_P:

  4770.             if (v->p_frame_skipped)

  4771.                 vc1_decode_skip_blocks(v);

  4772.             else

  4773.                 vc1_decode_p_blocks(v);

  4774.             break;

  4775.         case AV_PICTURE_TYPE_B:

  4776.             if (v->bi_type) {

  4777.                 if (v->profile == PROFILE_ADVANCED)

  4778.                     vc1_decode_i_blocks_adv(v);

  4779.                 else

  4780.                     vc1_decode_i_blocks(v);

  4781.             } else

  4782.                 vc1_decode_b_blocks(v);

  4783.             break;

  4784.         }

  4785.     }

  4786. }

  4787. 

  4788. #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER

  4789. 

  4790. typedef struct {

  4791.     /**

  4792.      * Transform coefficients for both sprites in 16.16 fixed point format,

  4793.      * in the order they appear in the bitstream:

  4794.      *  x scale

  4795.      *  rotation 1 (unused)

  4796.      *  x offset

  4797.      *  rotation 2 (unused)

  4798.      *  y scale

  4799.      *  y offset

  4800.      *  alpha

  4801.      */

  4802.     int coefs[2][7];

  4803. 

  4804.     int effect_type, effect_flag;

  4805.     int effect_pcount1, effect_pcount2;   ///< amount of effect parameters stored in effect_params

  4806.     int effect_params1[15], effect_params2[10]; ///< effect parameters in 16.16 fixed point format

  4807. } SpriteData;

  4808. 

  4809. static inline int get_fp_val(GetBitContext* gb)

  4810. {

  4811.     return (get_bits_long(gb, 30) - (1 << 29)) << 1;

  4812. }

  4813. 

  4814. static void vc1_sprite_parse_transform(GetBitContext* gb, int c[7])

  4815. {

  4816.     c[1] = c[3] = 0;

  4817. 

  4818.     switch (get_bits(gb, 2)) {

  4819.     case 0:

  4820.         c[0] = 1 << 16;

  4821.         c[2] = get_fp_val(gb);

  4822.         c[4] = 1 << 16;

  4823.         break;

  4824.     case 1:

  4825.         c[0] = c[4] = get_fp_val(gb);

  4826.         c[2] = get_fp_val(gb);

  4827.         break;

  4828.     case 2:

  4829.         c[0] = get_fp_val(gb);

  4830.         c[2] = get_fp_val(gb);

  4831.         c[4] = get_fp_val(gb);

  4832.         break;

  4833.     case 3:

  4834.         c[0] = get_fp_val(gb);

  4835.         c[1] = get_fp_val(gb);

  4836.         c[2] = get_fp_val(gb);

  4837.         c[3] = get_fp_val(gb);

  4838.         c[4] = get_fp_val(gb);

  4839.         break;

  4840.     }

  4841.     c[5] = get_fp_val(gb);

  4842.     if (get_bits1(gb))

  4843.         c[6] = get_fp_val(gb);

  4844.     else

  4845.         c[6] = 1 << 16;

  4846. }

  4847. 

  4848. static void vc1_parse_sprites(VC1Context *v, GetBitContext* gb, SpriteData* sd)

  4849. {

  4850.     AVCodecContext *avctx = v->s.avctx;

  4851.     int sprite, i;

  4852. 

  4853.     for (sprite = 0; sprite <= v->two_sprites; sprite++) {

  4854.         vc1_sprite_parse_transform(gb, sd->coefs[sprite]);

  4855.         if (sd->coefs[sprite][1] || sd->coefs[sprite][3])

  4856.             av_log_ask_for_sample(avctx, "Rotation coefficients are not zero");

  4857.         av_log(avctx, AV_LOG_DEBUG, sprite ? "S2:" : "S1:");

  4858.         for (i = 0; i < 7; i++)

  4859.             av_log(avctx, AV_LOG_DEBUG, " %d.%.3d",

  4860.                    sd->coefs[sprite][i] / (1<<16),

  4861.                    (abs(sd->coefs[sprite][i]) & 0xFFFF) * 1000 / (1 << 16));

  4862.         av_log(avctx, AV_LOG_DEBUG, "\n");

  4863.     }

  4864. 

  4865.     skip_bits(gb, 2);

  4866.     if (sd->effect_type = get_bits_long(gb, 30)) {

  4867.         switch (sd->effect_pcount1 = get_bits(gb, 4)) {

  4868.         case 7:

  4869.             vc1_sprite_parse_transform(gb, sd->effect_params1);

  4870.             break;

  4871.         case 14:

  4872.             vc1_sprite_parse_transform(gb, sd->effect_params1);

  4873.             vc1_sprite_parse_transform(gb, sd->effect_params1 + 7);

  4874.             break;

  4875.         default:

  4876.             for (i = 0; i < sd->effect_pcount1; i++)

  4877.                 sd->effect_params1[i] = get_fp_val(gb);

  4878.         }

  4879.         if (sd->effect_type != 13 || sd->effect_params1[0] != sd->coefs[0][6]) {

  4880.             // effect 13 is simple alpha blending and matches the opacity above

  4881.             av_log(avctx, AV_LOG_DEBUG, "Effect: %d; params: ", sd->effect_type);

  4882.             for (i = 0; i < sd->effect_pcount1; i++)

  4883.                 av_log(avctx, AV_LOG_DEBUG, " %d.%.2d",

  4884.                        sd->effect_params1[i] / (1 << 16),

  4885.                        (abs(sd->effect_params1[i]) & 0xFFFF) * 1000 / (1 << 16));

  4886.             av_log(avctx, AV_LOG_DEBUG, "\n");

  4887.         }

  4888. 

  4889.         sd->effect_pcount2 = get_bits(gb, 16);

  4890.         if (sd->effect_pcount2 > 10) {

  4891.             av_log(avctx, AV_LOG_ERROR, "Too many effect parameters\n");

  4892.             return;

  4893.         } else if (sd->effect_pcount2) {

  4894.             i = -1;

  4895.             av_log(avctx, AV_LOG_DEBUG, "Effect params 2: ");

  4896.             while (++i < sd->effect_pcount2) {

  4897.                 sd->effect_params2[i] = get_fp_val(gb);

  4898.                 av_log(avctx, AV_LOG_DEBUG, " %d.%.2d",

  4899.                        sd->effect_params2[i] / (1 << 16),

  4900.                        (abs(sd->effect_params2[i]) & 0xFFFF) * 1000 / (1 << 16));

  4901.             }

  4902.             av_log(avctx, AV_LOG_DEBUG, "\n");

  4903.         }

  4904.     }

  4905.     if (sd->effect_flag = get_bits1(gb))

  4906.         av_log(avctx, AV_LOG_DEBUG, "Effect flag set\n");

  4907. 

  4908.     if (get_bits_count(gb) >= gb->size_in_bits +

  4909.        (avctx->codec_id == CODEC_ID_WMV3IMAGE ? 64 : 0))

  4910.         av_log(avctx, AV_LOG_ERROR, "Buffer overrun\n");

  4911.     if (get_bits_count(gb) < gb->size_in_bits - 8)

  4912.         av_log(avctx, AV_LOG_WARNING, "Buffer not fully read\n");

  4913. }

  4914. 

  4915. static void vc1_draw_sprites(VC1Context *v, SpriteData* sd)

  4916. {

  4917.     int i, plane, row, sprite;

  4918.     int sr_cache[2][2] = { { -1, -1 }, { -1, -1 } };

  4919.     uint8_t* src_h[2][2];

  4920.     int xoff[2], xadv[2], yoff[2], yadv[2], alpha;

  4921.     int ysub[2];

  4922.     MpegEncContext *s = &v->s;

  4923. 

  4924.     for (i = 0; i < 2; i++) {

  4925.         xoff[i] = av_clip(sd->coefs[i][2], 0, v->sprite_width-1 << 16);

  4926.         xadv[i] = sd->coefs[i][0];

  4927.         if (xadv[i] != 1<<16 || (v->sprite_width << 16) - (v->output_width << 16) - xoff[i])

  4928.             xadv[i] = av_clip(xadv[i], 0, ((v->sprite_width<<16) - xoff[i] - 1) / v->output_width);

  4929. 

  4930.         yoff[i] = av_clip(sd->coefs[i][5], 0, v->sprite_height-1 << 16);

  4931.         yadv[i] = av_clip(sd->coefs[i][4], 0, ((v->sprite_height << 16) - yoff[i]) / v->output_height);

  4932.     }

  4933.     alpha = av_clip(sd->coefs[1][6], 0, (1<<16) - 1);

  4934. 

  4935.     for (plane = 0; plane < (s->flags&CODEC_FLAG_GRAY ? 1 : 3); plane++) {

  4936.         int width = v->output_width>>!!plane;

  4937. 

  4938.         for (row = 0; row < v->output_height>>!!plane; row++) {

  4939.             uint8_t *dst = v->sprite_output_frame.data[plane] +

  4940.                            v->sprite_output_frame.linesize[plane] * row;

  4941. 

  4942.             for (sprite = 0; sprite <= v->two_sprites; sprite++) {

  4943.                 uint8_t *iplane = s->current_picture.f.data[plane];

  4944.                 int      iline  = s->current_picture.f.linesize[plane];

  4945.                 int      ycoord = yoff[sprite] + yadv[sprite] * row;

  4946.                 int      yline  = ycoord >> 16;

  4947.                 ysub[sprite] = ycoord & 0xFFFF;

  4948.                 if (sprite) {

  4949.                     iplane = s->last_picture.f.data[plane];

  4950.                     iline  = s->last_picture.f.linesize[plane];

  4951.                 }

  4952.                 if (!(xoff[sprite] & 0xFFFF) && xadv[sprite] == 1 << 16) {

  4953.                         src_h[sprite][0] = iplane + (xoff[sprite] >> 16) +  yline      * iline;

  4954.                     if (ysub[sprite])

  4955.                         src_h[sprite][1] = iplane + (xoff[sprite] >> 16) + FFMIN(yline + 1, (v->sprite_height>>!!plane)-1) * iline;

  4956.                 } else {

  4957.                     if (sr_cache[sprite][0] != yline) {

  4958.                         if (sr_cache[sprite][1] == yline) {

  4959.                             FFSWAP(uint8_t*, v->sr_rows[sprite][0], v->sr_rows[sprite][1]);

  4960.                             FFSWAP(int,        sr_cache[sprite][0],   sr_cache[sprite][1]);

  4961.                         } else {

  4962.                             v->vc1dsp.sprite_h(v->sr_rows[sprite][0], iplane + yline * iline, xoff[sprite], xadv[sprite], width);

  4963.                             sr_cache[sprite][0] = yline;

  4964.                         }

  4965.                     }

  4966.                     if (ysub[sprite] && sr_cache[sprite][1] != yline + 1) {

  4967.                         v->vc1dsp.sprite_h(v->sr_rows[sprite][1], iplane + FFMIN(yline + 1, (v->sprite_height>>!!plane)-1) * iline, xoff[sprite], xadv[sprite], width);

  4968.                         sr_cache[sprite][1] = yline + 1;

  4969.                     }

  4970.                     src_h[sprite][0] = v->sr_rows[sprite][0];

  4971.                     src_h[sprite][1] = v->sr_rows[sprite][1];

  4972.                 }

  4973.             }

  4974. 

  4975.             if (!v->two_sprites) {

  4976.                 if (ysub[0]) {

  4977.                     v->vc1dsp.sprite_v_single(dst, src_h[0][0], src_h[0][1], ysub[0], width);

  4978.                 } else {

  4979.                     memcpy(dst, src_h[0][0], width);

  4980.                 }

  4981.             } else {

  4982.                 if (ysub[0] && ysub[1]) {

  4983.                     v->vc1dsp.sprite_v_double_twoscale(dst, src_h[0][0], src_h[0][1], ysub[0],

  4984.                                                        src_h[1][0], src_h[1][1], ysub[1], alpha, width);

  4985.                 } else if (ysub[0]) {

  4986.                     v->vc1dsp.sprite_v_double_onescale(dst, src_h[0][0], src_h[0][1], ysub[0],

  4987.                                                        src_h[1][0], alpha, width);

  4988.                 } else if (ysub[1]) {

  4989.                     v->vc1dsp.sprite_v_double_onescale(dst, src_h[1][0], src_h[1][1], ysub[1],

  4990.                                                        src_h[0][0], (1<<16)-1-alpha, width);

  4991.                 } else {

  4992.                     v->vc1dsp.sprite_v_double_noscale(dst, src_h[0][0], src_h[1][0], alpha, width);

  4993.                 }

  4994.             }

  4995.         }

  4996. 

  4997.         if (!plane) {

  4998.             for (i = 0; i < 2; i++) {

  4999.                 xoff[i] >>= 1;

  5000.                 yoff[i] >>= 1;

  5001.             }

  5002.         }

  5003. 

  5004.     }

  5005. }

  5006. 

  5007. 

  5008. static int vc1_decode_sprites(VC1Context *v, GetBitContext* gb)

  5009. {

  5010.     MpegEncContext *s     = &v->s;

  5011.     AVCodecContext *avctx = s->avctx;

  5012.     SpriteData sd;

  5013. 

  5014.     vc1_parse_sprites(v, gb, &sd);

  5015. 

  5016.     if (!s->current_picture.f.data[0]) {

  5017.         av_log(avctx, AV_LOG_ERROR, "Got no sprites\n");

  5018.         return -1;

  5019.     }

  5020. 

  5021.     if (v->two_sprites && (!s->last_picture_ptr || !s->last_picture.f.data[0])) {

  5022.         av_log(avctx, AV_LOG_WARNING, "Need two sprites, only got one\n");

  5023.         v->two_sprites = 0;

  5024.     }

  5025. 

  5026.     if (v->sprite_output_frame.data[0])

  5027.         avctx->release_buffer(avctx, &v->sprite_output_frame);

  5028. 

  5029.     v->sprite_output_frame.buffer_hints = FF_BUFFER_HINTS_VALID;

  5030.     v->sprite_output_frame.reference = 0;

  5031.     if (avctx->get_buffer(avctx, &v->sprite_output_frame) < 0) {

  5032.         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");

  5033.         return -1;

  5034.     }

  5035. 

  5036.     vc1_draw_sprites(v, &sd);

  5037. 

  5038.     return 0;

  5039. }

  5040. 

  5041. static void vc1_sprite_flush(AVCodecContext *avctx)

  5042. {

  5043.     VC1Context *v     = avctx->priv_data;

  5044.     MpegEncContext *s = &v->s;

  5045.     AVFrame *f = &s->current_picture.f;

  5046.     int plane, i;

  5047. 

  5048.     /* Windows Media Image codecs have a convergence interval of two keyframes.

  5049.        Since we can't enforce it, clear to black the missing sprite. This is

  5050.        wrong but it looks better than doing nothing. */

  5051. 

  5052.     if (f->data[0])

  5053.         for (plane = 0; plane < (s->flags&CODEC_FLAG_GRAY ? 1 : 3); plane++)

  5054.             for (i = 0; i < v->sprite_height>>!!plane; i++)

  5055.                 memset(f->data[plane] + i * f->linesize[plane],

  5056.                        plane ? 128 : 0, f->linesize[plane]);

  5057. }

  5058. 

  5059. #endif

  5060. 

  5061. static av_cold int vc1_decode_init_alloc_tables(VC1Context *v)

  5062. {

  5063.     MpegEncContext *s = &v->s;

  5064.     int i;

  5065. 

  5066.     /* Allocate mb bitplanes */

  5067.     v->mv_type_mb_plane = av_malloc (s->mb_stride * s->mb_height);

  5068.     v->direct_mb_plane  = av_malloc (s->mb_stride * s->mb_height);

  5069.     v->forward_mb_plane = av_malloc (s->mb_stride * s->mb_height);

  5070.     v->fieldtx_plane    = av_mallocz(s->mb_stride * s->mb_height);

  5071.     v->acpred_plane     = av_malloc (s->mb_stride * s->mb_height);

  5072.     v->over_flags_plane = av_malloc (s->mb_stride * s->mb_height);

  5073. 

  5074.     v->n_allocated_blks = s->mb_width + 2;

  5075.     v->block            = av_malloc(sizeof(*v->block) * v->n_allocated_blks);

  5076.     v->cbp_base         = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);

  5077.     v->cbp              = v->cbp_base + s->mb_stride;

  5078.     v->ttblk_base       = av_malloc(sizeof(v->ttblk_base[0]) * 2 * s->mb_stride);

  5079.     v->ttblk            = v->ttblk_base + s->mb_stride;

  5080.     v->is_intra_base    = av_mallocz(sizeof(v->is_intra_base[0]) * 2 * s->mb_stride);

  5081.     v->is_intra         = v->is_intra_base + s->mb_stride;

  5082.     v->luma_mv_base     = av_malloc(sizeof(v->luma_mv_base[0]) * 2 * s->mb_stride);

  5083.     v->luma_mv          = v->luma_mv_base + s->mb_stride;

  5084. 

  5085.     /* allocate block type info in that way so it could be used with s->block_index[] */

  5086.     v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5087.     v->mb_type[0]   = v->mb_type_base + s->b8_stride + 1;

  5088.     v->mb_type[1]   = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;

  5089.     v->mb_type[2]   = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);

  5090. 

  5091.     /* allocate memory to store block level MV info */

  5092.     v->blk_mv_type_base = av_mallocz(     s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5093.     v->blk_mv_type      = v->blk_mv_type_base + s->b8_stride + 1;

  5094.     v->mv_f_base        = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2));

  5095.     v->mv_f[0]          = v->mv_f_base + s->b8_stride + 1;

  5096.     v->mv_f[1]          = v->mv_f[0] + (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5097.     v->mv_f_last_base   = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2));

  5098.     v->mv_f_last[0]     = v->mv_f_last_base + s->b8_stride + 1;

  5099.     v->mv_f_last[1]     = v->mv_f_last[0] + (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5100.     v->mv_f_next_base   = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2));

  5101.     v->mv_f_next[0]     = v->mv_f_next_base + s->b8_stride + 1;

  5102.     v->mv_f_next[1]     = v->mv_f_next[0] + (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5103. 

  5104.     /* Init coded blocks info */

  5105.     if (v->profile == PROFILE_ADVANCED) {

  5106. //        if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)

  5107. //            return -1;

  5108. //        if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)

  5109. //            return -1;

  5110.     }

  5111. 

  5112.     ff_intrax8_common_init(&v->x8,s);

  5113. 

  5114.     if (s->avctx->codec_id == CODEC_ID_WMV3IMAGE || s->avctx->codec_id == CODEC_ID_VC1IMAGE) {

  5115.         for (i = 0; i < 4; i++)

  5116.             if (!(v->sr_rows[i >> 1][i & 1] = av_malloc(v->output_width))) return -1;

  5117.     }

  5118. 

  5119.     if (!v->mv_type_mb_plane || !v->direct_mb_plane || !v->acpred_plane || !v->over_flags_plane ||

  5120.         !v->block || !v->cbp_base || !v->ttblk_base || !v->is_intra_base || !v->luma_mv_base ||

  5121.         !v->mb_type_base)

  5122.             return -1;

  5123. 

  5124.     return 0;

  5125. }

  5126. 

  5127. /** Initialize a VC1/WMV3 decoder

  5128.  * @todo TODO: Handle VC-1 IDUs (Transport level?)

  5129.  * @todo TODO: Decypher remaining bits in extra_data

  5130.  */

  5131. static av_cold int vc1_decode_init(AVCodecContext *avctx)

  5132. {

  5133.     VC1Context *v = avctx->priv_data;

  5134.     MpegEncContext *s = &v->s;

  5135.     GetBitContext gb;

  5136.     int i;

  5137. 

  5138.     /* save the container output size for WMImage */

  5139.     v->output_width  = avctx->width;

  5140.     v->output_height = avctx->height;

  5141. 

  5142.     if (!avctx->extradata_size || !avctx->extradata)

  5143.         return -1;

  5144.     if (!(avctx->flags & CODEC_FLAG_GRAY))

  5145.         avctx->pix_fmt = avctx->get_format(avctx, avctx->codec->pix_fmts);

  5146.     else

  5147.         avctx->pix_fmt = PIX_FMT_GRAY8;

  5148.     avctx->hwaccel = ff_find_hwaccel(avctx->codec->id, avctx->pix_fmt);

  5149.     v->s.avctx = avctx;

  5150.     avctx->flags |= CODEC_FLAG_EMU_EDGE;

  5151.     v->s.flags   |= CODEC_FLAG_EMU_EDGE;

  5152. 

  5153.     if (avctx->idct_algo == FF_IDCT_AUTO) {

  5154.         avctx->idct_algo = FF_IDCT_WMV2;

  5155.     }

  5156. 

  5157.     if (ff_vc1_init_common(v) < 0)

  5158.         return -1;

  5159.     ff_vc1dsp_init(&v->vc1dsp);

  5160. 

  5161.     if (avctx->codec_id == CODEC_ID_WMV3 || avctx->codec_id == CODEC_ID_WMV3IMAGE) {

  5162.         int count = 0;

  5163. 

  5164.         // looks like WMV3 has a sequence header stored in the extradata

  5165.         // advanced sequence header may be before the first frame

  5166.         // the last byte of the extradata is a version number, 1 for the

  5167.         // samples we can decode

  5168. 

  5169.         init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);

  5170. 

  5171.         if (ff_vc1_decode_sequence_header(avctx, v, &gb) < 0)

  5172.           return -1;

  5173. 

  5174.         count = avctx->extradata_size*8 - get_bits_count(&gb);

  5175.         if (count > 0) {

  5176.             av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",

  5177.                    count, get_bits(&gb, count));

  5178.         } else if (count < 0) {

  5179.             av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);

  5180.         }

  5181.     } else { // VC1/WVC1/WVP2

  5182.         const uint8_t *start = avctx->extradata;

  5183.         uint8_t *end = avctx->extradata + avctx->extradata_size;

  5184.         const uint8_t *next;

  5185.         int size, buf2_size;

  5186.         uint8_t *buf2 = NULL;

  5187.         int seq_initialized = 0, ep_initialized = 0;

  5188. 

  5189.         if (avctx->extradata_size < 16) {

  5190.             av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);

  5191.             return -1;

  5192.         }

  5193. 

  5194.         buf2  = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5195.         start = find_next_marker(start, end); // in WVC1 extradata first byte is its size, but can be 0 in mkv

  5196.         next  = start;

  5197.         for (; next < end; start = next) {

  5198.             next = find_next_marker(start + 4, end);

  5199.             size = next - start - 4;

  5200.             if (size <= 0)

  5201.                 continue;

  5202.             buf2_size = vc1_unescape_buffer(start + 4, size, buf2);

  5203.             init_get_bits(&gb, buf2, buf2_size * 8);

  5204.             switch (AV_RB32(start)) {

  5205.             case VC1_CODE_SEQHDR:

  5206.                 if (ff_vc1_decode_sequence_header(avctx, v, &gb) < 0) {

  5207.                     av_free(buf2);

  5208.                     return -1;

  5209.                 }

  5210.                 seq_initialized = 1;

  5211.                 break;

  5212.             case VC1_CODE_ENTRYPOINT:

  5213.                 if (ff_vc1_decode_entry_point(avctx, v, &gb) < 0) {

  5214.                     av_free(buf2);

  5215.                     return -1;

  5216.                 }

  5217.                 ep_initialized = 1;

  5218.                 break;

  5219.             }

  5220.         }

  5221.         av_free(buf2);

  5222.         if (!seq_initialized || !ep_initialized) {

  5223.             av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");

  5224.             return -1;

  5225.         }

  5226.         v->res_sprite = (avctx->codec_tag == MKTAG('W','V','P','2'));

  5227.     }

  5228. 

  5229.     avctx->profile = v->profile;

  5230.     if (v->profile == PROFILE_ADVANCED)

  5231.         avctx->level = v->level;

  5232. 

  5233.     avctx->has_b_frames = !!avctx->max_b_frames;

  5234. 

  5235.     s->mb_width  = (avctx->coded_width  + 15) >> 4;

  5236.     s->mb_height = (avctx->coded_height + 15) >> 4;

  5237. 

  5238.     if (v->profile == PROFILE_ADVANCED || v->res_fasttx) {

  5239.         for (i = 0; i < 64; i++) {

  5240. #define transpose(x) ((x >> 3) | ((x & 7) << 3))

  5241.             v->zz_8x8[0][i] = transpose(ff_wmv1_scantable[0][i]);

  5242.             v->zz_8x8[1][i] = transpose(ff_wmv1_scantable[1][i]);

  5243.             v->zz_8x8[2][i] = transpose(ff_wmv1_scantable[2][i]);

  5244.             v->zz_8x8[3][i] = transpose(ff_wmv1_scantable[3][i]);

  5245.             v->zzi_8x8[i] = transpose(ff_vc1_adv_interlaced_8x8_zz[i]);

  5246.         }

  5247.         v->left_blk_sh = 0;

  5248.         v->top_blk_sh  = 3;

  5249.     } else {

  5250.         memcpy(v->zz_8x8, ff_wmv1_scantable, 4*64);

  5251.         v->left_blk_sh = 3;

  5252.         v->top_blk_sh  = 0;

  5253.     }

  5254. 

  5255.     if (avctx->codec_id == CODEC_ID_WMV3IMAGE || avctx->codec_id == CODEC_ID_VC1IMAGE) {

  5256.         v->sprite_width  = avctx->coded_width;

  5257.         v->sprite_height = avctx->coded_height;

  5258. 

  5259.         avctx->coded_width  = avctx->width  = v->output_width;

  5260.         avctx->coded_height = avctx->height = v->output_height;

  5261. 

  5262.         // prevent 16.16 overflows

  5263.         if (v->sprite_width  > 1 << 14 ||

  5264.             v->sprite_height > 1 << 14 ||

  5265.             v->output_width  > 1 << 14 ||

  5266.             v->output_height > 1 << 14) return -1;

  5267.     }

  5268.     return 0;

  5269. }

  5270. 

  5271. /** Close a VC1/WMV3 decoder

  5272.  * @warning Initial try at using MpegEncContext stuff

  5273.  */

  5274. static av_cold int vc1_decode_end(AVCodecContext *avctx)

  5275. {

  5276.     VC1Context *v = avctx->priv_data;

  5277.     int i;

  5278. 

  5279.     if ((avctx->codec_id == CODEC_ID_WMV3IMAGE || avctx->codec_id == CODEC_ID_VC1IMAGE)

  5280.         && v->sprite_output_frame.data[0])

  5281.         avctx->release_buffer(avctx, &v->sprite_output_frame);

  5282.     for (i = 0; i < 4; i++)

  5283.         av_freep(&v->sr_rows[i >> 1][i & 1]);

  5284.     av_freep(&v->hrd_rate);

  5285.     av_freep(&v->hrd_buffer);

  5286.     ff_MPV_common_end(&v->s);

  5287.     av_freep(&v->mv_type_mb_plane);

  5288.     av_freep(&v->direct_mb_plane);

  5289.     av_freep(&v->forward_mb_plane);

  5290.     av_freep(&v->fieldtx_plane);

  5291.     av_freep(&v->acpred_plane);

  5292.     av_freep(&v->over_flags_plane);

  5293.     av_freep(&v->mb_type_base);

  5294.     av_freep(&v->blk_mv_type_base);

  5295.     av_freep(&v->mv_f_base);

  5296.     av_freep(&v->mv_f_last_base);

  5297.     av_freep(&v->mv_f_next_base);

  5298.     av_freep(&v->block);

  5299.     av_freep(&v->cbp_base);

  5300.     av_freep(&v->ttblk_base);

  5301.     av_freep(&v->is_intra_base); // FIXME use v->mb_type[]

  5302.     av_freep(&v->luma_mv_base);

  5303.     ff_intrax8_common_end(&v->x8);

  5304.     return 0;

  5305. }

  5306. 

  5307. 

  5308. /** Decode a VC1/WMV3 frame

  5309.  * @todo TODO: Handle VC-1 IDUs (Transport level?)

  5310.  */

  5311. static int vc1_decode_frame(AVCodecContext *avctx, void *data,

  5312.                             int *data_size, AVPacket *avpkt)

  5313. {

  5314.     const uint8_t *buf = avpkt->data;

  5315.     int buf_size = avpkt->size, n_slices = 0, i;

  5316.     VC1Context *v = avctx->priv_data;

  5317.     MpegEncContext *s = &v->s;

  5318.     AVFrame *pict = data;

  5319.     uint8_t *buf2 = NULL;

  5320.     const uint8_t *buf_start = buf;

  5321.     int mb_height, n_slices1=-1;

  5322.     struct {

  5323.         uint8_t *buf;

  5324.         GetBitContext gb;

  5325.         int mby_start;

  5326.     } *slices = NULL, *tmp;

  5327. 

  5328.     if(s->flags & CODEC_FLAG_LOW_DELAY)

  5329.         s->low_delay = 1;

  5330. 

  5331.     /* no supplementary picture */

  5332.     if (buf_size == 0 || (buf_size == 4 && AV_RB32(buf) == VC1_CODE_ENDOFSEQ)) {

  5333.         /* special case for last picture */

  5334.         if (s->low_delay == 0 && s->next_picture_ptr) {

  5335.             *pict = s->next_picture_ptr->f;

  5336.             s->next_picture_ptr = NULL;

  5337. 

  5338.             *data_size = sizeof(AVFrame);

  5339.         }

  5340. 

  5341.         return 0;

  5342.     }

  5343. 

  5344.     if (s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) {

  5345.         if (v->profile < PROFILE_ADVANCED)

  5346.             avctx->pix_fmt = PIX_FMT_VDPAU_WMV3;

  5347.         else

  5348.             avctx->pix_fmt = PIX_FMT_VDPAU_VC1;

  5349.     }

  5350. 

  5351.     //for advanced profile we may need to parse and unescape data

  5352.     if (avctx->codec_id == CODEC_ID_VC1 || avctx->codec_id == CODEC_ID_VC1IMAGE) {

  5353.         int buf_size2 = 0;

  5354.         buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5355. 

  5356.         if (IS_MARKER(AV_RB32(buf))) { /* frame starts with marker and needs to be parsed */

  5357.             const uint8_t *start, *end, *next;

  5358.             int size;

  5359. 

  5360.             next = buf;

  5361.             for (start = buf, end = buf + buf_size; next < end; start = next) {

  5362.                 next = find_next_marker(start + 4, end);

  5363.                 size = next - start - 4;

  5364.                 if (size <= 0) continue;

  5365.                 switch (AV_RB32(start)) {

  5366.                 case VC1_CODE_FRAME:

  5367.                     if (avctx->hwaccel ||

  5368.                         s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)

  5369.                         buf_start = start;

  5370.                     buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);

  5371.                     break;

  5372.                 case VC1_CODE_FIELD: {

  5373.                     int buf_size3;

  5374.                     slices = av_realloc(slices, sizeof(*slices) * (n_slices+1));

  5375.                     if (!slices)

  5376.                         goto err;

  5377.                     slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5378.                     if (!slices[n_slices].buf)

  5379.                         goto err;

  5380.                     buf_size3 = vc1_unescape_buffer(start + 4, size,

  5381.                                                     slices[n_slices].buf);

  5382.                     init_get_bits(&slices[n_slices].gb, slices[n_slices].buf,

  5383.                                   buf_size3 << 3);

  5384.                     /* assuming that the field marker is at the exact middle,

  5385.                        hope it's correct */

  5386.                     slices[n_slices].mby_start = s->mb_height >> 1;

  5387.                     n_slices1 = n_slices - 1; // index of the last slice of the first field

  5388.                     n_slices++;

  5389.                     break;

  5390.                 }

  5391.                 case VC1_CODE_ENTRYPOINT: /* it should be before frame data */

  5392.                     buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);

  5393.                     init_get_bits(&s->gb, buf2, buf_size2 * 8);

  5394.                     ff_vc1_decode_entry_point(avctx, v, &s->gb);

  5395.                     break;

  5396.                 case VC1_CODE_SLICE: {

  5397.                     int buf_size3;

  5398.                     slices = av_realloc(slices, sizeof(*slices) * (n_slices+1));

  5399.                     if (!slices)

  5400.                         goto err;

  5401.                     slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5402.                     if (!slices[n_slices].buf)

  5403.                         goto err;

  5404.                     buf_size3 = vc1_unescape_buffer(start + 4, size,

  5405.                                                     slices[n_slices].buf);

  5406.                     init_get_bits(&slices[n_slices].gb, slices[n_slices].buf,

  5407.                                   buf_size3 << 3);

  5408.                     slices[n_slices].mby_start = get_bits(&slices[n_slices].gb, 9);

  5409.                     n_slices++;

  5410.                     break;

  5411.                 }

  5412.                 }

  5413.             }

  5414.         } else if (v->interlace && ((buf[0] & 0xC0) == 0xC0)) { /* WVC1 interlaced stores both fields divided by marker */

  5415.             const uint8_t *divider;

  5416.             int buf_size3;

  5417. 

  5418.             divider = find_next_marker(buf, buf + buf_size);

  5419.             if ((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD) {

  5420.                 av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");

  5421.                 goto err;

  5422.             } else { // found field marker, unescape second field

  5423.                 tmp = av_realloc(slices, sizeof(*slices) * (n_slices+1));

  5424.                 if (!tmp)

  5425.                     goto err;

  5426.                 slices = tmp;

  5427.                 slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5428.                 if (!slices[n_slices].buf)

  5429.                     goto err;

  5430.                 buf_size3 = vc1_unescape_buffer(divider + 4, buf + buf_size - divider - 4, slices[n_slices].buf);

  5431.                 init_get_bits(&slices[n_slices].gb, slices[n_slices].buf,

  5432.                               buf_size3 << 3);

  5433.                 slices[n_slices].mby_start = s->mb_height >> 1;

  5434.                 n_slices1 = n_slices - 1;

  5435.                 n_slices++;

  5436.             }

  5437.             buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);

  5438.         } else {

  5439.             buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);

  5440.         }

  5441.         init_get_bits(&s->gb, buf2, buf_size2*8);

  5442.     } else

  5443.         init_get_bits(&s->gb, buf, buf_size*8);

  5444. 

  5445.     if (v->res_sprite) {

  5446.         v->new_sprite  = !get_bits1(&s->gb);

  5447.         v->two_sprites =  get_bits1(&s->gb);

  5448.         /* res_sprite means a Windows Media Image stream, CODEC_ID_*IMAGE means

  5449.            we're using the sprite compositor. These are intentionally kept separate

  5450.            so you can get the raw sprites by using the wmv3 decoder for WMVP or

  5451.            the vc1 one for WVP2 */

  5452.         if (avctx->codec_id == CODEC_ID_WMV3IMAGE || avctx->codec_id == CODEC_ID_VC1IMAGE) {

  5453.             if (v->new_sprite) {

  5454.                 // switch AVCodecContext parameters to those of the sprites

  5455.                 avctx->width  = avctx->coded_width  = v->sprite_width;

  5456.                 avctx->height = avctx->coded_height = v->sprite_height;

  5457.             } else {

  5458.                 goto image;

  5459.             }

  5460.         }

  5461.     }

  5462. 

  5463.     if (s->context_initialized &&

  5464.         (s->width  != avctx->coded_width ||

  5465.          s->height != avctx->coded_height)) {

  5466.         vc1_decode_end(avctx);

  5467.     }

  5468. 

  5469.     if (!s->context_initialized) {

  5470.         if (ff_msmpeg4_decode_init(avctx) < 0 || vc1_decode_init_alloc_tables(v) < 0)

  5471.             return -1;

  5472. 

  5473.         s->low_delay = !avctx->has_b_frames || v->res_sprite;

  5474. 

  5475.         if (v->profile == PROFILE_ADVANCED) {

  5476.             s->h_edge_pos = avctx->coded_width;

  5477.             s->v_edge_pos = avctx->coded_height;

  5478.         }

  5479.     }

  5480. 

  5481.     /* We need to set current_picture_ptr before reading the header,

  5482.      * otherwise we cannot store anything in there. */

  5483.     if (s->current_picture_ptr == NULL || s->current_picture_ptr->f.data[0]) {

  5484.         int i = ff_find_unused_picture(s, 0);

  5485.         if (i < 0)

  5486.             goto err;

  5487.         s->current_picture_ptr = &s->picture[i];

  5488.     }

  5489. 

  5490.     // do parse frame header

  5491.     v->pic_header_flag = 0;

  5492.     if (v->profile < PROFILE_ADVANCED) {

  5493.         if (ff_vc1_parse_frame_header(v, &s->gb) < 0) {

  5494.             goto err;

  5495.         }

  5496.     } else {

  5497.         if (ff_vc1_parse_frame_header_adv(v, &s->gb) < 0) {

  5498.             goto err;

  5499.         }

  5500.     }

  5501. 

  5502.     if ((avctx->codec_id == CODEC_ID_WMV3IMAGE || avctx->codec_id == CODEC_ID_VC1IMAGE)

  5503.         && s->pict_type != AV_PICTURE_TYPE_I) {

  5504.         av_log(v->s.avctx, AV_LOG_ERROR, "Sprite decoder: expected I-frame\n");

  5505.         goto err;

  5506.     }

  5507. 

  5508.     // process pulldown flags

  5509.     s->current_picture_ptr->f.repeat_pict = 0;

  5510.     // Pulldown flags are only valid when 'broadcast' has been set.

  5511.     // So ticks_per_frame will be 2

  5512.     if (v->rff) {

  5513.         // repeat field

  5514.         s->current_picture_ptr->f.repeat_pict = 1;

  5515.     } else if (v->rptfrm) {

  5516.         // repeat frames

  5517.         s->current_picture_ptr->f.repeat_pict = v->rptfrm * 2;

  5518.     }

  5519. 

  5520.     // for skipping the frame

  5521.     s->current_picture.f.pict_type = s->pict_type;

  5522.     s->current_picture.f.key_frame = s->pict_type == AV_PICTURE_TYPE_I;

  5523. 

  5524.     /* skip B-frames if we don't have reference frames */

  5525.     if (s->last_picture_ptr == NULL && (s->pict_type == AV_PICTURE_TYPE_B || s->dropable)) {

  5526.         goto err;

  5527.     }

  5528.     if ((avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type == AV_PICTURE_TYPE_B) ||

  5529.         (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type != AV_PICTURE_TYPE_I) ||

  5530.          avctx->skip_frame >= AVDISCARD_ALL) {

  5531.         goto end;

  5532.     }

  5533. 

  5534.     if (s->next_p_frame_damaged) {

  5535.         if (s->pict_type == AV_PICTURE_TYPE_B)

  5536.             goto end;

  5537.         else

  5538.             s->next_p_frame_damaged = 0;

  5539.     }

  5540. 

  5541.     if (ff_MPV_frame_start(s, avctx) < 0) {

  5542.         goto err;

  5543.     }

  5544. 

  5545.     s->me.qpel_put = s->dsp.put_qpel_pixels_tab;

  5546.     s->me.qpel_avg = s->dsp.avg_qpel_pixels_tab;

  5547. 

  5548.     if ((CONFIG_VC1_VDPAU_DECODER)

  5549.         &&s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)

  5550.         ff_vdpau_vc1_decode_picture(s, buf_start, (buf + buf_size) - buf_start);

  5551.     else if (avctx->hwaccel) {

  5552.         if (avctx->hwaccel->start_frame(avctx, buf, buf_size) < 0)

  5553.             goto err;

  5554.         if (avctx->hwaccel->decode_slice(avctx, buf_start, (buf + buf_size) - buf_start) < 0)

  5555.             goto err;

  5556.         if (avctx->hwaccel->end_frame(avctx) < 0)

  5557.             goto err;

  5558.     } else {

  5559.         ff_er_frame_start(s);

  5560. 

  5561.         v->bits = buf_size * 8;

  5562.         if (v->field_mode) {

  5563.             uint8_t *tmp[2];

  5564.             s->current_picture.f.linesize[0] <<= 1;

  5565.             s->current_picture.f.linesize[1] <<= 1;

  5566.             s->current_picture.f.linesize[2] <<= 1;

  5567.             s->linesize                      <<= 1;

  5568.             s->uvlinesize                    <<= 1;

  5569.             tmp[0]          = v->mv_f_last[0];

  5570.             tmp[1]          = v->mv_f_last[1];

  5571.             v->mv_f_last[0] = v->mv_f_next[0];

  5572.             v->mv_f_last[1] = v->mv_f_next[1];

  5573.             v->mv_f_next[0] = v->mv_f[0];

  5574.             v->mv_f_next[1] = v->mv_f[1];

  5575.             v->mv_f[0] = tmp[0];

  5576.             v->mv_f[1] = tmp[1];

  5577.         }

  5578.         mb_height = s->mb_height >> v->field_mode;

  5579.         for (i = 0; i <= n_slices; i++) {

  5580.             if (i > 0 &&  slices[i - 1].mby_start >= mb_height) {

  5581.                 if(v->field_mode <= 0) {

  5582.                     av_log(v->s.avctx, AV_LOG_ERROR, "invalid end_mb_y %d\n", slices[i - 1].mby_start);

  5583.                     continue;

  5584.                 }

  5585.                 v->second_field = 1;

  5586.                 v->blocks_off   = s->mb_width  * s->mb_height << 1;

  5587.                 v->mb_off       = s->mb_stride * s->mb_height >> 1;

  5588.             } else {

  5589.                 v->second_field = 0;

  5590.                 v->blocks_off   = 0;

  5591.                 v->mb_off       = 0;

  5592.             }

  5593.             if (i) {

  5594.                 v->pic_header_flag = 0;

  5595.                 if (v->field_mode && i == n_slices1 + 2) {

  5596.                     if (ff_vc1_parse_frame_header_adv(v, &s->gb) < 0) {

  5597.                         av_log(v->s.avctx, AV_LOG_ERROR, "slice header damaged\n");

  5598.                         continue;

  5599.                     }

  5600.                 } else if (get_bits1(&s->gb)) {

  5601.                     v->pic_header_flag = 1;

  5602.                     if (ff_vc1_parse_frame_header_adv(v, &s->gb) < 0) {

  5603.                         av_log(v->s.avctx, AV_LOG_ERROR, "slice header damaged\n");

  5604.                         continue;

  5605.                     }

  5606.                 }

  5607.             }

  5608.             s->start_mb_y = (i == 0) ? 0 : FFMAX(0, slices[i-1].mby_start % mb_height);

  5609.             if (!v->field_mode || v->second_field)

  5610.                 s->end_mb_y = (i == n_slices     ) ? mb_height : FFMIN(mb_height, slices[i].mby_start % mb_height);

  5611.             else

  5612.                 s->end_mb_y = (i <= n_slices1 + 1) ? mb_height : FFMIN(mb_height, slices[i].mby_start % mb_height);

  5613.             if (s->end_mb_y <= s->start_mb_y) {

  5614.                 av_log(v->s.avctx, AV_LOG_ERROR, "end mb y %d %d invalid\n", s->end_mb_y, s->start_mb_y);

  5615.                 continue;

  5616.             }

  5617.             vc1_decode_blocks(v);

  5618.             if (i != n_slices)

  5619.                 s->gb = slices[i].gb;

  5620.         }

  5621.         if (v->field_mode) {

  5622.             v->second_field = 0;

  5623.             if (s->pict_type == AV_PICTURE_TYPE_B) {

  5624.                 memcpy(v->mv_f_base, v->mv_f_next_base,

  5625.                        2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2));

  5626.             }

  5627.             s->current_picture.f.linesize[0] >>= 1;

  5628.             s->current_picture.f.linesize[1] >>= 1;

  5629.             s->current_picture.f.linesize[2] >>= 1;

  5630.             s->linesize                      >>= 1;

  5631.             s->uvlinesize                    >>= 1;

  5632.         }

  5633. //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), s->gb.size_in_bits);

  5634. //  if (get_bits_count(&s->gb) > buf_size * 8)

  5635. //      return -1;

  5636.         if(s->error_occurred && s->pict_type == AV_PICTURE_TYPE_B)

  5637.             goto err;

  5638.         ff_er_frame_end(s);

  5639.     }

  5640. 

  5641.     ff_MPV_frame_end(s);

  5642. 

  5643.     if (avctx->codec_id == CODEC_ID_WMV3IMAGE || avctx->codec_id == CODEC_ID_VC1IMAGE) {

  5644. image:

  5645.         avctx->width  = avctx->coded_width  = v->output_width;

  5646.         avctx->height = avctx->coded_height = v->output_height;

  5647.         if (avctx->skip_frame >= AVDISCARD_NONREF)

  5648.             goto end;

  5649. #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER

  5650.         if (vc1_decode_sprites(v, &s->gb))

  5651.             goto err;

  5652. #endif

  5653.         *pict      = v->sprite_output_frame;

  5654.         *data_size = sizeof(AVFrame);

  5655.     } else {

  5656.         if (s->pict_type == AV_PICTURE_TYPE_B || s->low_delay) {

  5657.             *pict = s->current_picture_ptr->f;

  5658.         } else if (s->last_picture_ptr != NULL) {

  5659.             *pict = s->last_picture_ptr->f;

  5660.         }

  5661.         if (s->last_picture_ptr || s->low_delay) {

  5662.             *data_size = sizeof(AVFrame);

  5663.             ff_print_debug_info(s, pict);

  5664.         }

  5665.     }

  5666. 

  5667. end:

  5668.     av_free(buf2);

  5669.     for (i = 0; i < n_slices; i++)

  5670.         av_free(slices[i].buf);

  5671.     av_free(slices);

  5672.     return buf_size;

  5673. 

  5674. err:

  5675.     av_free(buf2);

  5676.     for (i = 0; i < n_slices; i++)

  5677.         av_free(slices[i].buf);

  5678.     av_free(slices);

  5679.     return -1;

  5680. }

  5681. 

  5682. 

  5683. static const AVProfile profiles[] = {

  5684.     { FF_PROFILE_VC1_SIMPLE,   "Simple"   },

  5685.     { FF_PROFILE_VC1_MAIN,     "Main"     },

  5686.     { FF_PROFILE_VC1_COMPLEX,  "Complex"  },

  5687.     { FF_PROFILE_VC1_ADVANCED, "Advanced" },

  5688.     { FF_PROFILE_UNKNOWN },

  5689. };

  5690. 

  5691. AVCodec ff_vc1_decoder = {

  5692.     .name           = "vc1",

  5693.     .type           = AVMEDIA_TYPE_VIDEO,

  5694.     .id             = CODEC_ID_VC1,

  5695.     .priv_data_size = sizeof(VC1Context),

  5696.     .init           = vc1_decode_init,

  5697.     .close          = vc1_decode_end,

  5698.     .decode         = vc1_decode_frame,

  5699.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY,

  5700.     .long_name      = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),

  5701.     .pix_fmts       = ff_hwaccel_pixfmt_list_420,

  5702.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  5703. };

  5704. 

  5705. #if CONFIG_WMV3_DECODER

  5706. AVCodec ff_wmv3_decoder = {

  5707.     .name           = "wmv3",

  5708.     .type           = AVMEDIA_TYPE_VIDEO,

  5709.     .id             = CODEC_ID_WMV3,

  5710.     .priv_data_size = sizeof(VC1Context),

  5711.     .init           = vc1_decode_init,

  5712.     .close          = vc1_decode_end,

  5713.     .decode         = vc1_decode_frame,

  5714.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY,

  5715.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),

  5716.     .pix_fmts       = ff_hwaccel_pixfmt_list_420,

  5717.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  5718. };

  5719. #endif

  5720. 

  5721. #if CONFIG_WMV3_VDPAU_DECODER

  5722. AVCodec ff_wmv3_vdpau_decoder = {

  5723.     .name           = "wmv3_vdpau",

  5724.     .type           = AVMEDIA_TYPE_VIDEO,

  5725.     .id             = CODEC_ID_WMV3,

  5726.     .priv_data_size = sizeof(VC1Context),

  5727.     .init           = vc1_decode_init,

  5728.     .close          = vc1_decode_end,

  5729.     .decode         = vc1_decode_frame,

  5730.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,

  5731.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9 VDPAU"),

  5732.     .pix_fmts       = (const enum PixelFormat[]){ PIX_FMT_VDPAU_WMV3, PIX_FMT_NONE },

  5733.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  5734. };

  5735. #endif

  5736. 

  5737. #if CONFIG_VC1_VDPAU_DECODER

  5738. AVCodec ff_vc1_vdpau_decoder = {

  5739.     .name           = "vc1_vdpau",

  5740.     .type           = AVMEDIA_TYPE_VIDEO,

  5741.     .id             = CODEC_ID_VC1,

  5742.     .priv_data_size = sizeof(VC1Context),

  5743.     .init           = vc1_decode_init,

  5744.     .close          = vc1_decode_end,

  5745.     .decode         = vc1_decode_frame,

  5746.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,

  5747.     .long_name      = NULL_IF_CONFIG_SMALL("SMPTE VC-1 VDPAU"),

  5748.     .pix_fmts       = (const enum PixelFormat[]){ PIX_FMT_VDPAU_VC1, PIX_FMT_NONE },

  5749.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  5750. };

  5751. #endif

  5752. 

  5753. #if CONFIG_WMV3IMAGE_DECODER

  5754. AVCodec ff_wmv3image_decoder = {

  5755.     .name           = "wmv3image",

  5756.     .type           = AVMEDIA_TYPE_VIDEO,

  5757.     .id             = CODEC_ID_WMV3IMAGE,

  5758.     .priv_data_size = sizeof(VC1Context),

  5759.     .init           = vc1_decode_init,

  5760.     .close          = vc1_decode_end,

  5761.     .decode         = vc1_decode_frame,

  5762.     .capabilities   = CODEC_CAP_DR1,

  5763.     .flush          = vc1_sprite_flush,

  5764.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9 Image"),

  5765.     .pix_fmts       = ff_pixfmt_list_420

  5766. };

  5767. #endif

  5768. 

  5769. #if CONFIG_VC1IMAGE_DECODER

  5770. AVCodec ff_vc1image_decoder = {

  5771.     .name           = "vc1image",

  5772.     .type           = AVMEDIA_TYPE_VIDEO,

  5773.     .id             = CODEC_ID_VC1IMAGE,

  5774.     .priv_data_size = sizeof(VC1Context),

  5775.     .init           = vc1_decode_init,

  5776.     .close          = vc1_decode_end,

  5777.     .decode         = vc1_decode_frame,

  5778.     .capabilities   = CODEC_CAP_DR1,

  5779.     .flush          = vc1_sprite_flush,

  5780.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9 Image v2"),

  5781.     .pix_fmts       = ff_pixfmt_list_420

  5782. };

  5783. #endif